TU: fix stale legacy_source extensions in the manifest

[andrewboudreau]

What

38 functions[].legacy_source values across 19 TU manifest entries still point at src/<name>.c when the file on disk is src/<name>.cpp — the file was migrated C -> C++ and the manifest row was never updated.

Each stale row blocks tu_promote on that entry: the promote path resolves legacy_source to a file that no longer exists.

Derivation

Re-derived against current origin/main (25cfc136) by script, not carried over from an older tree. For every config/tu_manifest.d/**/*.json, every legacy_source ending in .c was checked against git ls-tree -r --name-only origin/main. A row counts as stale iff the .c path is absent and the .cpp sibling is present.

• Stale rows found: 38, across 19 files.
• Inverse case (a .cpp legacy_source whose file is actually .c): 0 — nothing to fix there.
• No data[].legacy_source rows were affected; all 38 are under functions[].

Re-running the same derivation after the change reports 0 stale rows.

Entries changed

entry	rows
arm9/Actor.json	7
arm9/ActorDerived.json	1
ov002/Platform.json	2
ov002/Warp.json	1
ov009/CastleWater.json	2
ov009/Flag.json	2
ov009/MetalNet.json	1
ov010/LightBeam.json	2
ov010/PeachPainting.json	2
ov014/ShutterBob.json	1
ov019/IceSlideManager.json	1
ov019/RacingPenguin.json	2
ov020/HauntedChair.json	2
ov029/CageLift.json	1
ov029/RotatingPlatformWdw.json	2
ov029/SwitchActivatedPlank.json	2
ov029/WaterDiamond.json	2
ov062/KoopaFlag.json	2
ov077/Spiny.json	3

Shape

Manifest JSON only — no tooling, no source, no config outside config/tu_manifest.d/. Applied as a targeted textual substitution of the specific "legacy_source": "src/X.c" strings rather than a json.load/json.dump round-trip, so indentation, key order and trailing newlines are byte-identical outside the 38 changed values.

git diff --stat is 19 files, 38 insertions, 38 deletions; every changed line pair is verified to be an exact .c -> .cpp swap on the same base name and nothing else.

python tools/tu_manifest.py list still succeeds and still reports 89 TUs, unchanged from main.

Provenance

Salvaged from the closed PR #1884 — this is the one clean, mechanical win in it. It was re-derived from current main rather than cherry-picked, because #1884's row set was computed against a 411-commit-old tree. This supersedes the salvageable part of #1884.

🤖 Generated with Claude Code

https://claude.ai/code/session_01WdCK1xgrJdiJzPCh3bAJfQ

[tangos-validator]

✅ PR validation — Passed

Committed merge introduces no reconstruction or attribution regression.

Full merge validation

Check	Result
Committed test merge	yes
Byte-verified functions	10,998 / 11,347 (96.92%, +0)
Byte-verified code bytes	2,045,248 / 2,211,124 (92.50%, +0)
Claimed, not byte-verified	218 functions, 60,056 bytes (+0)
Perfect source moves	0 R100
Enrolled ranges (delinks complete)	11,058 functions, 2,052,772 bytes (92.84%, +0) -- differs from byte-verified by +60
Contributor credit	0 added, 0 changed, 0 lost
Relocation check	0 checked; no affected slots
Module fidelity	106/106 exact; 100.000000% compared bytes
Code linked from verified source	11,087 functions, 2,066,772 bytes (93.47%)
Module bytes from source	2,066,772 / 3,049,600 (67.8%); 811,492 (26.6%) are data no delink entry reaches
ROM data reproduced from source	456 symbol(s) exact, 241 partial, 9 differ

Byte-verified means the range carries complete in a delinks.txt, so the ROM build compiled it and compared it to the cartridge. The 218 claimed functions have a src/ file named after the symbol with no NONMATCHING banner, and nothing compiles them -- dsd fills their addresses with the ROM's own bytes. Both together are the 11,216 this project calls matched.

_{The private worker commits a test merge, builds the stock ROM profile, compares every executable module, measures matched and source-built code, checks contributor lineage, and verifies affected relocations. The mod profile is opt-in and is not part of this merge gate.}

[andrewboudreau]

Reviewed and queued to merge — the derivation is the right one (re-derived against origin/main via git ls-tree, not cherry-picked from #1884's 13bf9222f, whose row set was computed against a 411-commit-old tree). Regex-targeted substitution rather than a json.dump round-trip is exactly right here; it keeps the diff at 38/38 with byte-identical formatting elsewhere.

This supersedes the one salvageable hunk of #1884, which I closed along with #1880/#1882 (see #1880 for why: its plan_many change dropped every _ZTV/_ZTI/_ZTS section uncompared, a design main deliberately rejected in favour of #1987's deadstrip-data licensing).

Separate follow-up, not for this PR: the same scan found 75 other legacy_source values pointing at paths that exist in neither extension — all of ov070/daKpFr_c.json and ov070/daKrpa_c.json, plus ov002/daObjAbuku_c.json, arm9/ActorBase_SceneNode.json, and scattered func_* rows in ov010/PeachPainting, ov020/HauntedChair, ov029/WaterDiamond. Those are entries whose one-function files were consolidated into merged TUs, so the legacy path is gone rather than renamed. That is a different and non-mechanical problem — correctly left alone here, but it wants its own look, since a dangling legacy_source is exactly the kind of pointer that reads as evidence without being any.