refactor: decouple from site-specific CDN references #5

Merged
kfj001 merged 11 commits from feature/no_mention_of_pornhub into master 2026-08-30 17:07:52 -07:00
9 changed files with 337 additions and 100 deletions
Showing only changes of commit 4357ce4094 - Show all commits

No files matched your search

+7 -1
View File
@@ -11,7 +11,13 @@
Review

[SHOULD-FIX] F3
A tool-local settings file is committed to the shared repo and now carries absolute local paths to the author's machine: 'Bash(node /Users/kyle/.claude/jobs/1462df8d/tmp/transform2.mjs)' and 'Bash(open -a "Microsoft Edge" ...)'. These are machine-specific and should not live in the shared codebase; .claude/settings.local.json is typically user-scoped and gitignored.
Suggested fix: Drop .claude/settings.local.json from the PR (it is unrelated to the refactor) and/or add it to .gitignore so per-machine tool settings never get committed.

[SHOULD-FIX] F3 A tool-local settings file is committed to the shared repo and now carries absolute local paths to the author's machine: 'Bash(node /Users/kyle/.claude/jobs/1462df8d/tmp/transform2.mjs)' and 'Bash(open -a "Microsoft Edge" ...)'. These are machine-specific and should not live in the shared codebase; .claude/settings.local.json is typically user-scoped and gitignored. **Suggested fix:** Drop .claude/settings.local.json from the PR (it is unrelated to the refactor) and/or add it to .gitignore so per-machine tool settings never get committed.
Review

[SHOULD-FIX] F3
A tool-local settings file is committed to the shared repo and now carries absolute local paths to the author's machine: 'Bash(node /Users/kyle/.claude/jobs/1462df8d/tmp/transform2.mjs)' and 'Bash(open -a "Microsoft Edge" ...)'. These are machine-specific and should not live in the shared codebase; .claude/settings.local.json is typically user-scoped and gitignored.
Suggested fix: Drop .claude/settings.local.json from the PR (it is unrelated to the refactor) and/or add it to .gitignore so per-machine tool settings never get committed.

[SHOULD-FIX] F3 A tool-local settings file is committed to the shared repo and now carries absolute local paths to the author's machine: 'Bash(node /Users/kyle/.claude/jobs/1462df8d/tmp/transform2.mjs)' and 'Bash(open -a "Microsoft Edge" ...)'. These are machine-specific and should not live in the shared codebase; .claude/settings.local.json is typically user-scoped and gitignored. **Suggested fix:** Drop .claude/settings.local.json from the PR (it is unrelated to the refactor) and/or add it to .gitignore so per-machine tool settings never get committed.
"Bash(node -e \"console.log\\(\\(require\\('./package.json'\\).scripts||{}\\).build || '\\(no build script\\)'\\)\")",
"Bash(git check-ignore *)",
"Bash(node --check __tests__/simple_video.test.mjs)",
"Bash(node --check __tests__/run.mjs)"
Review

[SHOULD-FIX] F3
A tool-local settings file is committed to the shared repo and now carries absolute local paths to the author's machine: 'Bash(node /Users/kyle/.claude/jobs/1462df8d/tmp/transform2.mjs)' and 'Bash(open -a "Microsoft Edge" ...)'. These are machine-specific and should not live in the shared codebase; .claude/settings.local.json is typically user-scoped and gitignored.
Suggested fix: Drop .claude/settings.local.json from the PR (it is unrelated to the refactor) and/or add it to .gitignore so per-machine tool settings never get committed.

[SHOULD-FIX] F3 A tool-local settings file is committed to the shared repo and now carries absolute local paths to the author's machine: 'Bash(node /Users/kyle/.claude/jobs/1462df8d/tmp/transform2.mjs)' and 'Bash(open -a "Microsoft Edge" ...)'. These are machine-specific and should not live in the shared codebase; .claude/settings.local.json is typically user-scoped and gitignored. **Suggested fix:** Drop .claude/settings.local.json from the PR (it is unrelated to the refactor) and/or add it to .gitignore so per-machine tool settings never get committed.
"Bash(node --check __tests__/run.mjs)",
Review

[SHOULD-FIX] F3
A tool-local settings file is committed to the shared repo and now carries absolute local paths to the author's machine: 'Bash(node /Users/kyle/.claude/jobs/1462df8d/tmp/transform2.mjs)' and 'Bash(open -a "Microsoft Edge" ...)'. These are machine-specific and should not live in the shared codebase; .claude/settings.local.json is typically user-scoped and gitignored.
Suggested fix: Drop .claude/settings.local.json from the PR (it is unrelated to the refactor) and/or add it to .gitignore so per-machine tool settings never get committed.

[SHOULD-FIX] F3 A tool-local settings file is committed to the shared repo and now carries absolute local paths to the author's machine: 'Bash(node /Users/kyle/.claude/jobs/1462df8d/tmp/transform2.mjs)' and 'Bash(open -a "Microsoft Edge" ...)'. These are machine-specific and should not live in the shared codebase; .claude/settings.local.json is typically user-scoped and gitignored. **Suggested fix:** Drop .claude/settings.local.json from the PR (it is unrelated to the refactor) and/or add it to .gitignore so per-machine tool settings never get committed.
"WebSearch",
Review

[SHOULD-FIX] F3
A tool-local settings file is committed to the shared repo and now carries absolute local paths to the author's machine: 'Bash(node /Users/kyle/.claude/jobs/1462df8d/tmp/transform2.mjs)' and 'Bash(open -a "Microsoft Edge" ...)'. These are machine-specific and should not live in the shared codebase; .claude/settings.local.json is typically user-scoped and gitignored.
Suggested fix: Drop .claude/settings.local.json from the PR (it is unrelated to the refactor) and/or add it to .gitignore so per-machine tool settings never get committed.

