Repository navigation
Commit 93e3bc3
committed
fix(gate): check_vm_parity portable timeout — macOS runner ships no
Once the surface_marker false-positive stopped short-circuiting ci_gate.sh
before its step 3, the nightly macOS full gate surfaced 382 tree-walk e2e
"failures", every one with an empty diff. Root cause: check_vm_parity.sh
wrapped each e2e in raw `timeout 180 bash "$t"`, but hosted macOS runners ship
neither GNU `timeout` nor coreutils `gtimeout` → "command not found" (exit 127)
→ each e2e counted as a FAIL whose output matched no FAIL/error pattern (hence
the empty `()`). corpus passed because check_corpus_regression already uses the
run_bounded fallback. This was the last raw `timeout` caller missed by commit
264804b's sweep; ported to the same run_bounded helper (timeout → gtimeout →
unbounded). Reproduced the runner condition locally (timeout/gtimeout absent):
raw timeout → exit 127; run_bounded → exit 0, e2e green.
Smell-audited: 1: test-harness portability, no runtime logic change; verified via a clean-PATH macOS-runner simulation + smoke green.timeout
1 parent 9c78920 commit 93e3bc3
1 file changed
Lines changed: 14 additions & 1 deletion
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
35 | 35 | | |
36 | 36 | | |
37 | 37 | | |
| 38 | + | |
| 39 | + | |
| 40 | + | |
| 41 | + | |
| 42 | + | |
| 43 | + | |
| 44 | + | |
| 45 | + | |
| 46 | + | |
| 47 | + | |
| 48 | + | |
| 49 | + | |
| 50 | + | |
38 | 51 | | |
39 | 52 | | |
40 | 53 | | |
| |||
60 | 73 | | |
61 | 74 | | |
62 | 75 | | |
63 | | - | |
| 76 | + | |
64 | 77 | | |
65 | 78 | | |
66 | 79 | | |
| |||
0 commit comments