← board

Track T by default: the FAILING STEP named no owner. Line 1 of 1 is tools/optdiff.sh --shard 10/12. 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 6 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#shard10/12 at 285208414d3f in step 1/1, tools/optdiff.sh --shard 10/12 (auto-filed by twatch)

Repro

tools/testmgr.py --tier opt --job 'optdiff#shard10/12' at 285208414d3f332c0b13f2a084e671fbad114be7

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

OPT DIFF -O1: test/c_crtl_glob.c (rc 2 vs 2)
OPT DIFF -O2: test/c_crtl_glob.c (rc 2 vs 2)
OPT DIFF -O3: test/c_crtl_glob.c (rc 2 vs 2)
optdiff skip SKIPLIST: test_rtti_kind_numbering.pas
optdiff skip TIMEOUT-O0: c_unsigned_const_guard_folds.c test_qplus_narrowing_store.pas
optdiff skip BUILD-FAIL: c_ir_unsupported_reports_the_real_line.c c_obj_data_only.c c_obj_runtime_state_a.c c_pasunit_ansistring_result_fail.c c_unclosed_unsized_2d_init_fail.c c_undeclared_in_file_scope_init_refused.c cerror_directive_fail.c const_before_ctor_unit.pas cpasunit_strings.pas ctypes_lib.c except_b339_base.pas test_a_derailed_parse_names_the_appended_unit_as_the_compilers.pas test_a_half_dereferenced_call_result_is_refused.pas test_a_parameter_and_a_local_that_differ_only_in_case_are_two_symbols.pas test_an_interface_arity_mismatch_is_refused_fail.pas test_asm_rv32.pas test_bad_calls_all_report_fail.pas test_default_filefield_fail.pas test_default_textfile_fail.pas test_dynarray_in_a_variant_part_refused.pas test_generic_cycle_fail.pas test_generic_error_location_names_a_third_file_fail.pas test_generic_shadow_decl.pas test_header_static_body.pas test_overload_record_identity_fail.pas test_specialization_does_not_rename_after_a_dot.pas test_unit_finalization.pas test_unit_init_begin_form.pas test_unit_stray_token_refused.pas test_unterminated_comment_names_the_nested_brace.pas unit_c_bridge.pas
optdiff THREADSAFE-RETRY: test_atomic_counter.pas
optdiff shard 10/12: pass=186 skip=34 diff=1

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

Resolution (2026-09-16, frankb-56, Track C)

NOT A COMPILER DEFECT. The harness manufactured the difference, and the difference was argv[0]. optdiff built the four levels to four different paths and compared stdout+stderr. test/c_crtl_glob.c requires an argument, optdiff supplies none, so it exits 2 on its usage line — and that line prints argv[0]. The entire diff was the binary's own path:

-usage: /tmp/optdiff.N/d0 ...
+usage: /tmp/optdiff.N/d1 ...

rc 2 vs 2 was the tell and it was in the ticket from the start: matching exit codes on both sides mean the program ran to completion and disagreed only on text. Fixed in tools/optdiff.sh by building every level to ONE path.

THE COMPILER WAS NEVER IMPLICATED, AND I MEASURED THAT RATHER THAN INFERRING IT. Run properly with its argument at each level, c_crtl_glob.c is byte-identical across -O0/-O1/-O2/-O3 — 37 glob rows against glibc. This is the first time this program has actually been swept at four O levels, because under optdiff it only ever reached the usage path.

RESIDUAL, AND IT IS NOT CLOSED BY THIS FIX — owner: Track T. optdiff cannot supply arguments, so any argument-taking program in the corpus is swept only on its usage path and counts as a pass. After this fix these two pass honestly (all four levels agree) while covering nothing, which is a guard that cannot fail sitting inside the pass count. Sizing that population needs a full-corpus run, which this seat's gate does not include.