[SHOULD-FIX] F3 A tool-local settings file is committed to the shared repo and now carries absolute local paths to the author's machine: 'Bash(node /Users/kyle/.claude/jobs/1462df8d/tmp/transform2.mjs)' and 'Bash(open -a "Microsoft Edge" ...)'. These are machine-specific and should not live in the shared codebase; .claude/settings.local.json is typically user-scoped and gitignored. **Suggested fix:** Drop .claude/settings.local.json from the PR (it is unrelated to the refactor) and/or add it to .gitignore so per-machine tool settings never get committed.
"Bash(node *)",
Review

[SHOULD-FIX] F3
A tool-local settings file is committed to the shared repo and now carries absolute local paths to the author's machine: 'Bash(node /Users/kyle/.claude/jobs/1462df8d/tmp/transform2.mjs)' and 'Bash(open -a "Microsoft Edge" ...)'. These are machine-specific and should not live in the shared codebase; .claude/settings.local.json is typically user-scoped and gitignored.
Suggested fix: Drop .claude/settings.local.json from the PR (it is unrelated to the refactor) and/or add it to .gitignore so per-machine tool settings never get committed.

[SHOULD-FIX] F3 A tool-local settings file is committed to the shared repo and now carries absolute local paths to the author's machine: 'Bash(node /Users/kyle/.claude/jobs/1462df8d/tmp/transform2.mjs)' and 'Bash(open -a "Microsoft Edge" ...)'. These are machine-specific and should not live in the shared codebase; .claude/settings.local.json is typically user-scoped and gitignored. **Suggested fix:** Drop .claude/settings.local.json from the PR (it is unrelated to the refactor) and/or add it to .gitignore so per-machine tool settings never get committed.
"Bash(python3 *)",
Review

[SHOULD-FIX] F3
A tool-local settings file is committed to the shared repo and now carries absolute local paths to the author's machine: 'Bash(node /Users/kyle/.claude/jobs/1462df8d/tmp/transform2.mjs)' and 'Bash(open -a "Microsoft Edge" ...)'. These are machine-specific and should not live in the shared codebase; .claude/settings.local.json is typically user-scoped and gitignored.
Suggested fix: Drop .claude/settings.local.json from the PR (it is unrelated to the refactor) and/or add it to .gitignore so per-machine tool settings never get committed.

[SHOULD-FIX] F3 A tool-local settings file is committed to the shared repo and now carries absolute local paths to the author's machine: 'Bash(node /Users/kyle/.claude/jobs/1462df8d/tmp/transform2.mjs)' and 'Bash(open -a "Microsoft Edge" ...)'. These are machine-specific and should not live in the shared codebase; .claude/settings.local.json is typically user-scoped and gitignored. **Suggested fix:** Drop .claude/settings.local.json from the PR (it is unrelated to the refactor) and/or add it to .gitignore so per-machine tool settings never get committed.
"Bash(git stash *)",
Review

[SHOULD-FIX] F3
A tool-local settings file is committed to the shared repo and now carries absolute local paths to the author's machine: 'Bash(node /Users/kyle/.claude/jobs/1462df8d/tmp/transform2.mjs)' and 'Bash(open -a "Microsoft Edge" ...)'. These are machine-specific and should not live in the shared codebase; .claude/settings.local.json is typically user-scoped and gitignored.
Suggested fix: Drop .claude/settings.local.json from the PR (it is unrelated to the refactor) and/or add it to .gitignore so per-machine tool settings never get committed.

[SHOULD-FIX] F3 A tool-local settings file is committed to the shared repo and now carries absolute local paths to the author's machine: 'Bash(node /Users/kyle/.claude/jobs/1462df8d/tmp/transform2.mjs)' and 'Bash(open -a "Microsoft Edge" ...)'. These are machine-specific and should not live in the shared codebase; .claude/settings.local.json is typically user-scoped and gitignored. **Suggested fix:** Drop .claude/settings.local.json from the PR (it is unrelated to the refactor) and/or add it to .gitignore so per-machine tool settings never get committed.
"Bash(npm -v)",
Review

[SHOULD-FIX] F3
A tool-local settings file is committed to the shared repo and now carries absolute local paths to the author's machine: 'Bash(node /Users/kyle/.claude/jobs/1462df8d/tmp/transform2.mjs)' and 'Bash(open -a "Microsoft Edge" ...)'. These are machine-specific and should not live in the shared codebase; .claude/settings.local.json is typically user-scoped and gitignored.
Suggested fix: Drop .claude/settings.local.json from the PR (it is unrelated to the refactor) and/or add it to .gitignore so per-machine tool settings never get committed.

[SHOULD-FIX] F3 A tool-local settings file is committed to the shared repo and now carries absolute local paths to the author's machine: 'Bash(node /Users/kyle/.claude/jobs/1462df8d/tmp/transform2.mjs)' and 'Bash(open -a "Microsoft Edge" ...)'. These are machine-specific and should not live in the shared codebase; .claude/settings.local.json is typically user-scoped and gitignored. **Suggested fix:** Drop .claude/settings.local.json from the PR (it is unrelated to the refactor) and/or add it to .gitignore so per-machine tool settings never get committed.
"Bash(npm install *)"
Review

[SHOULD-FIX] F3
A tool-local settings file is committed to the shared repo and now carries absolute local paths to the author's machine: 'Bash(node /Users/kyle/.claude/jobs/1462df8d/tmp/transform2.mjs)' and 'Bash(open -a "Microsoft Edge" ...)'. These are machine-specific and should not live in the shared codebase; .claude/settings.local.json is typically user-scoped and gitignored.
Suggested fix: Drop .claude/settings.local.json from the PR (it is unrelated to the refactor) and/or add it to .gitignore so per-machine tool settings never get committed.

