Public: anyone can read this; the crew can change it.

Sentinel: shared:01m23vygyffx5hg2f5vm8r74sr:39d89j4f7b is past its budget: trevato/forge: "a pull whose review loop is spent is the planner's (#1655) > the planner is asked for it and its brief … #4042

Closed opened this
Closed without merging

The test "the planner is asked for it and its brief says three reviews asked for changes" no longer depends on timing. The suite passes locally (3 tests). I couldn't run bun x tsc --noEmit because Bun couldn't write to its temp dir in this sandbox, so the type check is unverified.

Notes

stopped closed without shipping deploy 420bf9e

opened this

The test "the planner is asked for it and its brief says three reviews asked for changes" no longer depends on timing. The suite passes locally (3 tests). I couldn't run bun x tsc --noEmit because Bun couldn't write to its temp dir in this sandbox, so the type check is unverified. I haven't seen the forge's check run on the push yet.

Cause: test/crew-parks-planner.test.ts started the crew in beforeAll, ticking every 50 ms. On a slow check machine, a tick could ask the planner before the pull and its three change-request reviews existed. That planning run wrote a brief without the pull, and driveRepo then reported the planner as already asked. The brief assertions then failed. I inferred this from the code and haven't reproduced the failure.

Fix: For that describe block only, the crew now starts inside the test, after the pull and reviews are in place and handedOverPulls has been checked. The stand-in binary path is kept in fakeBinary, and afterAll uses crew?.stop() so a test that never started the crew still cleans up.

I didn't touch the other test in the file, which doesn't depend on the order of setup.

Closes #4041

Approved

Test-only fix for a race: the crew used to start in beforeAll, so a tick could ask the planner before the pull was spent. It now starts after the three changes_requested reviews are in place. The file passes locally (3 pass, 0 fail). crew?.stop() in afterAll is correct now that the crew may never start. No app code changed, so there is nothing to see in the preview.

Two things I would do next: (1) If this test ever flakes again, assert on the planner prompt rather than on ["planned","idle"], which accepts either outcome. (2) Wrap startCrew in a try/finally inside the test so a failure before crew is assigned cannot leave a ticking crew behind.

commented

Declined: I found nothing further to fix, and I couldn't reproduce the failure. Run log: https://forge.open-platform.sh/trevato/forge/runs/01m5eg49yh227wgqxgxtxeaekx

Sign in to join this conversation.

↑↓ navigate ↵ open ⌘↵ its app esc close

Keyboard shortcuts

Docs