regression: test-core#src:test/test_string_to_pchar_auto.pas red at 8997639f144f (auto-filed by twatch)

• Type: NOT a regression — the shared-/tmp SQLite race.
• Status: done — fixed 2026-07-13 in ad5d9c89.
• Found: 2026-07-13T01:51:35Z
• Test source: test/test_string_to_pchar_auto.pas

Repro

tools/testmgr.py --tier full --job 'test-core#src:test/test_string_to_pchar_auto.pas' at 8997639f144fc690145e51c980d5c50b94986a89

Range

bad 8997639f144f, last good 8997639f144f, 0 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-2417024/string_to_pchar_auto26  [code=48921B  data=1488B  bss=9520B  procs=578]

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

CLOSED 2026-07-13 — same shared-/tmp race, fixed in ad5d9c89

test_string_to_pchar_auto.pas hardcoded its SQLite database path (/tmp/test_string_to_pchar_auto26.db). testmgr runs core jobs in PARALLEL, and optdiff runs the -O0/-O2/-O3 builds of one source CONCURRENTLY — so runs collided on a single database file, and the resulting output difference was reported as a failure.

Note the "log tail" the watcher captured: it shows a SUCCESSFUL compile and nothing else. That is what a race looks like in a report — the signal is the absence of a real error.

Fixed together with its two siblings (test_sqlite_crud.pas, test_sqlite_crud_lazy.pas) in ad5d9c89: the DB path now comes from ParamStr(0), which is unique per binary. They were all fixed at once precisely so this ticket would not re-file itself from the next sibling under a new SHA.

Same family as [[regression-optdiff-shard4-6]] and [[project_borg_red_harness_race_not_regression]].

Verified: test-core green, --tier opt 11/11 GREEN, --tier full 1214/1214 GREEN.