[SHOULD-FIX] F3 A tool-local settings file is committed to the shared repo and now carries absolute local paths to the author's machine: 'Bash(node /Users/kyle/.claude/jobs/1462df8d/tmp/transform2.mjs)' and 'Bash(open -a "Microsoft Edge" ...)'. These are machine-specific and should not live in the shared codebase; .claude/settings.local.json is typically user-scoped and gitignored. **Suggested fix:** Drop .claude/settings.local.json from the PR (it is unrelated to the refactor) and/or add it to .gitignore so per-machine tool settings never get committed.
]
}
}
+31
View File
@@ -185,4 +185,35 @@ describe('Data flow: popup -> storage -> viewer', () => {
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
throw new Error('Entry with invalid mediaType should be rejected');
}
});
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
it('an HLS entry carrying a scan-time referer passes schema + round-trips', () => {
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
// The referer is bound to the entry at scan time so it survives the
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
// watch-later flow (the source tab may be long closed). Because the schema
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
// sets additionalProperties:false, a `referer` the store carries must be an
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
// explicit property or storage-layer validation would reject the entry.
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
const scrapedEntry = {
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
url: 'https://pageA.example.com/watch',
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
src: 'https://hv-h.phncdn.com/hls/live/master.m3u8',
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
hlsManifest: 'https://hv-h.phncdn.com/hls/live/master.m3u8',
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
mediaType: 'hls',
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
poster: 'https://pageA.example.com/thumb.jpg',
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
tab: 'Referer-protected stream',
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
referer: 'https://pageA.example.com/watch',
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
};
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
const valid = validate(scrapedEntry);
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
if (!valid) {
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
throw new Error(`Entry with referer fails schema: ${JSON.stringify(validate.errors)}`);
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
}
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
// The referer must survive the storage round-trip intact so the viewer can
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
// send it to the proxy when the user opens a previously-scanned entry.
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
const roundTripped = JSON.parse(JSON.stringify(scrapedEntry));
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
if (roundTripped.referer !== 'https://pageA.example.com/watch') {
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
throw new Error('referer lost during round-trip');
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
}
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
if (!validate(roundTripped)) {
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
throw new Error('Round-tripped entry with referer fails schema');
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
}
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
});
Review

[FYI] F5
Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency.
Suggested fix: Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.

