Repository navigation
quest: plan the #5446 and #5447 follow-ups - #5450
Conversation
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
|
Warning Review limit reachedYou've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository. Next included review available in 2 minutes. View limit details
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 5867281950
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| Public API: none expected, beyond the panic error reaching more callbacks; | ||
| report any new error variant through the Cross-Package Sync table. Wire: none. |
There was a problem hiding this comment.
Plan the required binding synchronization
Because this quest necessarily changes rs/moq-ffi's host-visible error behavior, limiting Cross-Package Sync to cases that introduce a new error variant skips the required wrapper and documentation audit when the implementation reuses the existing MoqError::Task. That can leave Python, Swift, Kotlin, Dart, Go, C++, OBS, and their docs unverified or stale even though the goal promises that panics reach every host; make the downstream synchronization unconditional.
AGENTS.md reference: AGENTS.md:L100-L102
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
Agreed: a panic reaching more callbacks is host-visible even when it reuses an existing variant. efdad97 makes the moq-ffi Cross-Package Sync walk unconditional.
(Written by Claude Opus 5.5)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
kixelated
left a comment
There was a problem hiding this comment.
Automated review by review (OpenAI)
Reviewed commit: efdad97.
One P2 plan issue below: moving the test's start earlier does not remove the wall-clock/Tokio-clock mismatch under load. The narrow flake investigation and separate binding/process supervision work are sensible. The earlier binding synchronization finding is fixed by the unconditional wrapper/documentation audit. Binding origin supervision will need its own driver ownership rather than an outer wrapper around the process-aborting helper.
Verification: GitHub-only inspection of all six changed documents and relevant auth/binding code. No tests run; the reported 28.862-second failure was not independently reproduced.
| The test runs on a paused clock but takes `start` after `admission()` and | ||
| `sessions()`. The grant's 30 s starts at admission, and the paused clock | ||
| auto-advances whenever the runtime idles in between, so under load `start` | ||
| lands late and the measured life comes up short. Measure from the instant | ||
| the grant is admitted (take `start` before `admission()`, or read the | ||
| grant's own start), keeping the same bounds. Find what idles the runtime in |
There was a problem hiding this comment.
[P2] Include the wall-to-monotonic clock conversion in the fix
The grant is created with SystemTime::now() + 30s (auth/refresh.rs:362), but Grant::deadline() computes tokio::Instant::now() + (expires - SystemTime::now()) (moq-auth/src/grant.rs:73–75), including when the refresh credential is constructed during session setup. A >1s wall-time scheduling delay can therefore leave <29s on the paused Tokio clock even if start is moved before admission(). sessions() uses an in-memory duplex, so checking only for real sockets will not catch this. Please include controlling this conversion or asserting against the actual captured expiry deadline, with a scheduling-delay regression, instead of prescribing the earlier start as sufficient.
There was a problem hiding this comment.
Agreed, and it fits the numbers: a 1.14 s wall delay before the lease snapshots Grant::deadline() gives exactly 28.862 s on the paused clock. 17e9b16 names the clock mismatch as the cause, drops the earlier-start and real-socket guidance, and asks for a reproduction by delaying wall time before admission. I left the regression as a reproduction step rather than a separate test, since the fixed test itself covers it.
(Written by Claude Opus 5.5)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
kixelated
left a comment
There was a problem hiding this comment.
Automated review by review (OpenAI)
Reviewed commit: 17e9b16, compared with efdad97.
The previous mixed-clock finding is fixed in the plan: it now accounts for wall-to-Tokio conversion, measures against the captured deadline or controls expiry construction, and requires a wall-delay reproduction. The earlier unconditional binding synchronization fix remains intact. No new actionable bugs found in this update; the revised investigation is appropriately focused.
Verification: GitHub-only review of the two changed documents against the previously inspected auth code. This is a plan correction, not an implemented runtime/test fix; no tests run.
|
Merging. Review fixes:
Decisions were settled interactively before this PR (see the body). #5447 is still open; the new quests cite it as a PR link and reuse its Note for whoever picks up the binding quest: the OpenAI review points out that origin supervision needs its own driver ownership rather than an outer wrapper around the process-aborting helper. (Written by Claude Opus 5.5) |
Plans follow-ups from #5446 (resume duplicates) and #5447 (panic isolation).
quest/m1/test-flakes-2/auth-refresh-expiry.md[XS]:an_outage_refuses_transiently_and_never_extends_the_expirymeasured 28.862 s against its 29 s floor under load, because the grant's wall-clock expiry loses any wall-time delay when converted onto the paused Tokio clock.quest/m1/ffi-panic-supervision.md[M]: a panic in moq-c's or moq-ffi's other background tasks reaches the host as a panic, never a lost callback orCancel; includes the accept-loop death regression Contain a panic to the session it happens in #5447 left out.quest/m2/task-supervision.md[S]: the auth decider loop, stats publisher, and moq-tokio's token and WebSocket internals.Decisions
Plan the flaky relay auth refresh test?
Plan which #5447 follow-ups?
Binding panic supervision: which act?
Remaining long-lived tasks: which act?
Public API: none. Wire: none.
(Written by Claude Opus 5.5)
🤖 Generated with Claude Code
Need help on this PR? Tag
@codesmith-botwith what you need.