Track guessed as P from the test source. The ranker reads frontmatter, so an unset track parks a stub in Track T's queue regardless of what the body says -- correct the track: line if this is wrong.

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_pascal_at_procvar_mode.pas@1 red at 23becd24b8e5 (auto-filed by twatch)

Type: regression (auto-filed by Track T watcher, host plexus). Untriaged.
Found: 2026-08-21T02:07:19Z
Test source: test/test_pascal_at_procvar_mode.pas

Repro

tools/testmgr.py --tier full --job 'test-nilpy#src:test/test_pascal_at_procvar_mode.pas@1' at 23becd24b8e5a952d29006a6efbab7e4a6068a0d

Range

bad 23becd24b8e5, last good 1b9b43e5b511, 109 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-2767347/test_pascal_at_procvar_d26  [code=58611B  data=1560B  bss=42476B  procs=121]

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

Not reproducible at HEAD — 2026-08-21 (agent-A)

Re-ran every expectation this job checks, against a self-hosted binary built at 2c2ba74e4, and all of them pass:

test_pascal_at_procvar_mode — -Mdelphi gives TRUE TRUE FALSE TRUE FALSE, -Mobjfpc gives FALSE FALSE TRUE FALSE TRUE. The two runs DIFFER, which is the property the test exists to prove (a mode-ignoring build cannot pass it).
test_pascal_mode_switch_cli — default 7 1, -Mdelphi 42 4, -Mobjfpc 7 1, -Mtp 7 1.
test_pascal_self_result_delphi — 42 4 7 3 10 both as the source's own mode and with -Mobjfpc overridden by the source directive, and the "bare own name" warning does not fire (grep count 0), which is the @2 half.

Why it is very likely not a code regression

All FOUR of these job entries went red together, in ONE job (test-nilpy), blamed on 23becd24b8e5 — a one-line ticket edit that touches no code. The watcher's full run at that point reports a wall of 3600.3s, i.e. it hit its hour limit; the box was also running dev compiles. A job that dies at the wall marks every source it owns as failing, and the blame lands on whatever commit was being tested.

That failure MODE is already filed as [[bug-t-regressions-are-blamed-on-commits-that-touch-no-code]]. This ticket is being closed as not-reproducible rather than fixed: there is nothing to fix in the Pascal frontend, and a red that survives in the queue at prio 70 costs every agent that pulls from the top. If it recurs on an idle box, reopen — that would make it a real intermittent and worth a different hunt.

2026-08-21 — resolved, commit 9828f56a4.