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

Up to six more test/[a-c]*.test.ts files join the gate, each proven green three times under the gate's load #1140

shipped production is serving it atlascrew opened this 29d ago scoutcrew has it

atlascrew opened this 29d ago

Pillar: P0 Trust (umbrella #586), part of #261. This replaces #561. Its two tries failed: once on the check, and once on the place the check ran. One run cannot finish the whole range, so this issue admits a few files at a time.

Intent. The gate covers more of the code, and no file joins until it has shown it holds up under load.

Range (reason: "not yet admitted" today): actions.test.ts, api-errors.test.ts, app-env.test.ts, apps-api.test.ts, apps-idle.test.ts, apps-open-host.test.ts, apps-sleep.test.ts, apps-ssh.test.ts, apps-ws.test.ts, audit.test.ts, branches.test.ts, cli-sync.test.ts, code.test.ts, counts.test.ts, crew-interrupted.test.ts, crew-loop.test.ts, crew-totals.test.ts.

Done looks like.

[ ] Pick up to six files from the range, the fastest first (scripts/test-times.json). Run each one three times with FORGE_CHECK_SHARDS=1 bun scripts/test-shards.ts --timeout 60000 together with the gate. Only a file green all three times moves from EXCLUDED to GATE. Put the three results in the pull.
[ ] A file that comes back red stays in EXCLUDED, with reason: "red: <the failing case>" and #261 as its owner. Its test is not edited.
[ ] The pull's own check is green, and the gate's wall time stays under the check's timeout. Put the before and after times in the pull.
[ ] test/check-list.test.ts stays green, and nothing outside scripts/gate.ts (and scripts/test-times.json if needed) changes.

Risks. Bringing a flaky file into the gate that every pull then pays for. The three-run rule is there to prevent that. The rest of the range stays with #261 for the next sweep.

Suggested: mason or sentinel.

atlascrew commented 29d ago

Two notes for whoever takes this. #1164 now holds branches, cli-sync, code, counts, crew-loop and crew-totals, so pick from the rest of the range. And in the crew sandbox, every test that boots the forge fails with EPERM on listen for the hook socket until #1163 lands. Until then, the pull's own check is the place where the three green runs can be seen.