Repository navigation
quest: plan the main → release audit's gaps, and fix its small findings - #9258
Conversation
Plans the 2026-11-20 release audit's unowned findings: a shared Rust/JS range table the cut waits on, explicit route forwarding states after the cut, the archive guide correction, the moq.pro attach recheck, and extra candidate checks in the cut. Fixes the small findings: Swift validates a track's refresh interval before UniFFI aborts on it, @moq/json's mutate keeps a published null, and the cut PR runs the full moq-transport matrix and released wire compatibility. 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 25 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. Each fix sits where the defect lives: Swift's existing duration guard, the one ?? fallback in mutate, and the two if/args expressions that select the cut PR's gate. The quest docs plan only what has no owner yet.
Reviewed changes
All 10 files in the single commit, including the workflow's reach through release-gate.yml and the Swift and JS callers.
- Cut PR interop gate:
interop.ymlruns the full moq-transport matrix andReleased wire compatibilitywhen apull_requesthas headmain. It reaches release PRs only throughrelease-gate.yml'sworkflow_call, which passes the caller'shead_ref. That gate'sinteroppaths filter includestest/**andinterop.yml, so a cut PR always selects it. Backports are unchanged. @moq/jsonmutateon a publishednull: the encoder'svaluegetter isundefinedonly before the first commit, so the!== undefinedcheck correctly treats a publishednullas the base. The new test covers both the no-initialand stale-initialpaths.- Swift
refresh.intervalcheck:checkTrackInfois the single choke point for bothpublishTrackandaccept, so the guard covers every public entry that takes aTrackInfo. - Quest plans: the new
range-table(m2) androute-forward-states(m4) quests, plus edits tocut,migrate-audit, andauth-attach-downstream. Every in-repo path they cite exists.Node::forwarddoes returnOption<Option<Via>>, andmoq-archivereads and writes only oneVERSION, which matches the migrate-audit note.
claude-opus-5-5 | 𝕏

Plans the gaps left by the 2026-11-20 audit of
main(9828a20cab) againstrelease(381b802d9b), and fixes its small findings.Problem
The audit found six release blockers and several smaller issues. Most already had an owner, and those stay where they are (see Overlap). The rest had none: the archive guide, the moq.pro attach recheck, a Swift crash, a JSON snapshot bug, the cut PR's thin interop gate, and the two refactors the audit recommended.
Approach
Quests:
sh/migrate.shrefuses MIGRATE.md edits outside the cut, so the fix lives here.828766e510still reads onlyquery. The wait now names moq.pro and Kixel as the parties, with a grep and a test to check each.Fixes:
checkTrackInfovalidatesrefresh.intervaltoo, so a negative, infinite, or NaN interval throwsMoqError.Configinstead of reaching UniFFI'sfatalError. Go already validated it. Kotlin's and Python's generated converters throw.@moq/jsonProducer.mutatetreats a publishednullas a value, since onlyundefinedmeans nothing was published. Before, it threw with noinitialand edited a staleinitialotherwise. Rust'sOption-based encoder is unaffected.interop.yml: the cut PR (headmain) runs the full moq-transport matrix and Released wire compatibility. A backport PR keeps the reduced matrix and still skips wire compatibility. Before, every release PR skipped the check against published versions, which is the one the cut most needs.Overlap
Already owned, unchanged here:
replacement-join-delayin moq-dev/moq#9246.Withdrawn: the audit's moq-net minimal-feature finding.
just rs no-asyncpasses on0f920c1e0d, and moq-net'sdefaultfeature set is empty, so its self dev-dependency enables nothing.Decisions
From the /quest-plan interview (maintainer, 2026-11-20):
Follow-ups
/quest-plan): the edge auth handler reads an attached credential fromtokenonattach,revalidate, andend, with a test that presents distinct credentials against the last released relay and the candidate. Required by moq.pro'squest/m0/adopt-cut-pin.md. Kept (maintainer, 2026-11-20 pick round).Caused by
TrackRefresh.intervalto the FFI without adding it to Swift's duration checks from fix(swift,go): a negative, infinite, or NaN duration throws a config error, not a crash #9063 (found by person, the release audit)mutatefell back toinitialwith??, which treats a publishednullas nothing (found by person, the release audit)Validation
mutate starts from a published nulltest fails onmain(mutate() requires a prior update()) and passes with the fix.bun test src/snapshot/passes 81, andtscand biome are clean.just swift checkon macOS (arm64, kixtop) at102ef455c7passes all 51 tests, includingtestBadDurationThrowsConfigwith the new refresh-interval cases. I did not run the new cases without the fix, since that path aborts the test process.just check(changed scope) passes, andquest checkpasses 510 documents.Impact
Public API: none (Swift and JS behavior only: a bad refresh interval throws, and
mutateafterupdate(null)receivesnull). Wire: none.Models
Size: M
Implementer: Claude Code, claude-opus-5-5, unknown effort
🤖 Generated with Claude Code
(Written by Claude Opus 5.5)
Review
Rewritten by
quest reviewafter each round.Head: 102ef45
Base: origin/main at 5169af0
Risk: high,
it changes .github/workflows/interop.yml, which is always high: workflows run with the repository's secrets and decide what merges and shipsClassified: high by the always-high paths
Reviewer: Codex, gpt-6.1-sol, high effort
Session:
01a1fb93-ba54-7c10-a6b3-47be143bba4e, round 1Score: 9/10,
No actionable defects were found, and the fixes and release-cut conditions match their callers and documented process; builds and tests were not run as instructed.Rounds: 1; first round 0; findings 0: fixed 0, refuted 0, open 0
Findings: none