← board

origin/master has advanced 3 commit(s) since this sha. Re-verify at current HEAD before acting — the callback is tagged to the sha that was tested, which may no longer be the state of the tree.

regression: test-nilpy#src:test/test_cpyext_args_errors.npy red at 34c41bde6fd6 (auto-filed by twatch)

Repro

tools/testmgr.py --tier full --job 'test-nilpy#src:test/test_cpyext_args_errors.npy' at 34c41bde6fd66529206b2891337066a5a9fae50c

Range

bad 34c41bde6fd6, last good unknown, 0 commit(s) in range — the watcher narrows this by idle bisect; check tstate/TSTATE.md for the current range.

Log tail

pascal26:16: error: uses: unit source not found: /lib/cpyext/src/pyruntime.c
(tail)
pascal26:16: error: uses: unit source not found: /lib/cpyext/src/pyruntime.c
  near:  interface uses pxxcio  ../../lib/cpyext/src/pyruntime.c >>>  ./argerr_ext_module.c

Stub ticket: signal only. Track T agent (face 2) enriches or a dev track takes it from the repro line.

2026-08-06 — STALE STUB, closing: the watcher itself already reported it FIXED

Not a live red. The whole cpyext family (args_errors, containers, cython, hello, markupsafe) went red together at 34c41bde6fd6 and the watcher's very next report — tstate/reports/20260805T203501Z-aba953c-plexus.md, about an hour later — lists all five under FIXED. The auto-filed stub was never closed, so it kept surfacing at the head of the ranked queue.

Re-verified at HEAD 733be3321 (compiler snapshot sha256 cafd50517875), all five green:

PASS  test-nilpy#410  test/test_cpyext_hello.npy
PASS  test-nilpy#411  test/test_cpyext_args_errors.npy
PASS  test-nilpy#412  test/test_cpyext_containers.npy
PASS  test-nilpy#413  test/test_cpyext_markupsafe.npy
PASS  test-nilpy#414  test/test_cpyext_cython.npy

Also checked the failure mode itself rather than only the exit status. The log tail was uses: unit source not found: /lib/cpyext/src/pyruntime.c — note the LEADING SLASH, meaning the base directory the unit-relative '../../lib/cpyext/src/pyruntime.c' resolves against came out empty. That is a path-resolution shape worth being suspicious of, so it was probed directly at HEAD: absolute -Fu with cwd outside the repo, and relative -Fu with cwd at the repo root, both compile and link fine. Five tests sharing one uses line going red and green together points at the transient tree state at 34c41bde6fd6, not at the resolver.

No code change. The lesson is the process one CLAUDE.md already names: an async watcher callback is tagged to the sha it was tested at, and must be re-checked against current HEAD before acting — this one had been fixed before anybody read the ticket.

Possible Track T follow-up (NOT filed as a bug, since it may be intended): the watcher files a stub on NEW-RED but does not appear to close or annotate that stub when a later report moves the same job to FIXED, so a self-healing red leaves a permanent prio-70 item at the head of the queue. If that is not already handled, auto-closing would keep the ranked queue honest.