Track A from the job NAME test-emit-obj, not from its source. This job names a MECHANISM rather than a subject — the source it was fed (test/c_obj_data_dup_a.c) 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 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: test-emit-obj#src:test/c_obj_data_dup_a.c at 7b781ce15084 in step 25/261, for t in "" "--target=i386" "--target=riscv32 --platform=esp" "--target=xtensa --platform=esp"; do \ ./compiler/pascal2… (auto-filed by twatch)

• Type: regression (auto-filed by Track T watcher, host borg, twatch f35167cd4e55). Untriaged.
• Found: 2026-09-25T00:22:02Z
• Test source: test/c_obj_data_dup_a.c test/c_obj_data_dup_b.c +17
• Failing step: line 25 of 261 of the job's recipe; it names no source file of its own — so it is the JOB's sources, one line up, that are unproven here, not this step's.

  for t in "" "--target=i386" "--target=riscv32 --platform=esp" "--target=xtensa --platform=esp"; do \ ./compiler/pascal26 --emit-obj $t /tmp/fnp_forms.c /tmp/fnp_forms.o >/dev/null || { echo "test-emit-obj: the fn-pointer census fixture FAILED to build for [$t]"; exit 1; }; \ n=$(readelf -sW /tmp/fnp
  
  

Repro

tools/testmgr.py --tier full --job 'test-emit-obj#src:test/c_obj_data_dup_a.c' at 7b781ce150841c4c44a5bbc92a3983a28386db86

Range

bad 7b781ce15084, last good aef4ee1310f9, 33 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-2812258/cods_dup_a.o  [code=18853B  data=1432B  bss=34640B  procs=514  codeseg=18853B]
ok: /tmp/testmgr-scratch-2812258/cods_dup_b.o  [code=18853B  data=1432B  bss=34640B  procs=514  codeseg=18853B]
test-emit-obj: duplicate definition is rejected, as gcc rejects it
test-emit-obj: the fn-pointer census fixture FAILED to build for [--target=i386]

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

Log

• 2026-09-25 — the borg watcher saw test-emit-obj#src:test/c_obj_data_dup_a.c GREEN at dcb660109dad (tier full) and did NOT close this: this is a repeat stub (regression-test-emit-obj-c-obj-data-dup-a-2, not regression-test-emit-obj-c-obj-data-dup-a) — the job already went red, was closed, and came back, so one green is the outcome a live intermittent bug produces most of the time. The green is recorded because it is evidence and because a ticket that stops moving with no reason reads as forgotten; closing this one is a human's call.

2026-10-02 (frankuser): a second flake in the same job group, test-emit-obj#03

In two consecutive full tiers on plexus (nice -n 15, load ~24), #03 failed with an object it had just written already gone:

ok: /tmp/testmgr-scratch-1289354/cods_imp_x64.o  [code=19482B ...]
... five more ok: lines ...
/usr/bin/x86_64-linux-gnu-ld.bfd: cannot find /tmp/testmgr-scratch-1289354/cods_imp_x64.o
test-emit-obj: data-import object FAILED to link

Run alone (--job 'test-emit-obj#03'), it is green.

Eliminated, each by measurement or search, not reasoning:

• Another Makefile rm. A glob-expanding scan of every rm ... $(TESTTMP)/... line finds exactly one that matches cods_imp_x64.o: #03's own first line (Makefile ~37801).
• A tool or test script deleting *.o or the scratch dir. No script under tools/ or test/ does. The test/*.sh scripts handed $(TESTTMP) delete nothing.
• The six-hour /tmp wiper. A fresh --emit-obj object has a current mtime.
• A fixed intermediate path in the compiler. There is no '/tmp/' literal and no rename in the output path. Identical sizes for #03's imp_x64 and #04's dup_a in the failing run are normal; those two objects differ but always size alike.
• cwd-running programs (c_ofa26, c_xi26) unlink only their own named files.

FOUND, same day. It was neither a retry nor a second run. testmgr's split_jobs put #03's leading rm -f $(TESTTMP)/cods_*.o at the END of job #02, because a job starts at a compile line and the rm is not one. #02 runs concurrently with #03, so the rm deleted #03's fresh objects whenever the two overlapped, which at tier load they do. The shared-filename merge did not join them because a glob is not a filename. A scan of every target found one more case (espx_, #04 against #06). Both rm lines are removed.