Repository navigation
fix(moq-srt): the test target builds on its own - #9163
Conversation
Quest-Claim: d209b48a6637cc40
…uild alone nextest starts a test's timeout when it starts mbx's runner, so a test that waits for a host-wide scheduler permit is killed before it runs. Disable mbx test scheduling for this workspace; quest gate and nextest's thread count already bound test load. moq-srt's ts tests call `Catalog::subscribe`, which needs hang's `async` feature, so enable it in the dev-dependencies. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
No actionable comments were generated in the recent review. WalkthroughThe Merge Risk:
|
There was a problem hiding this comment.
✅ No new issues found.
Approach: sound. It fixes the cause, which is mbx queueing tests for a scheduler permit that counts against nextest's timeout, rather than raising timeouts or patching individual tests.
Reviewed changes
.mbx.toml: setsscheduler.tests = falseworkspace-wide, so directcargo nextest runcalls and CI get it too.rs/moq-srt/Cargo.toml: adds a path-onlyhangdev-dependency withasync, matching how internal dev-dependencies such asworkspace-hackare declared, socargo test -p moq-srtcompiles alone.nix/Cargo.nix: regenerated to match the new dev-dependency.quest/m0/test-flakes-3/srt-server-load.md: deletes the finished quest file.
claude-sonnet-5-5 | 𝕏
|
Maintainer decision (2026-11-19, /quest-complete): ✅ drop the (Written by Claude Opus 5.5 with high) |
…3/srt-server-load
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
There was a problem hiding this comment.
✅ No new issues found.
Reviewed changes since the prior pullfrog review (2a1fac9):
.mbx.tomlremoved: the workspace-widescheduler.tests = falseopt-out is gone, so host-wide test scheduling is unchanged.- Quest doc rewritten:
quest/m0/test-flakes-3/srt-server-load.mdnow tracks the remaining mbx/dotfiles fix with a## Waitingitem and the one-core reproduction. - Kept: the
moq-srthang(async) dev-dependency and the regeneratednix/Cargo.nix, which fix the standalone test build.
claude-sonnet-5-5 | 𝕏
|
Merge pick (2026-11-19, /quest-complete): ✅ /quest-merge at f12bbdc, applying the maintainer decision above (land the srt fix, fix the scheduler in mbx). (Written by Claude Opus 5.5 with high) |

Problem
moq-srt's test target did not compile alone (cargo test -p moq-srt):ts::testscallshang::Catalog::subscribe, which needs hang'sasyncfeature, and only workspace feature unification turned it on.Four
moq-srtserver::testsalso hit nextest's 60s timeout in a loadedjust check(moq-dev/moq#9123). That cause is outside this repo: the host's global mbx config setsscheduler.tests = true, so each test waits for a host-wide permit insidembx-test-runner, and nextest counts the wait toward the test's timeout. Per the maintainer's decision, that gets fixed in mbx or the dotfiles config, not here.Approach
rs/moq-srtdev-depends onhangwithasync(path-only, as internal dev-dependencies are), andnix/Cargo.nixis regenerated..mbx.toml: the checked-in opt-out from the earlier head is gone.## Waitingitem on the mbx/dotfiles fix and the one-core reproduction to check it.Validation
moq-srt's lib tests in a systemd scope capped at one CPU (CPUQuota=100%, 4 test threads,--stress-count 20,KIO_DENY_HOST_TIMERS=1) timed out 20 tests at 60s, including pure ones, each printingmbx: waiting for machine-wide scheduler capacity before starting this test; nextest counts this wait toward its slow-timeout. WithMBX_SCHEDULER_TESTS=falsethe same run passed 20/20, slowest 3.36s. That reproduction is now the quest's check.cargo test -p moq-srt --lib --no-runfailed with E0599 onCatalog::subscribebefore the dev-dependency and builds after it.cargo test -p moq-srtpassed 28/28, andjust checkpassed.Impact
Decisions
.mbx.tomlopt-out and land only the moq-srt dev-dependency fix (rejected: land the opt-out; raise nextest'sterminate-afterhere). Host-wide test scheduling stays on; the queued-test timeout is fixed in mbx or the dotfiles config.Caused by
ts::testscall toCatalog::subscribewithout enabling hang'sasyncfeature in moq-srt's dev-dependencies (found by person).Follow-ups
None. The rewritten quest quest/m0/test-flakes-3/srt-server-load.md keeps the remaining work.
Models
Size: XS
Implementer: Claude Code, claude-opus-5-5, medium effort
Implementer: Claude Code, claude-opus-5-5, medium effort (iterate, XS)
(Written by Claude Opus 5.5 with medium)
🤖 Generated with Claude Code
Review
Rewritten by
quest reviewafter each round.Head: f12bbdc
Base: origin/main at 83c9fdd
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:
01a1f332-e643-7292-a01a-c118e901083e, round 2Score: 9/10,
No actionable defects were found in the incremental changes, which remove the scheduler opt-out and document the upstream follow-up; builds and tests were not run as instructed.Rounds: 2; first round 0; findings 0: fixed 0, refuted 0, open 0
Findings: none