← board

Track guessed as B from the FAILING STEP — line 1 of 2, stable_linux_amd64/default/pinned -Fulib/rtl/platform/posix test/lib_dns_facade.pas /tmp/lib_dns_facade, which names test/lib_dns_facade.pas. Not from the job's name or its src: those describe what the job is ABOUT, and this job's recipe spans 2 source file(s). The ranker reads frontmatter, so this line — not the body — decides who works it; correct it if the guess is wrong.

This commit CANNOT be the cause. The job builds only with $(PXX_STABLE), and this commit moved no stable_linux_amd64/** — so the bytes that compiled it are unchanged. Look at flakiness or box load, not at the named sha; the bisect is unsound here and has been skipped.

origin/master has advanced 12 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: lib-test#src:test/lib_dns_facade.pas at 021cd94f10a9 in step 1/2, stable_linux_amd64/default/pinned -Fulib/rtl/platform/posix test/lib_dns_facade.pas /tmp/lib_dns_facade (auto-filed by twatch)

Repro

tools/testmgr.py --tier full --job 'lib-test#src:test/lib_dns_facade.pas' at 021cd94f10a97c051c6c71d75d308028adc34c43

Range

The named sha 021cd94f10a9 CANNOT be the cause — it touches no buildable file (docs / tickets / tstate only). It is the sha that was TESTED, i.e. the upper bound of an untested range; the cause is somewhere below it.

bad 021cd94f10a9, last good 1e37a55f6748, 14 observable commit(s) in range (it builds with $(PXX_STABLE), so compiler/ commits cannot have caused it and are dropped; pin moves, lib/ and test/ are kept) — the watcher narrows this by idle bisect.

Log tail

pascal26:47: error: undefined variable (PalVfork)
(tail)
pascal26:47: error: undefined variable (PalVfork)
  near: boundP ) ; pid := PalVfork >>> ; if pid

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

2026-09-01 (frankH) — resolved: a deliberate PAL rename, in EIGHT files not one

e2ba5a1e1 renamed the PAL entry PalVfork -> PalFork on purpose: the old name was false, and it had already cost a real conclusion — lib/crtl/src/unistd.c reasoned FROM the name that the PAL had no fork(), and left fork() an ENOSYS stub, against a PAL whose body had issued SYS_fork all along (busybox ash then failed with can't fork). The rename is right; the test callers were missed.

The ticket named one file. Grepping the shape first found eight, ten call sites: lib_dns_{resolve,facade,spoof,tcp,multins,chase}.pas plus lib_dns_cache_facade.pas and lib_dns_aaaa.pas with two each. Fixing only the filed file would have moved the red one row down and looked like a second bug.

Pure rename, same signature (function PalFork: Integer). PalVforkAndExec is deliberately untouched — it keeps its name per platform_backend.pas:1408 — and \bPalVfork\b does not match inside it.

Verified: make lib-test EXIT=0 end to end against stable v399, with synapse-ssl and reportlab-diff declared SKIPPED by the summary line itself.

For whoever tunes the auto-filer: both banners were correct and both pointed away

The cause was a rename in lib/rtl plus its test/ callers. The heuristic drops compiler/ commits from a $(PXX_STABLE) job's bisect range, which is sound, but the fallback advice it then prints assumes the residue is environmental. It is not: lib/ and test/ are still in range and are exactly where a $(PXX_STABLE) job's real regressions live. Suggest the fallback say "look in lib/ and test/ within the range" before it says "flakiness".