[FYI] F5 Consistent with the PR's goal ('no references in src/'), src/ is clean, but the NEW test files still reference the very providers being removed — phncdn.com in data_flow.test.mjs and referer_proxy.test.mjs, and thisvid.com in hls_qa.test.mjs / simple_video.test.mjs. These read as real-world fixtures, not logic coupling, so they do not reintroduce the coupling the PR removes; flagging for awareness / consistency. **Suggested fix:** Optional: swap the real provider hostnames in test fixtures for neutral 'example.com' hosts to keep the spirit of the refactor consistent across the test suite.
});
+129 -10
View File
@@ -38,7 +38,7 @@ function makeChromeEnv() {
let sessionRules = {};
const webRequest = {
onBeforeSendHeaders: { addListener: (cb) => { handlers.onBeforeSendHeaders = cb; } },
onBeforeSendHeaders: { addListener: (cb, filter) => { handlers.onBeforeSendHeaders = cb; handlers.onBeforeSendHeadersFilter = filter; } },
onBeforeRequest: { addListener: (cb) => { handlers.onBeforeRequest = cb; } },
};
@@ -113,7 +113,7 @@ describe('PR#5: per-host referer capture + proxy race', () => {
if (!cap) throw new Error('onBeforeSendHeaders listener was not registered');
await cap(videoDetails(11, 'https://cdnA.example.com/hls/live/seg-1.ts', 'https://pageA.example.com/watch'));
const rA = env.storage['goon_referers']?.['cdna.example.com'];
const rA = env.storage['goon_referers']?.['example.com'];
if (!rA) throw new Error('referer was not captured for cdnA');
if (rA.referer !== 'https://pageA.example.com/watch') {
throw new Error('captured referer wrong: ' + JSON.stringify(rA));
@@ -123,10 +123,10 @@ describe('PR#5: per-host referer capture + proxy race', () => {
await env.handlers.onRemoved(22);
const after = env.storage['goon_referers'];
if (!after || !after['cdna.example.com']) {
if (!after || !after['example.com']) {
throw new Error('closing tab 22 wiped goon_referers — host A referer lost');
}
if (after['cdna.example.com'].referer !== 'https://pageA.example.com/watch') {
if (after['example.com'].referer !== 'https://pageA.example.com/watch') {
throw new Error('closing tab 22 corrupted the host A referer: ' + JSON.stringify(after));
}
});
@@ -143,7 +143,7 @@ describe('PR#5: per-host referer capture + proxy race', () => {
// The very tab that captured the referer now closes.
await env.handlers.onRemoved(11);
const after = env.storage['goon_referers']?.['cdnb.example.com'];
const after = env.storage['goon_referers']?.['example.com'];
if (!after) throw new Error('goon_referers for the capturing host was wiped on its own tab close');
if (after.referer !== 'https://pageB.example.com/v') {
throw new Error('capturing-tab close corrupted referer: ' + JSON.stringify(after));
@@ -153,7 +153,7 @@ describe('PR#5: per-host referer capture + proxy race', () => {
// (b) Should-fix + coverage: two hosts proxied concurrently must each keep
// their own referer. With a shared rule id + non-atomic install this regressed;
// per-host rule ids + the per-host lock keep them distinct.
it('concurrent proxy fetches for two hosts keep distinct per-host referer rules', async () => {
it('concurrent proxy fetches for two distinct registrable domains keep distinct referer rules', async () => {
const env = makeChromeEnv();
let fetchCalls = 0;
global.chrome = env.chrome;
@@ -173,7 +173,7 @@ describe('PR#5: per-host referer capture + proxy race', () => {
const handler = env.handlers.onMessage[env.handlers.onMessage.length - 1];
let respA, respB;
handler({ type: 'GOON_PROXY_FETCH', url: 'https://cdnA.example.com/hls/live/seg-1.ts', referer: 'https://pageA.example.com/watch' }, {}, (r) => { respA = r; });
handler({ type: 'GOON_PROXY_FETCH', url: 'https://cdnB.example.com/hls/live/seg-1.ts', referer: 'https://pageB.example.com/watch' }, {}, (r) => { respB = r; });
handler({ type: 'GOON_PROXY_FETCH', url: 'https://cdnB.other.com/hls/live/seg-1.ts', referer: 'https://pageB.other.com/watch' }, {}, (r) => { respB = r; });
await new Promise(r => setTimeout(r, 30)); // let async sends + installs flush
@@ -184,8 +184,8 @@ describe('PR#5: per-host referer capture + proxy race', () => {
// Each host must have its OWN installed DNR rule with its OWN referer —
// host A must NOT be served host B's referer (the 412 regression).
const rules = env.getSessionRules();
const ruleA = Object.values(rules).find(r => r.condition?.requestDomains?.includes('cdna.example.com'));
const ruleB = Object.values(rules).find(r => r.condition?.requestDomains?.includes('cdnb.example.com'));
const ruleA = Object.values(rules).find(r => r.condition?.requestDomains?.includes('example.com'));
const ruleB = Object.values(rules).find(r => r.condition?.requestDomains?.includes('other.com'));
if (!ruleA) throw new Error('no DNR rule installed for cdna; rules=' + JSON.stringify(rules));
if (!ruleB) throw new Error('no DNR rule installed for cdnb; rules=' + JSON.stringify(rules));
if (ruleA.id === ruleB.id) {
@@ -196,7 +196,7 @@ describe('PR#5: per-host referer capture + proxy race', () => {
if (refA !== 'https://pageA.example.com/watch') {
throw new Error('cdnA served wrong referer: ' + refA);
}
if (refB !== 'https://pageB.example.com/watch') {
if (refB !== 'https://pageB.other.com/watch') {
throw new Error('cdnB served wrong referer: ' + refB);
}
if (refA === refB) throw new Error('both hosts got the same referer — cross-host bleed');
@@ -236,4 +236,123 @@ describe('PR#5: per-host referer capture + proxy race', () => {
throw new Error(`expected idempotent install, but re-added rule ${installs - installsAfterFirst} extra time(s)`);
}
});
// (c) Domain generalization — the core 412 fix. A provider referer captured for
// ONE edge subdomain must cover a SIBLING subdomain: the DNR rule is scoped to
// the registrable domain and the store is domain-keyed, so a proxy for ev-h
// needs no per-subdomain capture.
it('a proxy for a sibling subdomain reuses the captured referer via the registrable domain', async () => {
const env = makeChromeEnv();
global.chrome = env.chrome;
global.fetch = async () => ({
ok: true,
status: 200,
headers: { get: () => 'application/vnd.apple.mpegurl' },
arrayBuffer: async () => new Uint8Array([0]).buffer,
});
await loadBackground(env);
// Capture the provider referer for hv-h.phncdn.com (one edge subdomain).
await env.handlers.onBeforeSendHeaders(
videoDetails(11, 'https://hv-h.phncdn.com/hls/live/seg-1.ts', 'https://provider.example.com/watch'));
// Proxy a SIBLING subdomain ev-h.phncdn.com with NO referer in the message:
// the SW must resolve it from the domain-keyed store.
const handler = env.handlers.onMessage[env.handlers.onMessage.length - 1];
let resp;
handler({ type: 'GOON_PROXY_FETCH', url: 'https://ev-h.phncdn.com/hls/live/seg-1.ts' }, {}, (r) => { resp = r; });
await new Promise(r => setTimeout(r, 20));
if (!resp || !resp.ok) throw new Error('sibling proxy failed: ' + JSON.stringify(resp));
const rules = env.getSessionRules();
