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

Full suite red: test/looks.test.ts #12661

Merged opened this in answer to

Check · Review · Land · Live done — Landed and live in production.

Merged

Both over-budget tests in test/looks.test.ts are changed to run faster. I could only run one of them here. bun x tsc --noEmit and bun scripts/gate.ts showed no errors, and I did not run the full file again after the edit.

Notes

shipped production is serving it deploy b7253a6

opened this

Both over-budget tests in test/looks.test.ts are changed to run faster. I could only run one of them here. bun x tsc --noEmit and bun scripts/gate.ts showed no errors, and I did not run the full file again after the edit.

  • Never-answers test (#11520): it waited a real 2 s (answerWaitMs: 2_000) for an app that never answers, which was the whole 2.20 s. The wait is now 150 ms. The test's assertions are the same: no capture, one recorded reason, and a slow app that answers on the fourth ask still gets its pictures. It passes here.
  • Dark-capture test: it started two browsers one after the other. The light and dark captures now run together with Promise.all, so the wall time should be about half. The assertions on the corner pixels are the same.
  • Not verified: the dark-capture test could not run in this sandbox. Real-browser captures fail here (the same sandbox failure hits two other real-chromium tests in the file), so I did not measure its new time. The forge's check on this push is the first real measure of it.
  • Left out: I did not add test/looks.test.ts to SLOW in scripts/gate.ts, since both tests were made faster instead. If the dark test is still over 2 s on the gate box, the next step is to move it to a *.looks.test.ts file, which the gate tiers don't budget.

To try it: bun test ./test/looks.test.ts --rerun-each 10.

Closes #12637

commented

Both over-budget tests in test/looks.test.ts are changed to run faster. I could only run one of them here. bun x tsc --noEmit and bun scripts/gate.ts showed no errors, and I did not run the full file again after the edit.

  • Never-answers test (#11520): it waited a real 2 s (answerWaitMs: 2_000) for an app that never answers, which was the whole 2.20 s. The wait is now 150 ms. The test's assertions are the same: no capture, one recorded reason, and a slow app that answers on the fourth ask still gets its pictures. It passes here.
  • Dark-capture test: it started two browsers one after the other. The light and dark captures now run together with Promise.all, so the wall time should be about half. The assertions on the corner pixels are the same.
  • Not verified: the dark-capture test could not run in this sandbox. Real-browser captures fail here (the same sandbox failure hits two other real-chromium tests in the file), so I did not measure its new time. The forge's check on this push is the first real measure of it.
  • Left out: I did not add test/looks.test.ts to SLOW in scripts/gate.ts, since both tests were made faster instead. If the dark test is still over 2 s on the gate box, the next step is to move it to a *.looks.test.ts file, which the gate tiers don't budget.

To try it: bun test ./test/looks.test.ts --rerun-each 10.

Approved

Test-only change with no security surface. It trims two over-budget tests and keeps the same assertions. The never-answers wait drops to 150 ms, which is still about 10x the ~15 ms the fourth-ask case needs. The light and dark captures now run together. The pull is green on the forge check. Nothing else is needed.

Sign in to join this conversation.

↑↓ navigate ↵ open ⌘↵ its app esc close

Keyboard shortcuts

Docs