← board

Track guessed as B from the FAILING STEP — line 84 of 346, stable_linux_amd64/default/pinned --mimic-fpc -dPXX_DYNLIB_LIBC -Fuexternal/synapse -Fulib/rtl -Fulib/rtl/platform/posix, which names test/lib_synapse_tls_loopback.pas. Not from the job's name or its src: those describe what the job is ABOUT, and this job's recipe spans 52 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 7 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:tools/crtl_reachability.py at fca28056d8ec in step 84/346, stable_linux_amd64/default/pinned --mimic-fpc -dPXX_DYNLIB_LIBC -Fuexternal/synapse -Fulib/rtl -Fulib/rtl/platform/posi… (auto-filed by twatch)

Repro

tools/testmgr.py --tier full --job 'lib-test#src:tools/crtl_reachability.py' at fca28056d8ec2483d3a9a1b0f064842164f92b46

Range

The named sha fca28056d8ec 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 fca28056d8ec, last good 0e3ba86d5208, 4 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

pascal26:2039: error: undefined variable (SetString)
pascal26:2077: error: undefined variable (SetString)
pascal26:2178: error: undefined variable (SetString)
(tail)
k pinned to: stable_linux_amd64/default/pinned -> stable_pinned   (newest checkpoint: latest -> stable_latest)
frozen builtin RTL: stable_linux_amd64/default/builtin/ (13 src) -- isolates track A's compiler/builtin/ edits
=== lib-test: library smoke against stable_linux_amd64/default/pinned ===
crtl-reachability: OK -- 148 headers, 66 modules, every declared function reachable from its own header
crtl-map: OK -- 613 crtl functions mapped to 63 headers
  lib-units: 144 units compile (0 known-broken skipped)
ok: /tmp/testmgr-scratch-3871455/lib_ipv6  [code=163608B  data=6128B  bss=52028B  procs=555]
ok: /tmp/testmgr-scratch-3871455/lib_net6  [code=171800B  data=7168B  bss=52968B  procs=576]
ok: /tmp/testmgr-scratch-3871455/lib_asyncnet6  [code=171800B  data=6232B  bss=235724B  procs=585]
ok: /tmp/testmgr-scratch-3871455/crtl_exp2  [code=413464B  data=12928B  bss=74188B  procs=963]
  tk-nilpy: ok
ok: /tmp/testmgr-scratch-3871455/lib_urllib_server  [code=1363736B  data=98872B  bss=293084B  procs=2258]
note: urllib_request -> mimic_urllib_request (shim, subset)
note: urllib_error -> mimic_urllib_error (shim, subset)
ok: /tmp/testmgr-scratch-3871455/lib_urllib_client  [code=1761048B  data=125586B  bss=297620B  procs=2594]
note: urllib_request -> mimic_urllib_request (shim, subset)
note: urllib_error -> mimic_urllib_error (shim, subset)
ok: /tmp/testmgr-scratch-3871455/lib_urllib_refusals  [code=1748760B  data=123676B  bss=292404B  procs=2602]
  lib-test: mimic_urllib_request matches CPython
  lib-test: tkhtmlview renders identically under CPython
pascal26:2039: error: undefined variable (SetString)
  in: external/synapse/synautil.pas
  near: ( var APtr : PANSIChar ; >>> AEtx : PANSIChar 
pascal26:2077: error: undefined variable (SetString)
  in: external/synapse/synautil.pas
  near: do begin SearchForLineBreak ( APtr , >>> AEtx , bol 
pascal26:2178: error: undefined variable (SetString)
  in: external/synapse/synautil.pas
  near: <= AEtx ) and ( SynaFpc >>> . strlcomp (

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

2026-09-16, 20:4x UTC — THE CURRENTLY-RECORDED REASON DOES NOT REPRODUCE AT THE TESTED TREE

borg.json gives this job's reason, for the newest run, as:

... 4 more line(s) not shown | lib-units: FAIL mimic_reportlab_pdfgen | pascal26:60:
warning: #include <asm-generic/ioctls.h> resolved from the host system (/usr/include) …

That step is one command and it is reproducible off the tier. Run here at HEAD b160f6305:

tools/lib_units_compile.py                               ->  154 units compile, rc=0
PXX_STABLE=compiler/pascal26 tools/lib_units_compile.py   ->  154 units compile, rc=0
pinned  c599e8546121      HEAD compiler  b57f90696a01

Under the PINNED compiler — the tier's own configuration, which is what the tool defaults to — it passes. So this is not the RTL/compiler split that explains its sibling row lib-test#44 (crtl_atexit.c), where the pinned diagnostic is genuinely stale; that control comes out green here.

And the tree is not the variable. git diff --name-only acbc6fa04482 HEAD is entirely under devdocs/ — no lib/, no compiler/, no test/. The tree that failed on borg and the tree that passes here are identical in every buildable file, so the difference is not something that landed since.

What that leaves, and what it does not. The cause is not in the tree: it is flakiness or something about the box. I am NOT naming which — lib_units_compile.py compiles all 154 units concurrently (min(cpu_count, 16) workers) into one shared temp dir, and borg has more cores than this box, which makes a race a candidate rather than a finding. Track T owns that residual question; it owns the harness and the box, and neither is visible from here.

SCOPE, because this is one step of a 346-step recipe. This does NOT say the job is green. It says the reason currently published for it does not reproduce at the tree it was published against. The failing step named in the auto-filed body above is a DIFFERENT one — line 84, test/lib_synapse_tls_loopback.pas, with undefined variable (SetString) — from the 2026-09-09 run. A job that reports a different failing step on different runs is the shape a flaky harness makes, and it is also the shape a genuinely moving target makes; nothing here separates those two.

Measured by the toko-watch seat, check-in 2c. Not claiming the ticket; not re-laning it. The track: B guess in the frontmatter was derived from the OLD failing step and may now be wrong for the same reason the job name is.