const rule = Object.values(rules).find(r => r.condition?.requestDomains?.includes('phncdn.com'));
if (!rule) throw new Error('no domain-scoped DNR rule for phncdn.com; rules=' + JSON.stringify(rules));
const ref = rule.action.requestHeaders?.find(h => h.header.toLowerCase() === 'referer')?.value;
if (ref !== 'https://provider.example.com/watch') {
throw new Error('sibling edge served wrong referer: ' + ref);
}
// The rule is scoped to the registrable domain, NOT the exact subdomain.
if (rule.condition.requestDomains?.includes('ev-h.phncdn.com')) {
throw new Error('rule scoped to exact subdomain instead of registrable domain');
}
});
// (d) With no captured referer and none in the message, install NO rule: the
// self-origin fallback (guaranteed-wrong for hotlink protection) is dropped.
it('no captured referer and no message referer installs no DNR rule', async () => {
const env = makeChromeEnv();
let fetchCalls = 0;
global.chrome = env.chrome;
global.fetch = async () => {
fetchCalls++;
return {
ok: true,
status: 200,
headers: { get: () => 'application/vnd.apple.mpegurl' },
arrayBuffer: async () => new Uint8Array([0]).buffer,
};
};
await loadBackground(env);
const handler = env.handlers.onMessage[env.handlers.onMessage.length - 1];
let resp;
handler({ type: 'GOON_PROXY_FETCH', url: 'https://cdnA.example.com/hls/live/seg-1.ts' }, {}, (r) => { resp = r; });
await new Promise(r => setTimeout(r, 20));
if (!resp || !resp.ok) throw new Error('proxy fetch failed: ' + JSON.stringify(resp));
if (fetchCalls < 1) throw new Error('proxy fetch did not run');
const rules = env.getSessionRules();
if (Object.keys(rules).length !== 0) {
throw new Error('installed a rule with no referer (self-origin fallback not dropped): ' + JSON.stringify(rules));
}
});
// (e) GOON_GET_REFERERS (used by scan-time capture) resolves each manifest host
// via the domain-keyed store: a sibling subdomain gets the provider referer,
// and an uncaptured host is omitted.
it('GOON_GET_REFERERS resolves manifest hosts through the domain-keyed store', async () => {
const env = makeChromeEnv();
global.chrome = env.chrome;
await loadBackground(env);
// Capture for hv-h.phncdn.com only.
await env.handlers.onBeforeSendHeaders(
videoDetails(11, 'https://hv-h.phncdn.com/hls/live/seg-1.ts', 'https://provider.example.com/watch'));
const handler = env.handlers.onMessage[env.handlers.onMessage.length - 1];
let resp;
handler(
{ type: 'GOON_GET_REFERERS', hosts: ['ev-h.phncdn.com', 'hv-h.phncdn.com', 'cdnA.example.com'] },
{},
(r) => { resp = r; });
await new Promise(r => setTimeout(r, 15));
if (!resp || !resp.ok) throw new Error('GOON_GET_REFERERS failed: ' + JSON.stringify(resp));
const referers = resp.referers || {};
if (referers['ev-h.phncdn.com'] !== 'https://provider.example.com/watch') {
throw new Error('sibling subdomain did not resolve the captured referer: ' + JSON.stringify(referers));
}
if (referers['hv-h.phncdn.com'] !== 'https://provider.example.com/watch') {
throw new Error('captured subdomain did not resolve its own referer: ' + JSON.stringify(referers));
}
if ('cdnA.example.com' in referers) {
throw new Error('uncaptured host should be omitted from referers: ' + JSON.stringify(referers));
}
});
// Regression for the live 412 bug: the capture listener MUST request
// requestHeaders via extraInfoSpec, or details.requestHeaders is undefined and
// readReferer() returns '' for every request — so no referer is ever stored,
// every entry stays referer-less, and the phncdn proxy 412s. The callback-level
// tests above pass even without this (they hand requestHeaders in directly), so
// this asserts the registration itself.
it('the onBeforeSendHeaders capture listener requests requestHeaders via extraInfoSpec', async () => {
const env = makeChromeEnv();
global.chrome = env.chrome;
await loadBackground(env);
const filter = env.handlers.onBeforeSendHeadersFilter;
if (!filter) throw new Error('onBeforeSendHeaders listener registered no filter object');
if (!Array.isArray(filter.extraInfoSpec) || !filter.extraInfoSpec.includes('requestHeaders')) {
throw new Error('capture listener did not request requestHeaders via extraInfoSpec: ' + JSON.stringify(filter));
}
});
});
+4 -18
View File
@@ -8,7 +8,7 @@
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
"name": "vidplex",
"version": "0.1.0",
"dependencies": {
"@rollup/rollup-darwin-arm64": "^4.63.1"
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
"@rollup/rollup-darwin-arm64": "^4.60.4"
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
},
"devDependencies": {
"ajv": "^8.17.1",
@@ -1338,9 +1338,9 @@
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
]
},
"node_modules/@rollup/rollup-darwin-arm64": {
"version": "4.63.1",
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
"resolved": "https://registry.npmjs.org/@rollup/rollup-darwin-arm64/-/rollup-darwin-arm64-4.63.1.tgz",
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
"integrity": "sha512-FQNqd1lRy/0QhDk3xeRIkSBiCpXCiDnZO3YLVdcDKN1UBiKToNftCzcXYNLshmPDUMlu2TdeS8tGcsU6f3YF1Q==",
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
"version": "4.60.4",
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
"resolved": "https://registry.npmjs.org/@rollup/rollup-darwin-arm64/-/rollup-darwin-arm64-4.60.4.tgz",
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
"integrity": "sha512-tua0TaJxMOB1R0V0RS1jFZ/RpURFDJIOR2A6jWwQeawuFyS4gBW+rntLRaQd0EQ4bd6Vp44Z2rXW+YYDBsj6IA==",
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
"cpu": [
"arm64"
],
@@ -3423,20 +3423,6 @@
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
"fsevents": "~2.3.2"
}
},
"node_modules/rollup/node_modules/@rollup/rollup-darwin-arm64": {
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
"version": "4.60.4",
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
"resolved": "https://registry.npmjs.org/@rollup/rollup-darwin-arm64/-/rollup-darwin-arm64-4.60.4.tgz",
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
"integrity": "sha512-tua0TaJxMOB1R0V0RS1jFZ/RpURFDJIOR2A6jWwQeawuFyS4gBW+rntLRaQd0EQ4bd6Vp44Z2rXW+YYDBsj6IA==",
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
"cpu": [
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
"arm64"
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
],
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
"dev": true,
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
"license": "MIT",
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
"optional": true,
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
"os": [
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
"darwin"
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
]
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
},
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
"node_modules/rollup/node_modules/@types/estree": {
"version": "1.0.8",
"resolved": "https://registry.npmjs.org/@types/estree/-/estree-1.0.8.tgz",
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
Review

