Files
BFM-decomp/docs
Drew T 7ccf48f2f4 feat(phase-26): Task-8 reconcile wiring (§41c h_seq) + 463 reconcile-class banks
- BUILT the per-sibling reconcile: family_remap.remap_hseq_body (h_seq-remap a RAW crack draft: symbol +
  immediate + cross-address self-rename) + family_sweep.reconcile_remap_hseq + --reconcile-raw. Per sibling,
  remap the RAW crack then canon_sig_reconcile against that sibling's own TU (the h_seq port of the h_norm
  M2 path) — because a reconciled body is TU-specific and can't template plainly (validation: 0/4).
- HARVEST: the 4 triage isolation-cracks (func_80155800/80167540/801506A4/8016A73C) templated 463/0 x~133
  via --reconcile-raw (0 failures). Metrics: instr 58.5->58.9%, distinct 30.9->31.1%.
- each overlay SHA-gated by harvest_verify vs config/check.<ov>.sha (byte-identical = the match def, G3).
  FULL R22 clean-fleet DEFERRED until the concurrent Fable5 crack agents release asm/ (their m2c needs it);
  R22 fleet-confirm to follow post-window.
- cookbook §40c (the h_seq per-sibling reconcile technique, R30).
2026-07-12 00:03:47 -06:00
..