regression CASCADE: 17 jobs newly red at 110774a14648 (auto-filed by twatch)

• Type: regression cascade (auto-filed by Track T watcher, host xeon). Untriaged. 17 jobs went red in ONE sweep — treat as ONE root cause until triage proves otherwise; do NOT fan out per-job tickets.
• Found: 2026-07-31T16:56:14Z
• Root-cause suspects in the red set: fpc-bootstrap#src:compiler/compiler.pas, selfhost-fixedpoint#src:tools/selfhost_fixedpoint.sh

Repro (start with a suspect, or any listed job)

tools/testmgr.py --tier native --job '<job>' at 110774a1464878920dcdf27d2afae7ca36dae219

Newly red jobs

• fpc-bootstrap#src:compiler/compiler.pas
• selfhost-fixedpoint#src:tools/selfhost_fixedpoint.sh
• test-asm#src:test/test_asm_so.asm
• test-core#src:examples/tk/facade_and_paths.npy
• test-core#src:examples/tk/import_in_body.npy
• test-core#src:examples/tk/shadow_format_except.npy
• test-core#src:test/cprintf_ll_b252.c@2
• test-core#src:test/test_c_define_const.pas
• test-core#src:test/test_c_gtk.pas
• test-core#src:test/test_c_gtk_call.pas
• test-core#src:test/test_c_gtk_types.pas
• test-core#src:test/test_c_gtk_window.pas
• test-core#src:test/test_nilpy_c_define_const.npy
• test-core#src:test/test_sqlite_crud.pas
• test-core#src:test/test_sqlite_crud_autotyped.pas
• test-core#src:test/test_sqlite_crud_lazy.pas
• test-core#src:test/test_string_to_pchar_auto.pas

Cascade stub: one signal for one event. Track T agent (face 2) or the owning dev track triages the root; individual tickets only for whatever remains red after the root is fixed.

REJECTED — false positive, not a regression (Track T, 2026-07-31)

All 17 jobs were red because of the host, not the commit. 110774a14648 is a tstate-only commit (tstate(borg): opt f2f1a3a9add8 done) and touches no code at all, so it cannot have caused a single one of these.

This was the first run of a newly enrolled watcher box (xeon), and the reds decompose into exactly two causes, both measured at the sha, both now closed:

1. The suite tested the wrong binary (accounts for the two "root-cause suspects"). A fresh watcher clone is seeded with make seed-from-stable, which copies stable_linux_amd64/default/pinned onto compiler/pascal26. The copy lands with a newer mtime than compiler/compiler.pas, so testmgr's make compiler/pascal26 reported "up to date" and never self-hosted. The whole 1098-job sweep ran against the pinned binary rather than a compiler built from the checked-out sources. compiler/pascal26 was byte-identical to pinned and 13 minutes newer than the sources — measured, not inferred.

selfhost-fixedpoint is the ONLY job that can see this, and it did its job correctly: property 2 (the anti-Thompson agreement check) fired exactly as designed. After touch compiler/compiler.pas && make compiler/pascal26, the gate converges in 1 round and agrees. Not a bug — the gate working.

fpc-bootstrap is the separately ticketed, advisory FPC seed drift (urgent/regression-fpc-seed-drift-b1976-stale.md). It is red on borg too; it is red here for the same reason and is NOT part of this cascade.

2. Missing host dev packages (the other 15). Each verified fixed by installing the package, at the same sha, with no code change:

test-asm#src:test/test_asm_so.asm passes standalone at the same sha and needed no package — a genuine flake, folded in by the cascade grouping.

Kept as a record rather than deleted: this is the concrete reproduction of the known "phantom NEW-RED" complaint, and cause 1 is a real trap — the documented fresh-box step silently makes every job test a stale compiler, and only one job in the matrix can notice. Follow-ups filed separately: task-t-seed-from-stable-defeats-rebuild and task-t-suppress-autoticket-until-host-baselined.