← board

Track T by default: the FAILING STEP named no owner. Line 1 of 1 is tools/optdiff.sh --shard 4/12. The job's own src (tools/optdiff.sh, 1 file(s)) is NOT used here on purpose: it is what the job compiles, not what broke, and guessing a lane from it is what sent three reds in one job to the wrong lane. This is a FALLBACK, not a finding — nothing says the defect is Track T's. Re-lane it before working it.

origin/master has advanced 8 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: optdiff#shard4/12 at d74c7fbe9ffe in step 1/1, tools/optdiff.sh --shard 4/12 (auto-filed by twatch)

Repro

tools/testmgr.py --tier opt --job 'optdiff#shard4/12' at d74c7fbe9ffeca26a3991fcf80ca1d1b4318564d

Range

bad unknown, range unknown — there is no earlier passing sha to bound it, or the bound is not recorded. No idle bisect will happen; this one needs hand-triage.

Log tail

Terminated
OPT DIFF -O1: test/test_threadsafe_io_lock_foreign.pas (rc 0 vs 0)
OPT DIFF -O2: test/test_threadsafe_io_lock_foreign.pas (rc 0 vs 0)
OPT DIFF -O3: test/test_threadsafe_io_lock_foreign.pas (rc 0 vs 0)
optdiff skip TIMEOUT-O0: lib_mimic_urllib_request_server.pas test_signal_handlers.pas
optdiff skip BUILD-FAIL: c_cross_ns_arity.c c_def_hijack.c c_pasunit_missing_fail.c cpthread_needs_threadsafe_b.c csqlite_file_probe.c csqlite_suite.c except_b339_derived.pas kwpadprobe.pas macronest_fpcmode.pas macro_soup_lib.c test_auto_var_fail.pas test_four_independent_errors_report_fail.pas test_generic_constraint_longint_fail.pas test_incdiag_unit_fail.pas test_metaclass_narrowing_error.pas test_method_missing_args_report_fail.pas test_mode_delphi_unit_leak.pas test_parallel_for_capture.pas test_pascal_directive_error.pas test_sealed_abstract_method_fail.pas uopfrac.pas
optdiff shard 4/12: pass=160 skip=23 diff=1

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

Not a bug — the sweep was asking the wrong question. 2026-09-01, frankZ.

test/test_threadsafe_io_lock_foreign.pas puts four foreign threads through 50 Writelns each. The order is nondeterministic by design and its own header says so — three runs of ONE binary at a fixed -O give three different byte streams. optdiff compares raw stdout, so it reports DIFF by construction, at every level, forever.

The property the program actually asserts is ATOMICITY: a Writeln is two write(2) calls and without the I/O lock two threads interleave them into a line that is not 300 identical characters. That is what its Makefile row asks (grep -c -E '^(A{300}|B{300}|C{300}|D{300})$' must be 200, plus done), and raw stdout comparison cannot express it.

Measured at HEAD (b9fd008f89ef), five runs at each of -O0/-O1/-O2/-O3: 200 done twenty times out of twenty. The invariant holds at every level; only the order moves.

So this goes to tools/optdiff.skip, and the entry says which question is being declined and which tool asks it instead. Note that this was NEVER the same bug as the five-shard hang — that one was [[bug-a-dce-miscompiles-every-threaded-program-and-o3-turns-it-on]], a timeout (rc 0 vs 124) rather than an output divergence, which is exactly the distinction frank-user drew when it said shard4 was a name collision. It was right.