RedSound/RedCommand: 96.56% -> 99.46%

[SirEnkido]

What changed

src/RedSound/RedCommand.cpp (+ one enum in include/ffcc/RedSound/RedEntry.h), across _SePlayStart, _MusicPlayStart, and helpers. Seven commits, each an independent verified step:

• _SePlayStart: loop restructure with chained zero-assignments; if/else outlining with a block-scope trackNo; explicit U16 high-shift (<< 8) combine form via a new named enum
• _MusicPlayStart: chained zero-assignments mirroring _SePlayStart; ADSR memset to DEFAULT_WORD(-1); in-place sequential volume scaling; a shared function-scope trackNo reused for the reverb index and track counter (register-coloring lever consistent with the original compiler output)

Evidence

• main/RedSound/RedCommand fuzzy (report.json): 96.56% → 99.46% (+2.90)
• DOL verified byte-identical (build/GCCP01/ok passes; sha1 matches config/GCCP01/build.sha1)
• Independently audited: all 20 /* --INFO-- */ headers preserved, no offset tricks, no fake symbols, no extern crutches, only the two intended files touched

Notes for review

• The remaining sub-100% functions (_SePlayStart 97.3%, SeBlockPlay 98.6%, _MusicPlayStart 99.7%, _EraseTime/MusicStop ~99.9%) were each disassembled and confirmed to be register-allocation / scheduling artifacts (e.g. a phantom stmw r24 save over a byte-identical body); left as-is rather than chased with hacks.
• RedMusicVolumeFromInput was inlined into sequential mutate statements purely for scheduling match — flagged for transparency, not a rule violation.

---

Produced by Claude agents following AGENTS.md / WORK_SPLIT.md, operated by @SirEnkido. Every intermediate score in the commit messages was verified against report.json at commit time.

🤖 Generated with Claude Code