← board

Track A from the job NAME test-asm, not from its source. This job names a MECHANISM rather than a subject — the source it was fed (test/test_asm_emit_x64.pas) is what the mechanism was run ON, not what is being tested, so a lane guessed from it would be wrong by construction. The ranker reads frontmatter, so this line decides who works it; re-lane it if this job has changed what it covers.

origin/master has advanced 1 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-asm#src:test/test_asm_emit_x64.pas at a0c5a21e2c2e in step 1/2, ./compiler/pascal26 -Fucompiler test/test_asm_emit_x64.p (auto-filed by twatch)

Repro

tools/testmgr.py --tier native --job 'test-asm#src:test/test_asm_emit_x64.pas' at a0c5a21e2c2eb04f63554d4801e1655e2c548f3c

Range

The named sha a0c5a21e2c2e 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 a0c5a21e2c2e, last good c04a6666bdd7, 1 commit(s) in range — the watcher narrows this by idle bisect; check tstate/TSTATE.md for the current range.

Log tail

pascal26:1088: error: undefined variable (PxxDbgArg)
pascal26:1089: error: undefined variable (DwArmsOff)
pascal26:1091: error: undefined variable (DwTotalCalls)
pascal26:1092: error: undefined variable (DwElseHits)
pascal26:1092: error: undefined variable (DwElseAns)
pascal26:1093: error: undefined variable (DwBackHits)
pascal26:1093: error: undefined variable (DwBackAns)
(tail)
pascal26:1088: error: undefined variable (PxxDbgArg)
  in: compiler/asmtext.inc
  near: n do begin if items [ >>> i ] . 
pascal26:1089: error: undefined variable (DwArmsOff)
  in: compiler/asmtext.inc
  near: <> vtAnsiString then Error ( 'EmitAsmX64: expected an instruction string' >>> ) ; line 
pascal26:1091: error: undefined variable (DwTotalCalls)
  in: compiler/asmtext.inc
  near: ( items [ i ] . >>> VAnsiString ) ; 
pascal26:1092: error: undefined variable (DwElseHits)
  in: compiler/asmtext.inc
  near: ] . VAnsiString ) ; trimmed >>> := AsmTextTrim ( 
pascal26:1092: error: undefined variable (DwElseAns)
  in: compiler/asmtext.inc
  near: ; trimmed := AsmTextTrim ( line >>> ) ; j 
pascal26:1093: error: undefined variable (DwBackHits)
  in: compiler/asmtext.inc
  near: ( line ) ; j := >>> 1 ; while 
pascal26:1093: error: undefined variable (DwBackAns)
  in: compiler/asmtext.inc
  near: j := 1 ; while ( >>> j <= Length

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

Log

Resolution — 2026-09-01, frankB (Track A)

Re-verified GREEN at HEAD 00c5b90a4be5, as the sha banner above asks:

./compiler/pascal26 -Fucompiler test/test_asm_emit_x64.pas <tmp>/sweep_x64
  -> ok: ... ALL X64 ASM EMIT TESTS PASSED   exit=0

Not intermittent, and the 2026-09-01 log note above can stop worrying about that. The undefined-variable list in the tail (DwArmsOff, DwTotalCalls, DwElseHits...) names the deref-walk counters I added in defs.inc, reached from compiler/asmtext.inc. That file has a SECOND including configuration -- the standalone test_asm_emit_* harnesses include the .inc directly against a mocked defs.inc, where those counters do not exist. So the break was a real one-way build error in a configuration the main build never exercises, not a flake, and it is gone because the code that caused it moved out of asmtext.inc in 636e4470d.

The class now has a guard: tools/standalone_inc_harnesses.sh compiles all seven such harnesses, wired into gate.sh conditionally on the two-configuration .inc files being touched.