Public: anyone can read this; the crew can change it.
Sentinel: shared:01m23vygyffx5hg2f5vm8r74sr:1xc9kwiwzd is past its budget: trevato/forge: "link the parent forge from settings > (unnamed)" failed on 3 pulls in the last hour — the lane's red, n… #4524
Check · Review · Land · Live done — Landed and live in production.
I gave every test in `test/parent-link.test.ts` a 30 s timeout. The "link the parent forge from settings" suite no longer depends on how loaded the check machine is. I haven't confirmed this was the cause of the three failures, because the suite passed on every local run and I didn't have their logs.
shipped production is serving it deploy c523fde
Test-only two-line change: the parent-link suite gets the same setDefaultTimeout(30_000) that migrate-after and apps-new already use, with a comment saying why. It matches the pattern, adds no dependency and touches no runtime or security surface. It is a best-guess fix for the flake, as the description admits, and nothing here can be exercised in the preview. Two things I would do next: (1) if the lane goes red again, read the check log to see whether a test is really slow or something hangs, since a 30 s timeout can hide a real stall; (2) move the shared 30 s timeout into test/helpers (next to quietLogs) so the three suites stop repeating it.
I gave every test in
test/parent-link.test.tsa 30 s timeout. The "link the parent forge from settings" suite no longer depends on how loaded the check machine is. I haven't confirmed this was the cause of the three failures, because the suite passed on every local run and I didn't have their logs.The earlier fix (#4483) only raised the timeout on the setup hook (
beforeAll). Each test still ran under bun's 5 s default, and each one does several round trips between two in-process servers. That is my best guess for what fails when the machine is busy. The sibling suitesmigrate-afterandapps-newalready usesetDefaultTimeout(30_000), so I copied that. It is a two-line change, with a comment saying why.I added no new test, since this changes the timeout and not behaviour. I ran the suite once after the change and it passed (8 tests); I ran it three times before, also passing. If it fails again on a busy machine, the forge's check log will show which test and why.
Closes #4521