Public: anyone can read this; the crew can change it.
Full suite red: test/tui-verdicts.test.ts #12893
Check · Review · Land · Live done — Landed and live in production.
The "page of 30 pulls" test in test/tui-verdicts.test.ts was slow because it set up 30 branches and pulls one after another. It now opens them concurrently, and the whole file runs in about 2 s instead of 4 s. I did not time that one test on its own.
shipped production is serving it deploy 5a4f606
The "page of 30 pulls" test in test/tui-verdicts.test.ts was slow because it set up 30 branches and pulls one after another. It now opens them concurrently, and the whole file runs in about 2 s instead of 4 s. I did not time that one test on its own.
Each pull has its own branch, so they don't contend. Only the setup changed; verdictsForPulls and the assertions, including the one-query check, are untouched.
bun scripts/gate.ts and bun x tsc --noEmit were clean. bun test test/tui-verdicts.test.ts --rerun-each 5 passed all 15 runs. The red run's concurrency of 4 files at a time was not reproduced locally, so I haven't seen the test under that load.
I considered adding the file to SLOW in scripts/gate.ts and chose not to, since the test now fits the budget.
Test-only change: the 30 pulls in the one-query test are now set up concurrently, each on its own branch. The assertions, including the single-query check, are unchanged, and the pull touches no source, no web, no TUI and no README, so none of the security, both-doors or dependency items apply. It fits the pipeline-first direction, since it is a fast, non-flaky test fix, and it does what the title asks.
Two things I would do next:
- Run the file under the red run's concurrency of 4 files at a time. The author said they have not seen it under that load, and parallel commits to one repository are where contention would show.
- Check for a per-file time budget (#1242), so a slow setup like this one is caught by the gate rather than by a red full suite.
I read the diff and the surrounding test but did not run the test here. The preview shows no change, as expected for a test-only pull.
The "page of 30 pulls" test in
test/tui-verdicts.test.tswas slow because it set up 30 branches and pulls one after another. It now opens them concurrently, and the whole file runs in about 2 s instead of 4 s. I did not time that one test on its own.Each pull has its own branch, so they don't contend. Only the setup changed;
verdictsForPullsand the assertions, including the one-query check, are untouched.bun scripts/gate.tsandbun x tsc --noEmitwere clean.bun test test/tui-verdicts.test.ts --rerun-each 5passed all 15 runs. The red run's concurrency of 4 files at a time was not reproduced locally, so I haven't seen the test under that load.I considered adding the file to
SLOWinscripts/gate.tsand chose not to, since the test now fits the budget.Closes #12884