Decompile Enemy/bosseel#128
Open
KakarottoCake wants to merge 34 commits into
Open
Conversation
Report for GMSJ01 (a4e0950 - 794f6a7)📈 Matched code: 35.09% (+0.25%, +9068 bytes) ✅ 95 new matches
...and 65 more new matches 📈 54 improvements in unmatched items
...and 24 more improvements in unmatched items |
…t refinements __sinit registers the JAL sound lists (MSBgm/MSSetSound(Grp)/JALSeMod*), which come from MSSetSound.hpp - it was missing from the includes so __sinit was 9.7%. Adding it before MSoundBGM.hpp (same as MapObjSample.cpp) reproduces the exact registration order: 9.7% -> 100%. Also refine the TBEelTears/TOilBall moveObject push handling and TNerveBEelTearsMarioRecover. Unit 91.9% -> 93.4%. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…init The target float-copies mPosition into mInitialPosition (lfs/stfs); operator= was compiling to an integer lwz/stw copy. 76.3% -> 78.1% on init. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The target constructs the Mario target vec with y=10 then overwrites y=15; that redundant write was missing (mine built it with y=15 directly). 93.1% -> 99.6%. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The target inlines mMActor->getModel() (mMActor->unk4) and caches it, but getModel() is out-of-line so it emitted a bl each time. Using mMActor->getModel() directly (cached in a local for the skin-deform block) matches the inline fetch. 76.3% -> 80.5%. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Same as init: getModel() is out-of-line; the target inlines mMActor->getModel(). 76.9% -> 78.8%. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The map marks ExecBackNerve_Sub as local, but it was declared in the header (extern) and emitted global. Drop the header declaration and make the definition static. Verified with tools/validate-symbol-order.py. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
getFrameCtrl(0)->getEnd() returns an int; the target doubles it as an integer (slwi) before the float divide, not getEnd() * 2.0f. 97.3% -> 99.9%. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Same as ExecBackNerve_Sub: getEnd() * 2 (integer slwi) not * 2.0f. 95.7% -> 98.3%. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
mPosition (0x70, on TSharedParts) is copied with a batched float load/store, not the integer operator=. Spelling it set(position.x, position.y, position.z) evaluates all three loads before the stores. 92.8% -> 100%. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Contributor
Author
|
Marking ready for review. The whole TU is implemented end-to-end (every function and symbol from the map present, reverse definition order preserved for the State: ~94%, 91/142 functions at 100% (22 more at ≥99%, 11 at 95–99%). Recent structural passes:
Known-remaining residuals, all verified as the human-tier category (instruction streams and logic match): stack-frame size, register allocation, and the 🤖 Posted by Claude Code |
The symbol-order validator flagged this UNUSED helper as the sole non-weak symbol out of order: it sat after the whole nerve block, but the map emits it between theNerve(FirstSpin) and execute(WaitAppear). Since -inline deferred emits in reverse of source order, moving its definition to just after the WaitAppear nerve (before FirstSpin) reproduces the map order. validate-symbol-order.py now passes (non-weak order + linkage + presence). No fuzzy change (ordering doesn't affect per-function matching). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Contributor
Author
|
Ran the new symbol-order validator (the CI check from the
No fuzzy change from either fix — ordering/linkage don't affect per-function matching; the TU stays at ~94% (91/142 at 100%). 🤖 Posted by Claude Code |
Fixes the lint CI check (variable-declaration alignment and multi-line continuation indentation). Whitespace only; no codegen change. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Enemy/bosseel.cppend-to-end, including the tears/oil-ball subsystem, manager parameters, teeth, eyes, vortex, heart coins, collision actors, Boss Eel initialization, the mainperformloop, and the complete battle nerve graph-inline deferredbuildmario.MAPUNUSED entries; reconstruct the recoverable inline helpers and mark dead-stripped bodies with focused human-review TODOs and retail sizesMatching
mario/Enemy/bosseel: 79.24% fuzzy matchTBossEel::perform: 76.81%Several nerves and support routines are above 90%, with many weak functions and small routines fully matching.
Validation
ninja changes_allpassesgit diff --checkpassesThe existing warnings in
JGadget/std-list.hppandMoveBG/MapObjGeneral.hppare unchanged.Human review
The remaining TODOs are primarily dead-stripped UNUSED routines whose original bodies do not survive in linked assembly. Each is explicitly labeled with its retail MAP size so it can be reviewed or completed separately.