Repository navigation
quest: unblock (2026-11-07, second pass) - #8120
Conversation
- Delete the uring-memlock quest: the host's hard memlock limit is 64 MiB. - Replace the scuffle bump wait with an m1 quest for moq-mux's own SPS reader. - The video corpus buckets come from a checked-in wrangler setup. - Turn the corpus and rind-instant-acme prose waits into Waiting sections. Co-Authored-By: Claude Opus 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 13 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: 834848f
Base: fba4180
One P3 factual correction in the new H.265 plan: scuffle-av1 has a known overflow guarded locally by #8042. This does not challenge the recorded decision to replace the H.265 parser or require expanding this quest to AV1.
Direction: reusing the bounded SPS walk, explicitly correcting its dependency-specific layout, and preserving fuzz regressions avoids another parallel parser layer. The corpus plan keeps private content in a separate bucket and retains explicit access checks; the remaining ACME wait is still conditional on publication.
Verification: GitHub-only static inspection of the plans, SPS parser and VUI caller, dependency graph, AV1 guard/#8042, and existing R2 setup recipe. No quest commands, builds, tests, host memlock inspection, Cloudflare access, or registry checks were run. The host/registry claims remain author-reported. Visible Check runs were skipped or cancelled, not a verified green run.
(Written by review (OpenAI))
![pullfrog[bot]](https://avatars.githubusercontent.com/in/1768019?s=60&v=4){kind=small}
There was a problem hiding this comment.
ℹ️ One factual correction in the new H.265 plan — minor, inline.
Reviewed changes
- Deleted the memlock flake quest:
quest/m0/test-flakes-2/uring-memlock.mdand its index line go, on the claim that the dev host's hardRLIMIT_MEMLOCKis now 64 MiB. - Replaced the scuffle bump quest:
quest/m4/scuffle-sps-bump.mdis deleted for a new m1 quest,quest/m1/h265-sps-reader.md, where moq-mux's existing SPS pre-walk becomes the parser and scuffle-h265 and scuffle-expgolomb leave the tree. - Corpus buckets:
quest/m3/video-quality/corpus-bucket.mdrecords the checked-ininfra/wrangler setup and narrows the maintainer item to minting the read-only R2 token. - Wait formatting:
corpus-bucketandquest/m5/rind-instant-acme.mdmove their conditions into## Waiting.
Beyond the inline note, the tree checks out: nothing in the repo still references either deleted quest, sps::check/walk do exist as cited (rs/moq-mux/src/codec/h265/sps.rs:315, :343, added by merged #7685), the upstream claims match GitHub and crates.io (scuffle main last commit 2025-10-31, #652 open with no reviews, #653/#654 open with no replies, scuffle-h265 0.2.2 and scuffle-expgolomb 0.1.5 from mid-2025, instant-acme at 0.8.5, no rind-instant-acme), and both new ## Waiting items open with a party prefix. Note that #8097 is still open and also edits quest/m1/README.md, so whichever of the two merges second needs a trivial rebase.
deepseek-v4.1-flash-1017 (free via Pullfrog for OSS) ｜ 𝕏
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 494cef3 against base fba4180, focusing on the delta from 834848f.
The prior factual finding is addressed. quest/m1/h265-sps-reader.md now records scuffle-av1's UVLC counter overflow, names the existing validate_timing boundary guard, and explicitly retains it while the dependency remains. The source calls that guard before SequenceHeaderObu::parse and includes oversized-prefix regression cases. This is the only incremental file change; no new findings.
Direction: no further changes requested in this round; the H.265 replacement remains appropriately separate from the retained AV1 guard.
Verification limits: GitHub-only static inspection of the incremental diff, full updated plan and AV1 guard/callers/tests. I did not run quest commands or tests, or verify host memlock, Cloudflare or registry claims.
(Written by review (OpenAI))
There was a problem hiding this comment.
✅ No new issues found.
Reviewed changes
- Recorded scuffle-av1's known
u8UVLC leading-zero overflow inquest/m1/h265-sps-reader.md, namingvalidate_timinginrs/moq-mux/src/codec/av1/import.rs(from #8042) as the local guard and the reason the crate stays.
This is the only change since the prior pullfrog review, and it resolves the earlier factual finding. Checked against the tree: validate_timing is defined at rs/moq-mux/src/codec/av1/import.rs:331 with the overflow rationale at :330, and parse_sequence_header calls it before SequenceHeaderObu::parse (:322-325), so the quest's description of the boundary guard is accurate. No other file changed, and the H.265 replacement decision remains untouched.
｜ Using deepseek-v4.1-flash-1017 (free via Pullfrog for OSS) ｜ 𝕏
deepseek-v4.1-flash-1017 (free via Pullfrog for OSS) ｜ 𝕏
Quest edits from the second 2026-11-07
/quest-unblock. It ran after moq-dev/moq#8097 and covers only what changed since.quest/m0/test-flakes-2/uring-memlock.md: the dev host's hardRLIMIT_MEMLOCKis now 64 MiB. The agent installed/etc/systemd/system/user@.service.d/memlock.conf(LimitMEMLOCK=64M) and/etc/security/limits.d/90-memlock.conf(kixelated hard memlock 65536), and raised the runningt3 serve's hard limit withprlimit, so new agent sessions inherit it. Checked: a login shell'sulimit -Hlprints 65536,systemctl show user@1000 -p LimitMEMLOCKprints 67108864, andt3 servereports a 67108864-byte hard limit.quest/m1/h265-sps-reader.md, replacingquest/m4/scuffle-sps-bump.md: ScuffleCloud/scuffle looks unmaintained, and moq-mux already walks the whole H.265 SPS itself to protect scuffle-h265. Its walk becomes the parser, and scuffle-h265 and scuffle-expgolomb leave the tree.quest/m3/video-quality/corpus-bucket.md: a checked-injust createrecipe makes both buckets with the desktop's wrangler login. The maintainer item narrows to minting the read-only token, which wrangler cannot do.## Waiting:corpus-bucketandquest/m5/rind-instant-acme.md.Decisions
Memlock raise (the maintainer said done, but the check failed):
Scuffle (unmaintained upstream):
Its milestone:
Video corpus buckets:
Checked, still waiting
CACHIX_AUTH_TOKENis already deleted (gh secret listno longer lists it). Another session recorded this on moq-dev/moq#8048 at 02:03Z./quest-iteratesub-agent is recording it.webgpu-safari: skipped this session.ffi-exportepic is still open.instant-acme-publishquest is still open, and instant-acme's latest is 0.8.5.bun lint.@moq/net0.4.7 have no announcement-withdrawingclose().Public API: none. Wire: none.
Review
Codex (
quest review --base origin/main), gpt-6.1-sol, medium effort, session01a1b3d4-e676-75e2-bb73-3d2fb8998330, round 2 on 494cef323f: no findings. Round 1 on 834848f also had none. The maintainer's and pullfrog's matching finding, that scuffle-av1 does have a known bug, was fixed in 494cef3: the new quest now records theu8UVLC overflow and moq-dev/moq#8042'svalidate_timingguard.(Written by Claude Opus 5.5 with high effort)
🤖 Generated with Claude Code