Go to file
Levi Neuwirth c9cafe9290 Agent K M2 prep: close review findings #4 (spelling domain) and #1 (migrate tests)
- valuegen::spelling(nth) is now injective over the full u8 (nominal = nth % 7,
  octave = nth / 7), so distinct selectors always give distinct PitchSpellings.
  Removes the silent mod-7 collapse footgun the M1 review flagged; no call-site
  changes needed (existing hex-looking selectors now genuinely differ).
- migrate.rs gains a unit-test module covering the reconstruction branches the
  corpus-driven equivalence gate never reaches: Tie/Beam create round-trip
  exactly; Spanner create is Irreversible (anchor-based, not event-ref —
  documented in operation_catalog §CreateCrossCutting); the tuplet-compensation
  variants migrate (ReplaceWithRest preserves rest id+duration, voice recovered
  at reduction per finding #3); respell recovers from context else Irreversible
  (P12-K1).

Gates green: build/fmt/clippy -D warnings; cargo test --workspace (ops migrate
tests + all criteria). No production logic changed beyond the spelling() token
domain.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-24 13:20:55 -04:00
.github/workflows Phase 2: spelling + decomposition pre-passes (Agent H) with F merge gate 2026-06-23 16:33:39 -04:00
crates Agent K M2 prep: close review findings #4 (spelling domain) and #1 (migrate tests) 2026-06-24 13:20:55 -04:00
spec Agent K M1 review follow-up: correct two doc-accuracy findings 2026-06-24 13:06:22 -04:00
.gitignore Phase 2: spelling + decomposition pre-passes (Agent H) with F merge gate 2026-06-23 16:33:39 -04:00
Cargo.lock Phase 2 (Agent I): visible-slice scaffold — Bravura SVG renderer vs stub 2026-06-23 20:01:40 -04:00
Cargo.toml Phase 2 (Agent I): visible-slice scaffold — Bravura SVG renderer vs stub 2026-06-23 20:01:40 -04:00