← board

Track T by default: the FAILING STEP named no owner. Line 2 of 4 is tools/expect_same.sh test_nilpy_to_bytes26 "$(/tmp/test_nilpy_to_bytes26)" "$(printf '8\n10\n0\n10\n254\n255\n-2\n255\n0. The job's own src (test/test_nilpy_to_bytes.npy, 2 file(s)) is NOT used here on purpose: it is what the job compiles, not what broke, and guessing a lane from it is what sent three reds in one job to the wrong lane. This is a FALLBACK, not a finding — nothing says the defect is Track T's. Re-lane it before working it.

The SLUG names test_nilpy_to_bytes, and that is not what broke. The slug is derived from the job's src: selector so that it stays stable across a renumbering — it is the dedupe key and the close key — but src: says what the job is ABOUT. The failing step names expect_same. Read the Failing step: bullet, not the file name in the title, before you reproduce anything: the row the slug names may be passing.

origin/master has advanced 5 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_nilpy_to_bytes.npy at 5dbee723e228 in step 2/4, tools/expect_same.sh test_nilpy_to_bytes26 "$(/tmp/test_nilpy_to_bytes26)" "$(printf '8\n10\n0\n10\n254\n255\n-2\n255\n… (auto-filed by twatch)

Repro

tools/testmgr.py --tier full --job 'test-nilpy#src:test/test_nilpy_to_bytes.npy' at 5dbee723e228cfb986a5561e78ccffb7e9a153ec

Range

bad 5dbee723e228, last good 1704e17de6f3, 1 commit(s) in range — the watcher narrows this by idle bisect; check tstate/TSTATE.md for the current range.

Log tail

ok: /tmp/testmgr-scratch-3707340/test_nilpy_to_bytes26  [code=1445656B  data=87338B  bss=56740B  procs=2432]
expect_same: MISMATCH [test_nilpy_to_bytes26]
--- expected
+++ actual
@@ -26,7 +26,7 @@
 71
 201
 10
-8 101
+8 0
 9
 8 0 1
 1 70

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

Log

CENSUS 2026-09-19 (frankS) — does not reproduce; closed by events

Run through testmgr's own job runner, using this ticket's exact Repro line — the instrument that filed it, and the one auto-pin reads. Not through make, and not through a hand-run of the fixture:

tools/testmgr.py --tier full --job <this ticket's own literal job selector>
->  testmgr: GREEN, 1/1 pass

All 17 open NilPy regressions were run that way and all 17 came back GREEN.

The census carries a positive control drawn from the same population, because seventeen greens from an instrument nobody has shown can fail are not evidence. Same runner, same tier, on the xmlreader job — which is still built by the PINNED compiler and therefore still broken — the identical form returns:

expect_same: MISMATCH [lib_mimic_xmlreader.1]
--- expected
+++ actual
@@ -1 +1 @@
-25
+24

testmgr: RED

So GREEN here means the job passed, not that the runner is blind.

This does not say the report was never real. It was real at its filing sha; the tree has moved past it. Closed as NOT REPRODUCING, so it stops occupying a ranked slot and stops being dispatched to.

Why a whole pile went stale at once, which is the part worth keeping: these are auto-filed by the Track T watcher, and the watcher DOES retire them — 19 open tickets have a twin in done/ and every one of those twins carries the line auto-closed by the borg watcher. The defect is narrower than "nothing closes them": the auto-close WRITES the closed copy into done/ and does not remove the backlog/ original. The duplicate then keeps a real prio, so it goes on sorting alongside live work and a seat gets dispatched to a subject that has been passing for weeks — and progress.sh resolve refuses it as an ambiguous slug, which is how the pattern surfaced at all.