[FYI] F4
The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency.
Suggested fix: Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.

[FYI] F4 The PR also churns package-lock.json: it adds a top-level 'dependencies': {@rollup/rollup-darwin-arm64:'^4.60.4'} to the root '' entry and clears the 'dev':true/'optional':true flags on the darwin-arm64 package, while package.json now has an empty 'dependencies': {}. This looks like an npm-install side effect on the author's ARM-Mac that leaked in, unrelated to the decoupling refactor. Build still passes locally (copy-hlds.mjs copies the hls bundle), so it is not a functional break, but the lock churn is noise and a platform-specific optional dep should not be promoted to a hard dependency. **Suggested fix:** Re-generate package-lock.json on a clean install and avoid forcing @rollup/rollup-darwin-arm64 into 'dependencies'; keep it as the platform-optional dev dep it already is, or drop it entirely.
+3
View File
@@ -23,5 +23,8 @@
"typescript": "^5.7.0",
"vite": "^6.0.0",
"vitest": "^3.0.0"
},
"dependencies": {
"@rollup/rollup-darwin-arm64": "^4.60.4"
}
}
+98 -62
View File
@@ -227,6 +227,11 @@ async function injectFreshToken(url) {
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// re-query storage for a freshly-rotated token.
function invalidateTokenCache() { tokenCache.clear(); }
// CDN registrable domains whose referer we have already captured this SW
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// lifetime, so we do not re-query the tab for every segment of a stream.
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// Lost if the SW is killed; persistReferer's storage dedupe covers a fresh SW.
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
const refererCapturedDomains = new Set();
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
chrome.webRequest.onBeforeRequest.addListener(
async (details) => {
if (details.tabId < 0) return;
@@ -235,6 +240,28 @@ chrome.webRequest.onBeforeRequest.addListener(
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
if (isVideoRequest(details.url)) {
console.log('[goon-bg] webRequest: video request in tab', details.tabId, ':', details.url.substring(0, 180));
await persistVideoRequest(details.tabId, details.url);
// Capture the provider referer for this CDN host from the source tab's
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// URL: the browser sends the referring page's URL as the Referer for a
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// cross-origin video request, and that IS the tab's URL. extraInfoSpec
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// (reading the actual request headers) is rejected by some Chromium
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// builds and fails the whole service worker, so we derive the referer
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// from the tab instead. The `tabs` + host_permissions already allow it.
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
const host = hostOf(details.url);
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
const domain = host ? registrableDomain(host) : '';
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
if (domain && !refererCapturedDomains.has(domain)) {
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
try {
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
const tab = await chrome.tabs.get(details.tabId);
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
const ref = tab && tab.url;
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
if (ref) {
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
await persistReferer(host, ref);
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
refererCapturedDomains.add(domain);
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
console.log('[goon-bg] captured referer for', domain, '->', ref);
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
}
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
} catch {
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// Tab may have closed mid-stream; a later request from a live tab
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// re-captures. Best-effort.
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
}
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
}
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
}
// Also capture HLS manifest URLs (original logic)
@@ -258,7 +285,7 @@ chrome.webRequest.onBeforeRequest.addListener(
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
);
// Clean up stale PER-TAB entries when tabs are closed. NOTE: 'goon_referers' is
// a host-keyed GLOBAL store — it must outlive its capture tab so a proxy fetch
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// a domain-keyed GLOBAL store — it must outlive its capture tab so a proxy fetch
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// for that host still finds its referer after the source tab is closed. It is
// therefore NOT removed here (only the per-tab vr_requests_/wr_manifests_ keys
// are). Stale referers are pruned opportunistically in persistReferer.
@@ -308,87 +335,74 @@ function hostOf(url) {
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
try { return new URL(url).host.toLowerCase(); } catch { return ''; }
}
// Best-effort origin for a host when no referer was captured for it. Used only
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// as a fallback; a captured referer always wins.
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
function originFromHost(host) {
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
return host ? 'https://' + host + '/' : '';
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// Registrable domain of a CDN host — the label a single provider's referer
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// applies to. A referer-protecting CDN serves manifests and segments from many
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// edge subdomains that share ONE
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// provider referer, so we key the referer store, the DNR rule, and proxy-host
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// matching on this domain rather than the exact subhost.
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
//
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// Heuristic: the last two DNS labels. Correct for the overwhelmingly common
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// single-label public suffixes (.com, .net, .org, .io, ...). Known caveat:
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// multi-label public suffixes such as .co.uk / .com.au collapse to the public
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// suffix itself (foo.co.uk -> co.uk); that is an exotic edge case and no worse
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// than the exact-host behavior it replaces. A single-level host (no dot, or a
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// bare TLD) is returned unchanged.
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
function registrableDomain(host) {
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
if (!host) return '';
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
const parts = host.split('.');
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
if (parts.length < 2) return host;
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
return parts.slice(-2).join('.');
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
}
// Read the `Referer` request header (the browser may title-case it; the
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// (onBeforeSendHeaders cannot read request headers in MV3 without a blocking API,
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// so we fall back to the page origin for hosts that send no referer).
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
function readReferer(headers) {
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
if (!Array.isArray(headers)) return '';
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
for (const h of headers) {
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
if (h && h.name && h.name.toLowerCase() === 'referer') return h.value || '';
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
}
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
return '';
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
}
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// How long a captured referer is trusted before it is pruned on the next
// persist. Referers are short-lived session tokens; this keeps the per-host
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// global store from accumulating stale entries without tying cleanup to tab
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// close (a still-open source tab's referer must survive other tabs closing).
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
const REFERER_TTL_MS = 30 * 60 * 1000;
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// persist. Kept long enough for the scan-now / watch-later flow (a user scans a
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// page, then opens the viewer much later) without tying cleanup to tab close (a
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// still-open source tab's referer must survive other tabs closing).
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
const REFERER_TTL_MS = 7 * 24 * 60 * 60 * 1000;
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// Persist the captured referer per CDN host so the proxy can re-use it even
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// after this capture tab closes. Keyed by host so a proxy for that host can
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// look its referer up without knowing the source tab. Each entry is stamped
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// with a capture time; entries older than the TTL are pruned opportunistically
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// here (never on tab close).
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// Persist the captured referer per CDN registrable domain so the proxy can
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// re-use it even after the capture tab closes. Keyed by the host's registrable
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// domain so a single provider referer covers every edge subdomain (a CDN serves
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// manifests/segments from many subhosts that share one referer), and so a proxy
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// for any sibling subhost can look it up without knowing the source tab. Each
