regression: test-aarch64#src:test/test_cross_sysopen_family.pas red at a5fc06ee29b6 (auto-filed by twatch)

• Type: regression (auto-filed by Track T watcher, host borg). Untriaged.
• Found: 2026-07-13T05:54:30Z
• Test source: test/test_cross_sysopen_family.pas tools/run_target.sh

Repro

tools/testmgr.py --tier full --job 'test-aarch64#src:test/test_cross_sysopen_family.pas' at a5fc06ee29b63a859c5a8f0ac17ad2e8435c7144

Range

bad a5fc06ee29b6, last good a5fc06ee29b6, 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-3463314/test_aarch64_sysopen_family  [code=74812B  data=176B  bss=9468B  procs=65]
ok: /tmp/testmgr-scratch-3463314/test_aarch64_sysopen_family_x64  [code=32928B  data=224B  bss=9468B  procs=65]

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

Triage 2026-07-13 (Track A) — NOT A REGRESSION, transient

Green at HEAD: tools/testmgr.py --tier full --job 'test-aarch64#src:test/test_cross_sysopen_family.pas' → 1/1 pass. The watcher itself already recorded it as FIXED one SHA later (tstate/reports/20260713T060754Z-9b93d1c-borg.md, verdict GREEN) with no intervening change to the aarch64 backend or to that test.

Three things say "harness", not "codegen":

• the range is bad a5fc06ee, last good a5fc06ee — 0 commits, i.e. the same commit was both good and bad;
• the log tail ends with BOTH binaries emitted ok: and then nothing — no wrong output, no diff, no non-zero exit line. A miscompile fails loudly at the compare; a killed/timed-out qemu run fails exactly like this;
• a5fc06ee (advanced-record constructors) is a frontend commit that cannot reach the aarch64 backend.

Same shape as the earlier cjson/lua borg reds that turned out to be a scratch-dir race, so the suspicion is the run side (qemu launch under the memory scope), not the emitted code. Closing as rejected rather than done: nothing was fixed. If it comes back, instrument the harness (capture qemu's exit status and stderr into the log tail) before suspecting the backend — the current log tail cannot distinguish "ran and printed the wrong thing" from "never ran".