Chord round-trip breaks for a seventh with add11/add13: "C7add13" prints as "C713", which reparses as C13 with a 9th and 11th

[matt-edmondson]

What's wrong

Since #293 (ad24784), Chord.Parse accepts add11 and add13 after a seventh. For example, "C7add13" parses as a dominant seventh with tensions {13} only, and a check gives tones [0,4,7,10,21].

The formatter can't write that chord back. AppendTensions (Semantics.Music/Chord.cs:598-622) writes addN only when there is no seventh. With a seventh it writes the highest natural tension as a bare stacked number ("13", "11", "9"). On reparse, ApplyExtensions expands 13 into 9 + 11 + 13.

Observed (scratch app referencing Semantics.Music)

Input	ToString()	Reparsed tones	Equal
C7add13	C713	[0,4,7,10,14,17,21] vs original [0,4,7,10,21]	False
C7add11	C711	gains a 9th	False
C9add13	C713	gains an 11th	False
Cmaj7add13	Cmaj713	gains a 9th and 11th	False

The round-trip corpus (ChordRoundTripTests.cs:12-18) has no seventh+addN case, so the tests don't catch this.

Suggested fix

When there is a seventh:

1. Write the highest contiguous natural stack as a number: 9 → 9, 9+11 → 11, 9+11+13 → 13.
2. Write each remaining natural tension that isn't implied by that number as addN. For example, {13} gives 7add13 and {9,13} gives 9add13.

The alternative is to reject addN after a seventh in the parser again, but that would undo #293.

Acceptance criteria

C7add13, C7add11, C9add13 and Cmaj7add13 are added to the round-trip corpus, and each one round-trips to equal tones.

[matt-edmondson]

Triage

• Category: Bug
• Priority: Medium. It breaks the parse/format round-trip. ToString() of a valid chord reparses as a different chord with extra tones, so any caller that stores chords as text silently corrupts them. Only seventh chords with add11 or add13 are affected.
• Area / suggested assignment: Semantics.Music/Chord.cs (AppendTensions, lines 598-622) and ChordRoundTripTests.cs
• Duplicates: none found among open issues in the org. It is a regression surface from Read C69, Cadd11 and Cadd13 as added tones, and reject leftover text [patch] #293.
• In progress: no matching open PR

Notes: The suggested fix is right: emit the highest contiguous natural stack as a number, then addN for the rest. Keeping #293 is better than rejecting addN again. Add the four failing inputs to the round-trip corpus first, so the fix is test-driven.

---

Generated by Claude Code

[matt-edmondson]

Covered by #360, which redesigns this bug cluster as a whole. Implement it through that issue rather than individually.

---

Generated by Claude Code