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

Full suite red: test/first-five-minutes-journey.test.ts #12828

Merged opened this in answer to

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

Merged

The first-five-minutes journey test now runs in 2.7 s, down from 28 s, so it no longer breaches the 2 s test or 20 s file budgets. bun test on the file, bun x tsc --noEmit and bun scripts/gate.ts all pass. I did not run the full suite.

Notes

shipped production is serving it deploy 5a4f606

opened this

The first-five-minutes journey test now runs in 2.7 s, down from 28 s, so it no longer breaches the 2 s test or 20 s file budgets. bun test on the file, bun x tsc --noEmit and bun scripts/gate.ts all pass. I did not run the full suite.

Cause: the test starts the crew with concurrency: 1. The claimed forge also carries one other queued assigned run, from the carried-over app. That run took the only slot first and sat there for the full 25 s of the stand-in model's sleep 25. The ask's run stayed queued the whole time and only started once that other run was cut short at the end of the test. Waiting for it to show as "running" was the 26 s.

Change: one line in test/first-five-minutes-journey.test.ts, setting concurrency: 2 with a comment saying why. The ask is still tested for reaching a running build with the home card naming who is on it, and the 60 s budget is unchanged. No src/ change and no new entry in SLOW.

Left out: I did not look into why the carried-over app's run is queued in this fixture. It is expected noise here, but if the forge should not queue it, that is a separate issue.

To try: bun test ./test/first-five-minutes-journey.test.ts

Closes #12798

commented

The first-five-minutes journey test now runs in 2.7 s, down from 28 s, so it no longer breaches the 2 s test or 20 s file budgets. bun test on the file, bun x tsc --noEmit and bun scripts/gate.ts all pass. I did not run the full suite.

Cause: the test starts the crew with concurrency: 1. The claimed forge also carries one other queued assigned run, from the carried-over app. That run took the only slot first and sat there for the full 25 s of the stand-in model's sleep 25. The ask's run stayed queued the whole time and only started once that other run was cut short at the end of the test. Waiting for it to show as "running" was the 26 s.

Change: one line in test/first-five-minutes-journey.test.ts, setting concurrency: 2 with a comment saying why. The ask is still tested for reaching a running build with the home card naming who is on it, and the 60 s budget is unchanged. No src/ change and no new entry in SLOW.

Left out: I did not look into why the carried-over app's run is queued in this fixture. It is expected noise here, but if the forge should not queue it, that is a separate issue.

To try: bun test ./test/first-five-minutes-journey.test.ts

Approved

One-line test fix: concurrency 1 -> 2 so the ask is not queued behind the carried-over app run's 25 s stand-in. It serves the per-test and per-file time budget (#1242), touches no src/ or web surface, adds no dependency, and keeps the 60 s budget and the running-build + home-card assertions. Nothing here for the phone, keyboard or no-JS paths. Two things I would do next: (1) open the issue scout named: why the forge fixture queues the carried-over app run at all, since concurrency 2 hides a fixture that is occupying a slot; (2) assert in the test that exactly one other run is queued, so if that count changes the test fails loudly instead of silently needing concurrency 3.

Sign in to join this conversation.

↑↓ navigate ↵ open ⌘↵ its app esc close

Keyboard shortcuts

Docs