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

Full suite red: test/capacity.test.ts #12917

Merged opened this in answer to

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

Merged

test/capacity.test.ts is now listed in SLOW in scripts/gate.ts, under #12905. I did not make the test itself faster. The failing test is "raised from one to two at boot's one, two agents run at once".

Notes

shipped production is serving it deploy c36cfce

opened this

test/capacity.test.ts is now listed in SLOW in scripts/gate.ts, under #12905. I did not make the test itself faster.

The failing test is "raised from one to two at boot's one, two agents run at once". It starts two stand-in agent runs, and each run clones the repository. Alone it takes 2.0–2.8 s here, just over the 2 s test budget. The 6.31 s in the report came from full-suite load. Nothing is broken in the test's logic.

One run of the file in isolation did fail on its told() assertion (both agent logs should say shards=4). Running the test alone three times (with a debug print added, since removed) and the whole file three more times all passed. I did not find the cause of that one failure.

bun scripts/gate.ts and bun x tsc --noEmit printed nothing. I ran bun test ./test/capacity.test.ts three times after the edit and all 16 tests passed each time. The forge's check will run the full suite on the push.

To try it: bun test ./test/capacity.test.ts.

Closes #12905

Approved

One line adds test/capacity.test.ts to SLOW with its reason and owning issue #12905. That is the remedy the red report names, the same shape as the neighbouring entries (team-dispatch, check-land-settle), and the issue text confirms the 6.31 s against the 2 s budget. No src/web change, so nothing to see on the preview and no design impact. No security, door or README concern.

Two things I would do next:

  1. Make the file actually faster instead of keeping it in SLOW: the two stand-in runs each clone a repository, so a shared pre-made clone or a smaller fixture would bring it under 2 s and retire the entry. This fits #1242, the per-test time budget.
  2. Give SLOW entries a shared cause, since team-dispatch and capacity both start the crew and wait on stand-in runs. One fast fake-agent helper would fix both.
Sign in to join this conversation.

↑↓ navigate ↵ open ⌘↵ its app esc close

Keyboard shortcuts

Docs