Repository navigation
fix(moq-mux): the TS follower judges a publisher leaving by its error's retry answer - #9070
Conversation
Quest-Claim: 359179191b719553
…'s retry answer Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
|
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 10 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. |
There was a problem hiding this comment.
✅ No new issues found.
Approach: sound — left asks moq_net::Error::retry whether the source failed to deliver instead of keeping a second, hand-maintained variant list, so the TS export and moqsrc share one rule; keeping Remote(_) as the single matches! is the moq-transport exception the maintainer settled in #8955.
Reviewed changes
leftclassification. The old true set survives intact (every previously matched variant isRetry::SameorReroute), and the newly counted codes are exactly the delivery failuresretrynames:Cancel,Closed,GoingAway,Dropped,GoawayTimeout, and streamInternal/Cancel/GoingAway/ControlTimeout.- Call sites. All four uses in
container/ts(follower.rsRunning/Resuming/poll_switch,export.rsprepare_pending/leave) now follow the same rule, so a left failure retires the program and reports the error after the buffered tail instead of failing mid-stream. With the default--follow 0sthe follower still reports the error immediately. - Docs. Both comments now state the
retryrule and the moq-transport PUBLISH_DONE exception rather than enumerating variants, which keeps them from drifting again. - Tests.
a_follower_fails_at_once_on_its_own_failurepins the verdicts (MalformedTrack,NotFound, streamMalformedTrack) anda_follower_waits_out_the_follow_on_a_lost_routeaddsCancel,GoingAway, and streamInternal/GoingAway. I re-ran the suite: the 20a_follower_tests pass, and fullcontainer::tsis 749 passed with only the 20 sandbox-only missing-ffmpeg/ffprobefailures, consistent with the author's 769. - Quest.
quest/m2/ts-follow-retry.mdis removed in the same PR as other landed quests, and no reference to it remains in the tree.
deepseek-v4.1-flash | 𝕏

moq_mux::container::ts::follower::leftnow asksmoq_net::Error::retrywhether the publisher left, instead of matchingmoq_net::Errorvariants, so the TS export and moq-gst'smoqsrcagree on what ends a run. It keepsRemote(_)as the one case it still matches, since over moq-transport a relay's lost upstream and a publisher's own failure both end with an unregistered code.Settled by the maintainer in #8955: a cancel, GOAWAY, or INTERNAL_ERROR reset counts as the publisher leaving, as
moqsrcdoes.Public API and wire
None. A
Followernow waits out its follow (instead of failing at once) onCancel,Closed,GoingAway,Dropped, and the stream codesInternal,Cancel,GoingAway, andControlTimeout. A stitched replacement failing with one of them before it commits also keeps the old program instead of failing the export.Tests
a_follower_fails_at_once_on_its_own_failurenow uses verdicts (MalformedTrack,NotFound, streamMalformedTrack);a_follower_waits_out_the_follow_on_a_lost_routeaddsCancel,GoingAway, and streamInternal/GoingAway.Validation
cargo test -p moq-mux --lib container::ts(769 passed) andjust checkpass. No other behavior to exercise: the change is a pure error classification, pinned by the two tests above.Models
Size: S
Implementer: Claude, claude-sonnet-5-5, unknown effort (arm B)
(Written by Sonnet 5.5)
Review
Rewritten by
quest reviewafter each round.Head: 796de52
Base: origin/main at 0a4a977
Risk: medium,
the base has no RISK.md to name what is always high or customer-facing, so every change is mediumClassified: medium by default
Reviewer: Codex, gpt-6.1-sol, medium effort
Session:
01a1ecf6-eeae-70c1-b4e7-63f9247c2712, round 1Score: 9/10,
The focused implementation matches the intended retry behavior and preserves the moq-transport exception, with no actionable defects found by source review; tests were not run as instructed.Rounds: 1; first round 0; findings 0: fixed 0, refuted 0, open 0
Findings: none