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// entry is stamped with a capture time; entries older than the TTL are pruned
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// opportunistically here (never on tab close).
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
async function persistReferer(host, referer) {
if (!host || !referer) return;
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
const domain = registrableDomain(host);
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
if (!domain || !referer) return;
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
try {
const stored = await chrome.storage.local.get('goon_referers');
const map = stored['goon_referers'] || {};
const now = Date.now();
// Opportunistic per-host prune: drop entries older than the TTL.
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// Opportunistic per-domain prune: drop entries older than the TTL.
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
for (const [h, entry] of Object.entries(map)) {
if (h !== host && (!entry || !entry.at || now - entry.at > REFERER_TTL_MS)) {
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
if (h !== domain && (!entry || !entry.at || now - entry.at > REFERER_TTL_MS)) {
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
delete map[h];
}
}
const current = map[host];
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
const current = map[domain];
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
if (current && current.referer === referer) return;
map[host] = { referer, at: now };
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
map[domain] = { referer, at: now };
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
await chrome.storage.local.set({ goon_referers: map });
} catch (e) {
console.warn('[goon-bg] persistReferer failed:', e && e.message ? e.message : e);
}
}
// Look up the referer captured for a host. '' when none has been captured
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// (or the stored entry is malformed).
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// Look up the referer captured for a host's registrable domain. '' when none
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// has been captured (or the stored entry is malformed).
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
async function getReferer(host) {
if (!host) return '';
try {
const stored = await chrome.storage.local.get('goon_referers');
const entry = (stored['goon_referers'] || {})[host];
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
const entry = (stored['goon_referers'] || {})[registrableDomain(host)];
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
return entry && entry.referer ? entry.referer : '';
} catch {
return '';
}
}
// Capture the referer the real page sends for its video requests.
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
chrome.webRequest.onBeforeSendHeaders.addListener(
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
async (details) => {
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
if (details.tabId < 0) return;
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
if (!isVideoRequest(details.url)) return;
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
const host = hostOf(details.url);
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
if (!host) return;
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
const referer = readReferer(details.requestHeaders);
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
if (referer) {
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// The browser already built a referer from the source page — use it.
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
await persistReferer(host, referer);
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
} else {
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// No referer was sent; assume the page's origin is the expected referer.
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
await persistReferer(host, originFromHost(host));
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
}
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
},
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
{ urls: ['<all_urls>'] }
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
);
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// Install (idempotently) a DNR rule that sets the captured referer for a given
// CDN host. The value comes from capture (or a fallback origin), never a
// hardcoded site. Scoped to the requested host so only proxy fetches for that
@@ -404,7 +418,8 @@ const refererRuleInstalled = new Map();
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
const refererRuleLocks = new Map();
async function ensureRefererRule(host, referer) {
if (!host || !referer) return;
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
const domain = registrableDomain(host);
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
if (!domain || !referer) return;
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
const api = chrome.declarativeNetRequest;
if (!api || typeof api.updateSessionRules !== 'function') {
console.warn('[goon-bg] declarativeNetRequest unavailable; skipping referer rule');
@@ -413,10 +428,10 @@ async function ensureRefererRule(host, referer) {
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// Serialize per host so the idempotency check + updateSessionRules is atomic.
const run = (async () => {
if (refererRuleInstalled.has(host) && refererRuleInstalled.get(host) === referer) {
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
if (refererRuleInstalled.has(domain) && refererRuleInstalled.get(domain) === referer) {
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
return;
}
const ruleId = refererRuleIdFor(host);
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
const ruleId = refererRuleIdFor(domain);
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
try {
// Remove only THIS host's own rule id, then add it back scoped to the host.
await api.updateSessionRules({
@@ -435,17 +450,17 @@ async function ensureRefererRule(host, referer) {
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
condition: {
// Host + subdomains ONLY, so only proxy fetches for this host get the
// rewritten referer; the source page's own requests keep theirs.
requestDomains: [host],
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
requestDomains: [domain],
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
},
}],
});
refererRuleInstalled.set(host, referer);
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
console.log('[goon-bg] DNR referer rule installed for', host, '(' + ruleId + ')');
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
refererRuleInstalled.set(domain, referer);
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
console.log('[goon-bg] DNR referer rule installed for', domain, '(' + ruleId + ')');
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
} catch (e) {
console.warn('[goon-bg] ensureRefererRule failed:', e && e.message ? e.message : e);
}
})().finally(() => { refererRuleLocks.set(host, null); });
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
refererRuleLocks.set(host, run);
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
})().finally(() => { refererRuleLocks.set(domain, null); });
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
refererRuleLocks.set(domain, run);
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
return run;
}
@@ -533,10 +548,13 @@ chrome.runtime.onMessage.addListener((message, _sender, sendResponse) => {
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
const proxyHost = hostOf(message.url);
if (proxyHost) {
// Prefer the provider's real referer captured from the live
// tab; fall back to a viewer-supplied or host-derived origin.
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// tab (domain-aware), or the entry's scan-time referer the viewer sent.
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// With neither, install no rule: a self-origin referer is
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// guaranteed-wrong for hotlink protection, so we never fall
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// back to one (it would just 412).
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
const captured = await getReferer(proxyHost);
const referer = captured || message.referer || originFromHost(proxyHost);
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
await ensureRefererRule(proxyHost, referer);
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
const referer = captured || message.referer || '';
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
if (referer) await ensureRefererRule(proxyHost, referer);
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
}
// hls.js resolves sub-playslists/segments relative to the master
// and can DROP the token query, handing us a token-less
@@ -571,6 +589,24 @@ chrome.runtime.onMessage.addListener((message, _sender, sendResponse) => {
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
return true;
}
case 'GOON_GET_REFERERS': {
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// Resolve the provider referer captured for each manifest host
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// (domain-aware: one provider referer covers every edge subdomain).
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// Used at scan time to bind the referer to a stored video entry so it
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
// survives the watch-later flow.
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
async function getReferers(message, sendResponse) {
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
const hosts = Array.isArray(message.hosts) ? message.hosts : [];
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
const referers = {};
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
for (const host of hosts) {
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
const r = await getReferer(host);
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
if (r) referers[host] = r;
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
}
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
sendResponse({ ok: true, referers });
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
}
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
getReferers(message, sendResponse);
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
return true;
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
}
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
default:
console.warn('[goon-plugin] Unknown message type:', message.type);
}
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.

