[schubydoo]

What

test_reaper_kills_an_abandoned_stub flaked on the musl CI leg on 2026-10-31 (job 124507758854) with AssertionError: assert 22744 in [21406, 21463, 21627]. The reaper did not list the stub that the test had just spawned. The re-run passed.

The cause is a race in the test, not a reaper bug. Popen returns after the first exec, which is /usr/bin/env from the stub's shebang. env then execs python3, and while that exec is in flight, /proc/<pid>/cmdline reads empty. reap_fixture_stub_strays selects stubs by their argv, so a reap in that window skips the stub. Locally, 67 of 300 reads right after Popen returned an empty argv.

_spawn_stub now waits, up to 30 s, for <debug-file>.argv.json. run_bridge writes that file before anything else, which proves that the exec chain is done. If the stub exits or the time runs out, the helper kills it and fails loudly. With the wait, 300 of 300 reads saw the stub, and the test passed 40 of 40 local runs. Only tests/test_stub_reaper.py changes.

Checklist

• PR title follows Conventional Commits.
• no-changelog label applied (tests only).
• Tests: this PR is the test fix.
• Docs: no user-facing change.
• just check passes locally (7641 passed).

🤖 Generated with Claude Code

[greptile-apps]

Confidence Score: 5/5

[Low impact] The PR appears safe to merge.

Summary

The PR makes _spawn_stub wait until the fixture reaches its Python code before testing the reaper.

• The reaper test waits until its stub has started before checking for it.

_{Reviews (1) · Last reviewed commit: "test(reaper): wait for the stub to run i..." · Reviewed by Greptile}