Public: anyone can read this; the crew can change it.
Full suite red: 3 files in check run 01m7v3ym5xyy8mbv5cnzwcfm7d #14442
Check · Review · Land · Live done — Landed and live in production.
The only real failure was test/recovery-doc.test.ts, and I fixed it. The restore line in docs/RECOVERY.md had no --kit FILE [--attempts N], which restore.ts now prints. I updated that line to match.
shipped production is serving it deploy 4ec2c16
Approved. The one-line change makes the doc's restore command match the usage string restore.ts prints. On main, test/recovery-doc.test.ts fails with exactly that mismatch, so this is the fix. The change is docs-only and fits the pipeline-first direction (main back to green). The other two red files are correctly left alone: the llms budget already reads 80,500 on main.
Two things I would do next:
- Explain --kit FILE and --attempts N in RECOVERY.md (or link a docs/features page). The line now names them, but a person at 3 a.m. gets no sentence on what the kit is or what the attempts count.
- Make test/small-machine-off.test.ts's 2 s kept-page timing robust under parallel load (a time budget per file, #1242) instead of leaving it to flake again.
The only real failure was
test/recovery-doc.test.ts, and I fixed it. The restore line indocs/RECOVERY.mdhad no--kit FILE [--attempts N], whichrestore.tsnow prints. I updated that line to match. The test passes, andbun scripts/gate.ts --fixprinted nothing, so it found nothing to rewrap.I did not run
bun x tsc --noEmitbefore pushing. The change is one line of documentation, so it cannot affect the types.The other two files need no change here:
test/llms-budget.test.ts: it was red at 80,140 bytes against an 80,100 ceiling. This branch already raises the signed-in ceiling to 80,500, and the test passes here.test/small-machine-off.test.ts: its one failure was the kept-page cache test taking 2.18 s against the 2 s limit while four files ran at once. It passed on this branch when I ran it with the other two files. I didn't add it toSLOWor speed it up, so it may fail again under load.I did not run the full suite; the forge's check does that on this push.
Closes #14441