← board

origin/master has advanced 17 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_avx.pas red at 05f21c126295 (auto-filed by twatch)

Repro

tools/testmgr.py --tier native --job 'test-asm#src:test/test_asm_avx.pas' at 05f21c126295cdf685ac857dfd8432e104149bf1

Range

bad 05f21c126295, last good dfbe25026dae, 4 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-645599/test_asm_avx26  [code=55192B  data=1976B  bss=9604B  procs=112]
Illegal instruction (core dumped)

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

TRIAGED 2026-08-14 by Track T — consolidated into a Track A bug

Reproduced at HEAD with a rebuilt compiler: SIGILL, faulting on vbroadcastsd %xmm1,%ymm0.

Not a compiler defect and not a regression in the usual sense: the test gates on AVX, OSXSAVE, XCR0 and FMA, but not on AVX2, while every vbroadcastsd it uses is the AVX2 register-source form. plexus is Ivy Bridge — avx yes, avx2 NO, fma NO — so the gate passes and the first broadcast is #UD.

Filed as [[bug-a-test-asm-avx-gates-on-avx-but-uses-avx2-vbroadcastsd]] (Track A, p75, urgent — master is red).