ci: run all darwin test lanes on main / opt-in only for now - #39191
Conversation
The mac agent pool cannot keep up with PR volume, so PR builds were waiting hours on the darwin 14 aarch64 lane. Extend the existing gate from the 26 lane to every darwin test lane: they run on main, on manual builds, or when the commit subject has [macos tests]. darwin still builds on every PR. No-Verification-Needed: pipeline generator only; checked with --dry-run
|
Warning Review limit reachedYou’ve reached a temporary PR review limit under our Fair Usage Limits Policy. Next review available in: 34 minutes Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available. How can I continue?After more reviews become available, a review can be triggered using the To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews. How do review limits work?CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability. For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window. Please refer docs for additional details. Review details⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: ASSERTIVE Plan: Pro Run ID: 📒 Files selected for processing (1)
Comment |
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
There was a problem hiding this comment.
LGTM — straightforward extension of the existing darwin test-lane gate.
What was reviewed:
- Filter change from
os !== "darwin" || tier !== "latest" || …toos !== "darwin" || …correctly drops all three darwin test lanes on PRs while keeping them on main/manual/opt-in. testStepKeys(release gate) is only consumed on main, wheredarwinTestsEnabledis true, so no dangling depends_on.traceOrderTargetsdarwin entry is gated separately onisMainBranch()/[generate symbol order]and unaffected.- Comment above
testPlatformsupdated to match; no other references to the old variable name.
Extended reasoning...
Overview
This PR extends the existing darwin-test-lane gate in .buildkite/ci.mjs from covering only the tier: "latest" (macOS 26) lane to covering all three darwin test lanes. The change is a rename (darwinLatestLaneEnabled → darwinTestsEnabled), a simplified filter predicate (drops the tier !== "latest" clause and the now-unused tier destructure), and a matching comment update above testPlatforms. Darwin builds still run on every PR; only the test lanes are gated to main / manual / [macos tests] opt-in.
Security risks
None. This is CI pipeline generation logic that decides which test lanes to schedule; it touches no auth, crypto, or user-facing code paths.
Level of scrutiny
Low. It's a ~10-line CI config change that widens an existing, well-commented gate pattern (introduced in #37980) from one lane to three. The PR description includes --dry-run output confirming the three expected states (PR, PR+opt-in, main), and the change is explicitly framed as temporary/revertable.
Other factors
Checked downstream consumers: relevantTestPlatforms feeds testStepKeys, which the release step depends on — but that step only runs when isMainBranch(), where darwinTestsEnabled is always true, so no darwin depends_on is dropped there. The traceOrderTargets darwin-aarch64 entry lives in the build section under its own isMainBranch()/commit-tag gate and doesn't read relevantTestPlatforms. The unused tier destructure was correctly dropped from the filter callback. No lingering references to the old variable name.
…n take (#39195) #39191 took darwin tests off PR builds because the version-specific lanes were queueing for hours. The arm64 pool is not actually short of capacity for PR volume; it is fragmented: the agents behind the `26` lane sit mostly idle (that lane is main-only) while the `14` lane backs up. This gives PR builds **one untiered `darwin aarch64` test lane** whose agent selector is just `queue=test-darwin, os=darwin, arch=aarch64`, so every arm64 mac agent serves it regardless of macOS version. `main`, manual builds and `[macos tests]` commits keep the three version-specific lanes. `getTestAgent` only emits `release-tier` when a tier is set, and the untiered lane sets no `EXPECTED_PLATFORM_RELEASE`, so the runner's platform assert checks os/arch only. `--dry-run`: ``` PR: darwin-aarch64-any-test-bun agents={queue: test-darwin, os: darwin, arch: aarch64} depends_on=darwin-aarch64-build-bun parallelism=2 main: darwin-aarch64-26-test-bun agents={..., release-tier: latest} EXPECTED_PLATFORM_RELEASE=26 darwin-aarch64-14-test-bun agents={..., release-tier: previous} darwin-x64-14-test-bun agents={queue: test-darwin, os: darwin, arch: x64} ```
x64 darwin tests came off PR builds in #39191 when the fleet was down to 4 Intel minis. MacStadium returned the other 5 today, so there are 9 again. This adds `darwin-x64-any-test-bun` to PR builds, shaped exactly like the existing aarch64 PR lane: untiered (`queue=test-darwin, os=darwin, arch=x64`, any Intel box), 2 shards, and the same `--skip-slower-than=10000` stopgap so it fits the pool at PR volume (~6-7 min shards; 9 slots ≈ 216 slot-hours/day vs ~180 needed). `main`, manual and `[macos tests]` builds still run the full `darwin-x64-14` lane. `--dry-run`: PR → `darwin-aarch64-any-test-bun` + `darwin-x64-any-test-bun` (both skip-slow, par 2); main → `26` / `14` / `x64-14` unchanged. The x64 lane depends on `darwin-x64-build-bun`, which already runs on PRs.
2 participants
PR builds have been waiting hours for the
darwin 14 aarch64test lane; the mac agent pool cannot keep up with current PR volume. This extends the existing gate (#37980, which covered only the 26 lane) to every darwin test lane: they run onmain, on manual builds, or when the commit subject contains[macos tests]. darwin still builds on every PR, so compile breakage is still caught pre-merge.--dry-run:Revert when the pool can take PR volume again.