Blocking: 'goon_referers' is a host-keyed GLOBAL store meant to outlive its capture tab (see the docstring just above and the PR body: "persist it per CDN host ... even after this capture tab closes"). Adding it to the per-tab onRemoved cleanup means closing ANY tab wipes referers captured by other still-open source tabs, so the next proxy fetch for those hosts loses its referer and 412s. Remove 'goon_referers' from this array (keep the per-tab vr_requests_ / wr_manifests_ keys). If you want cleanup, key it per host with an expiry instead of on every tab removal.
Review

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).

Concurrent-proxy race: REFERER_RULE_ID is a single shared id and the idempotency check + updateSessionRules({removeRuleIds, addRules}) are not atomic. Two hosts proxied in parallel (common in the multi-panel grid) can interleave so the live rule ends up host B's while refererRuleInstalled records host A as installed -> host A's fetches get host B's referer -> 412. Now that the rule is per-host, use a distinct rule id per host (or add all hosts' rules in one updateSessionRules call) and make the install+record atomic (per-host async lock / queue).
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
Review

[SHOULD-FIX] F2
Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The extraHeaders permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism.
Suggested fix: Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.

[SHOULD-FIX] F2 Documentation/behavior mismatch. The PR body, commit 66c6218 message, and the code comments at :311 (:311 'via chrome.webRequest.onBeforeSendHeaders, which sees the request headers') describe an onBeforeSendHeaders + extraHeaders header-reading capture. The actual code uses onBeforeRequest (cannot read request headers) and derives the referer from the source TAB's URL instead (chrome.tabs.get). The `extraHeaders` permission is not declared in manifest.json (permissions: storage, tabs, scripting, webRequest, declarativeNetRequest). The commit message for :245-248 ('extraInfoSpec ... is rejected by some Chromium builds and fails the whole service worker') explains why, but the PR body / high-level comments were never updated to match, so a reviewer is misled about the mechanism. **Suggested fix:** Reword the PR body and the :311/:308-314 comments to say the referer is derived from the source tab URL (via onBeforeRequest + chrome.tabs.get) rather than 'read from request headers via onBeforeSendHeaders'. The actual mechanism is sound and arguably safer than reading real headers; the mismatch is a documentation issue, not a correctness one.
+27
View File
@@ -1147,6 +1147,7 @@ async function scanVideos(tabsToScan) {
}
const items = [...seen.values()];
await attachReferers(items);
const skipped = scrapeResults.filter(r => r === null).length;
log('scanVideos complete: scanned=', tabResults.length, 'skipped=', skipped, 'unique items=', items.length);
@@ -1157,6 +1158,31 @@ async function scanVideos(tabsToScan) {
return { items, scanned: tabResults.length, skipped };
}
/**
* Bind the provider referer the SW captured (domain-aware) to each item so it
* survives the watch-later flow and covers the sibling edge subdomains a segment
* may resolve to. Best-effort: playback still falls back to the SW's live store.
*/
async function attachReferers(items) {
const hostOf = (url) => { try { return new URL(url).host.toLowerCase(); } catch { return ''; } };
const hosts = [];
for (const it of items) {
const host = hostOf(it.hlsManifest || it.src || '');
if (host) hosts.push(host);
}
if (hosts.length === 0) return;
let referers = {};
try {
const resp = await chrome.runtime.sendMessage({ type: 'GOON_GET_REFERERS', hosts });
if (resp && resp.referers) referers = resp.referers;
} catch { /* referer is best-effort; the SW still resolves it at play time */ }
if (!Object.keys(referers).length) return;
for (const it of items) {
const host = hostOf(it.hlsManifest || it.src || '');
if (host && referers[host]) it.referer = referers[host];
}
}
/** Store video entries in chrome.storage.local. */
async function storeVideos(items) {
const stored = items.map(v => ({
@@ -1165,6 +1191,7 @@ async function storeVideos(items) {
tab: v.tab,
mediaType: v.mediaType || 'video',
hlsManifest: v.hlsManifest || undefined,
referer: v.referer || undefined,
}));
await chrome.storage.local.set({
[STORAGE_KEY]: stored,
+5
View File
@@ -33,6 +33,11 @@
"type": "string",
"enum": ["direct", "hls"],
"description": "Media delivery format. Defaults to 'direct' when omitted. Set to 'hls' for HTTP Live Streaming."
},
"referer": {
"type": ["string", "null"],
"format": "uri",
"description": "Provider Referer captured at scan time for the manifest's CDN; the proxy sends it so a referer-protecting CDN accepts sibling edge subdomains. Optional."
}
},
"additionalProperties": false,
+33 -9
View File
@@ -178,7 +178,7 @@ const panelsEl = document.getElementById('panels');
// Route through the service worker to bypass the CDN's CORS/Referer
// restrictions. The SW adds the required Referer and returns base64.
console.log('[goon-viewer] Proxying via SW:', (url || '').substring(0, 120));
chrome.runtime.sendMessage({ type: 'GOON_PROXY_FETCH', url, referer: originForUrl(url) }, (resp) => {
chrome.runtime.sendMessage({ type: 'GOON_PROXY_FETCH', url, referer: refererForUrl(url) }, (resp) => {
if (self._aborted) return;
const lastErr = chrome.runtime.lastError;
if (lastErr || !resp || !resp.ok) {
@@ -264,19 +264,34 @@ HlsLoader.canSupportLoader = function () { return true; };
// hardcode a particular CDN.
const PROXY_HOSTS_KEY = 'goon_proxy_hosts';
const proxyHosts = new Set();
// Provider referers bound to a stored entry at scan time, keyed by the entry's
// manifest registrable domain. A single provider referer covers every edge
// subdomain, so a segment on a sibling subhost resolves to the same referer.
const refererByDomain = new Map();
(async () => {
try {
const stored = await chrome.storage.local.get(PROXY_HOSTS_KEY);
const hosts = stored[PROXY_HOSTS_KEY];
if (Array.isArray(hosts)) for (const h of hosts) proxyHosts.add(String(h).toLowerCase());
if (Array.isArray(hosts)) for (const h of hosts) proxyHosts.add(registrableDomainOf(String(h).toLowerCase()));
} catch { /* ignore */ }
})();
// Origin for a URL's host — used as the Referer to request when a proxy fetch
// is made for that host (the SW also captures the live referer per host).
function originForUrl(url) {
if (!url) return '';
try { return new URL(url).origin + '/'; } catch { return ''; }
// Registrable domain of a CDN host — the label a single provider referer
// applies to (a referer-protecting CDN serves from many edge subdomains that
// share one referer). Last-two-labels heuristic; mirrors the SW's copy.
function registrableDomainOf(host) {
if (!host) return '';
const parts = host.split('.');
if (parts.length < 2) return host;
return parts.slice(-2).join('.');
}
// The provider referer to send for a proxied request: the referer bound to the
// entry at scan time (domain-keyed), or '' when none was captured. No self-origin
// fallback — a self-origin referer is guaranteed-wrong for hotlink protection.
function refererForUrl(url) {
const ref = refererByDomain.get(registrableDomainOf(hostOfUrl(url)));
return ref || '';
}
// Whether a given video URL should be proxied: its host is a known
@@ -284,7 +299,7 @@ HlsLoader.canSupportLoader = function () { return true; };
function needsProxy(url) {
if (!url || typeof url !== 'string') return false;
const host = hostOfUrl(url);
return host ? proxyHosts.has(host) : false;
return host ? proxyHosts.has(registrableDomainOf(host)) : false;
}
// host of a URL (lowercased); '' when malformed.
@@ -1184,11 +1199,20 @@ async function init() {
// Probe CORS on each video URL. Hosts that block extension-origin
// requests are learned as referer-protected and persisted so the viewer
// routes them through the service-worker proxy.
// Bind each entry's scan-time referer to its manifest's registrable
// domain, so the proxy sends it for every edge subdomain the provider
// serves from (segments often resolve to a sibling subhost).
for (const entry of videoUrls) {
const ref = entry && entry.referer;
if (!ref) continue;
const dom = registrableDomainOf(hostOfUrl(entry.src || entry.hlsManifest || ''));
if (dom) refererByDomain.set(dom, ref);
}
const urlsToTest = videoUrls.map(v => v.src || v.hlsManifest || 'none').filter(u => u !== 'none' && typeof u === 'string');
const blockedHosts = new Set();
for (const url of [...new Set(urlsToTest)]) {
const corsOk = await corsCheck(url);
if (!corsOk) blockedHosts.add(hostOfUrl(url));
if (!corsOk) blockedHosts.add(registrableDomainOf(hostOfUrl(url)));
console.log('[goon-viewer] CORS', corsOk ? 'OK' : 'BLOCKED:', url.substring(0, 120));
}
if (blockedHosts.size > 0) {