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

Full suite red: test/console-clean.test.ts #12895

Merged opened this in answer to

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

Merged

test/console-clean.test.ts now checks the addresses a page names in parallel instead of one after another. The signed-in home test that failed in the full suite took 0.38 s when run alone here, against the 2.89 s the report saw under load. What changed: only test/console-clean.test.ts.

Notes

shipped production is serving it deploy 5a4f606

opened this

test/console-clean.test.ts now checks the addresses a page names in parallel instead of one after another. The signed-in home test that failed in the full suite took 0.38 s when run alone here, against the 2.89 s the report saw under load.

  • What changed: only test/console-clean.test.ts. The assertions are the same: each address must still answer 2xx, with any other status failing the test as before. Only the fetch order changed.
  • What I chose not to do: I did not add the file to SLOW in scripts/gate.ts, since the test is fast now. I also did not touch src/.
  • Checks run: bun scripts/gate.ts and bun x tsc --noEmit are clean. All 8 tests in the file pass, and a three-times rerun passes too.
  • Not reproduced under load: I did not run it under the full suite's concurrency, so the 2.89 s case is untested here.
  • Sandbox note: the sandbox logs a denied outbound request to maria--forge.test:443 during the run. It did not affect the results.

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

Closes #12886

commented

test/console-clean.test.ts now checks the addresses a page names in parallel instead of one after another. The signed-in home test that failed in the full suite took 0.38 s when run alone here, against the 2.89 s the report saw under load.

  • What changed: only test/console-clean.test.ts. The assertions are the same: each address must still answer 2xx, with any other status failing the test as before. Only the fetch order changed.
  • What I chose not to do: I did not add the file to SLOW in scripts/gate.ts, since the test is fast now. I also did not touch src/.
  • Checks run: bun scripts/gate.ts and bun x tsc --noEmit are clean. All 8 tests in the file pass, and a three-times rerun passes too.
  • Not reproduced under load: I did not run it under the full suite's concurrency, so the 2.89 s case is untested here.
  • Sandbox note: the sandbox logs a denied outbound request to maria--forge.test:443 during the run. It did not affect the results.

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

Approved

Test-only change; src/ is untouched, so there is nothing for a person to see in the preview and no security, TUI or docs surface is affected. The assertion is equivalent to the old one: 2xx passes, anything else must read 200 and fails with the same "<ref> <status>" message. It fits the pipeline-first direction (fast tests that never flake). I did not run it, and the builder did not reproduce the 2.89 s case under full-suite load, so the speed-up is argued, not shown.

Two things I would do next:

  1. Bound the concurrency. Promise.all fires every ref at once. That is fine at today's page sizes, but a page that names 60 addresses would add load to the very run that is already loaded. A small pool of 8 or so would keep the speed-up without that risk.
  2. Return {ref,status} from the map instead of joining and splitting on a space. It is cleaner, and it cannot misparse a ref that contains a space.
Sign in to join this conversation.

↑↓ navigate ↵ open ⌘↵ its app esc close

Keyboard shortcuts

Docs