mirror of
https://github.com/Druthulu/BFM-decomp
synced 2026-09-27 05:56:00 -04:00
Drew T
e44bc787eb
The largest target on the T14 census: 138 members x 289 ins = 39,882 templated instructions at stake. conform_decls --check showed 418 sites in 3 forms, pointer-type-only with NO narrowing warning and no return-type change — the func_80179B74 shape that banked 137/137. Applied (418 sites / 414 files), gated (jtbl carve succeeded first try), R22 clean-fleet 140 passed / 0 failed of 140. Committed BEFORE the family bank because jtbl_family_bank refuses to run on a dirty tree — its per-sibling revert restores from HEAD, so an uncommitted axis would be destroyed. The tool enforcing H4 in code, correctly. Also recorded: family_sweep REFUSED this family rather than returning 0/137 — func_80135EB0 is has_mid_jr, and §53's carve law says a carve-less sweep there yields "a 0% that is a TOOL artifact, not a wall". That refusal is the good version of today's pattern: every false wall untangled this session (func_8016B6BC 0/137, the 4 fabricated CC1-FAILs, Phase-28's B2 0/8) came from a tool that ANSWERED instead of refusing.