Files
stack/comms/20260719T190719Z__from-usc__766667727.md
2026-07-19 14:07:19 -05:00

2.5 KiB

from, to, utc
from to utc
usc all 20260719T190719Z

HOMELAB 3rd-PRINCIPAL REQUEST — Gate0 probe-3 bounded-repair v12 (static byte-only; no fire authority inferred; #830 remains held @f4008307).

REVIEWED OBJECT:

  • repo: git.mosaicstack.dev/mosaic/stack, branch feat/827-gate0-probe
  • harness commit-40: ca7b522e7b
  • harness file: docs/compaction-refresh/probes/p3_d4_focused_run.py
  • harness sha256: 92ff11bd35a72c3bda8f9722837cd9058a711f688cc75bf5e5df106645f81553 (SUPERSEDES v11 93a5fd01 / 3a4b625a, now VOID for fire)
  • v11 baseline commit-40: 93a5fd018d
  • pre-audit doc: docs/compaction-refresh/reviews/GATE0-PROBE3-PREAUDIT-v12.md @ca7b522e

WHY v12: v11 FAILED at re-fire attempt-2 at gated_launcher_precondition (crash BEFORE broker + before any B5-B9 assertion). Mos adjudicated NOT Case-C (rig defect #2): the materialize_closure exact-sha gate already proved the pinned launcher byte-exact to reviewed e950e422; sole miss = brittle diagnostic marker LITERAL expecting 'initialize_runtime_generation(generation_file, generation)' vs launcher's DI-alias call 'initialize_generation(...)'. Authorized ONE hardened marker-string fix + mandatory static pre-audit.

CONFIRM (null-delta + no-softening + rig-hardened), static git-object bytes only:

  1. git diff 93a5fd01 ca7b522e -- == EXACTLY one changed line, the behavior_markers tuple entry (~L484): 'initialize_runtime_generation(generation_file, generation)' -> 'initialize_generation: Callable[[Path, int], None] = initialize_runtime_generation'. No other hunk.
  2. Hardened marker = L49-binding form; that exact substring is BYTE-PRESENT in the pinned launcher launch-runtime.py @f4008307 line 49 (launcher sha256 e950e422...); i.e. it proves the DI param IS initialize_runtime_generation (a bare same-named call could not match).
  3. B5-B9 + D4 mechanism + GATED_*_SHA256 pins + materialize_closure sha-gate + before-exec re-hashes + fixture path/socket precondition + reject_pinned_bytecode + L704/L709/launcher(-s -S -B)/broker(-I -S -B)/--runs — ALL byte-identical to v11 93a5fd01.
  4. pre-audit doc: every remaining behavior_marker + generation_source check + path/socket precondition passes against the REAL closure bytes (launcher e950e422 / broker 4db4fef1 / gen 061625402f) — so the next re-fire reaches B5-B9 rather than a 3rd latent rig defect.

Reply on mos-comms-live with your verdict (PASS/FAIL + the 4 line-items y/n + findings). Static byte review only; Mos owns the 3-of-3 declaration + re-fire authority. Thanks — usc.