Repository navigation
quest: plan follow-ups from the 2026-11-01 /quest-complete run - #7576
Conversation
New: mbx argument-length (m0/ci), the relay's all-features stack test (m0/test-flakes-2), moq export ending with its broadcast (m1), and byte-stream ladders (m1). Folds TRACK_INFO connection credit into fastest-reader. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
The model never responded, so Pullfrog stopped this run after 2 minutes. The request went out but nothing came back. This is usually temporary, so re-trigger to try again.
|
|
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 5 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. |
kixelated
left a comment
kixelated
left a comment
There was a problem hiding this comment.
Automated review by review (OpenAI)
Reviewed commit: 5c0f7cd.
No actionable bugs found in this plans-only change. The direction is sound: reuse the existing paused-stream credit work for TRACK_INFO starvation, diagnose export completion before changing its behavior, and fix the build/stack failures at their source rather than weakening tests. Keeping the ladder API decision ahead of the release cut avoids a separate later break.
Verification: static review of the eight changed plans/indexes and relevant importer, export-timing, stack-test and recipe-test code. I did not run tests, quest validation, or reproduce the reported failures.
(Written by OpenAI)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
kixelated
left a comment
kixelated
left a comment
There was a problem hiding this comment.
Automated review by review (OpenAI)
Reviewed commit: a14db93, focusing on the new fetched-copy SUBSCRIBE quest since my review of 5c0f7cd. The base is unchanged.
No new actionable bugs found. The plan preserves the live-edge requirement while targeting the missing upstream-to-downstream completion, rather than masking it with cached or stale edge data. Its direct/relay coverage on drafts 18 and 22 fits the existing fetch-only harness. The underlying hang is already recorded in #7545's discussion, not a newly introduced defect here; this PR plans the fix, it does not implement it.
Verification: static inspection of the plan delta, linked discussion, existing integration harness and relevant IETF response code. I did not run tests or reproduce the hang. My previous review had no findings to resolve.
(Written by OpenAI)
![pullfrog[bot]](https://avatars.githubusercontent.com/in/1768019?s=60&v=4){kind=small}
There was a problem hiding this comment.
✅ No new issues found.
Reviewed changes
This run re-reviews the delta since the prior review (5c0f7cd38e → a14db93c5d), plus the plans-only PR it lands on.
- Added
quest/m1/relay-fetched-copy-subscribe.md, planning a downstream moq-transport history SUBSCRIBE on a copy kept after a FETCH to get its SUBSCRIBE_OK from the upstream live edge rather than the cache or a stale reported edge. - Indexed the new quest in
quest/m1/README.mdand extended that milestone's 2026-11-01 "Added" paragraph. - Recorded the surrounding follow-up plans: the mbx argument-length fix in
quest/m0/ci, the--all-featuresWindows stack fix inquest/m0/test-flakes-2, the byte-stream ladder inquest/m1/stream-ladder.md,moq exportbroadcast-end inquest/m1/export-broadcast-end.md, and the folded TRACK_INFO connection-credit case inquest/m1/unbounded-groups/fastest-reader.md.
The plan's test target is accurate: rs/moq-net/tests/integration/fetch_only.rs already parametrizes moq-transport-18/moq-transport-22 with direct and relayed helpers, and the new quest's README link resolves with a matching description. This is a documentation-only change with no code or wire surface, so no inline comments.
deepseek-v4.1-flash (free via Pullfrog for OSS) ｜ 𝕏
Plans the follow-ups that came out of the 2026-11-01
/quest-completerun.Decisions (maintainer, 2026-11-01)
TRACK_INFO connection credit (follow-up of moq-dev/moq#7328): group data left unread while waiting for TRACK_INFO can use up the connection receive window, so TRACK_INFO itself can't arrive. FETCH has the same weakness. Where does the fix go?
quest/m1/unbounded-groups/fastest-reader.md, whose cap on paused-stream credit covers both cases.mbx "Argument list too long" (seen in a
just cirun on moq-dev/moq#7567):stack::debug_relay_load_and_run_fit_windows_stackunder--all-features(seen on moq-dev/moq#7211):moq exportnever reports the broadcast's end (found on moq-dev/moq#6987):Byte-stream ladders (follow-up of moq-dev/moq#7029):
lowerlands before the cut.Splitting the range-fill rules out of fetch-range (follow-up of moq-dev/moq#7487):
lite.mdrequires fetch-range only until moq-dev/moq#7545 lands, then drops that requirement.This PR doesn't carry decision 6.
lite.md's requirement on fetch-range exists only on quest(m1): lite-07 ranges wait for fetch-range #7487's branch, so the entry is removed once both quest(m1): lite-07 ranges wait for fetch-range #7487 and feat(net)!: end bounded reads and owe cached backfill (fetch as a range, part 1) #7545 have landed.A relay never answers a SUBSCRIBE on a copy it kept after a fetch (found on moq-dev/moq#7545; it hangs on main too):
Review
This change touches only quests, so it is simple under the simple-review rule (moq-dev/quest#364). The local review is skipped, and the online reviewers cover it.
quest check --base origin/mainpasses.(Written by Claude Opus 5.5)
🤖 Generated with Claude Code