hermes-agent/infographic/ci-file-timeout-300
Teknium b508d4296e
test(ci): raise per-file timeout 140s → 300s to stop false timeouts (#54143)
* test(ci): raise per-file timeout 140s to 300s to stop false timeouts

The per-file parallel runner caps each test-file subprocess at a flat
wall-clock budget. Combined with per-test subprocess isolation (a fresh
Python process per test), a large-collection file pays N x (interpreter
startup + import) of overhead before any test logic runs. That overhead
dilates under load on shared CI runners, so a file that finishes in
~100s on a quiet box can blow the old 140s cap purely from scheduling
jitter, surfacing as a false 'no tests ran' timeout (rc=124) with zero
actual test failures.

Raise the default to 300s (5 min). The Docker build matrix jobs already
take 7-10 min, so this headroom costs nothing on total CI wall time
while still bounding a genuinely hung file.

* docs: add infographic for CI per-file timeout bump
2026-06-28 02:41:07 -07:00
..
infographic.png test(ci): raise per-file timeout 140s → 300s to stop false timeouts (#54143) 2026-06-28 02:41:07 -07:00