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

The cell suites pass on a pull's check: spawn-cell-e2e and crew-sandbox #560 are green on today's main #10541

shipped production is serving it opened this has it
opened this

Intent: two pulls for #10405 (#10436, #10492) changed only src/web/pages/origins.ts and a new test, and both went red on the same three cell tests their diff never touched, so unrelated work gets parked.

The reds (check run 01m71msjzc8vvx97tk4dn9xz8k, base 87a718808, diskFree 0.16):

  • a child forge spawned into a cell serves its app at <app>--<child>--<parent> (test/spawn-cell-e2e.test.ts:15)
  • a real child forge > boots in its own forge:<name> cell and answers / through the forward; its first request does not wait for it
  • the shell sandbox denies the whole data directory but the agent's own (#560) > the agent still reaches its own working directory and its own run's temp (test/crew-sandbox.test.ts:247; last touched by 636aa0355, the child temp directory move)

Done:

  • [x] The cause of each red is named in the pull: the code (a regression from 636aa0355 / 0dfe641b7 / 9cecaf275), the test's own fixture, or the machine (disk at 16% free).
  • [x] The code or the test is fixed so all three pass in the forge's full check on a pull that changes nothing in src/core; if the test was wrong, the pull says so.
  • [x] If the cause is load or disk, the test fails with that cause in its message instead of a timeout.

Files: test/spawn-cell-e2e.test.ts, test/crew-sandbox.test.ts, and whichever of src/core/cells*/src/core/crew sandbox code the cause is in. Risks: these suites are skipped in a builder's FORGE_TEST_SOCKET sandbox, so the builder proves the fix through the forge's check on push. Do not loosen the sandbox to make #560 pass: it guards that an agent cannot read the data directory. Crew: a builder with the platform/cells brief; a person assigns.

commented

Reviewed pull #10551 at its head: ticked The cause of each red is named in the pull: the code (a regression from 636aa0355 / 0dfe641b7 / 9cecaf275), the test's own fixture, or the machine (disk at 16% free).; The code or the test is fixed so all three pass in the forge's full check on a pull that changes nothing in src/core; if the test was wrong, the pull says so.; If the cause is load or disk, the test fails with that cause in its message instead of a timeout..

Sign in to join this conversation.

↑↓ navigate ↵ open ⌘↵ its app esc close

Keyboard shortcuts

Docs