• it_802ADA1C
• it_802ADAF0
• it_802ADB48
• it_802ADB80
• it_802ADBB0
• it_802ADBE4

https://github.com/doldecomp/melee/blob/8fb93d47102263fd4bef40841ec62e638ca50843/src/sysdolphin/baselib/pobj.h#L22C33-L22C33

I wrote this comment probably 6 years ago, and I don't remember what I was thinking here but the vtxcount is clear a 16 bit value due to :

As discussed on Discord, the ObjUpdate functions of various HSD types were changed to have enum_t for the type argument, when it is in fact an unsigned int that uses a list of #define rather than an enum. An example of this is MObjUpdateFunc.

This branch addresses some of that in commits as well as incorrect prototype declarations that were part of CObj, if someone wants to cherry-pick them. master...PsiLupan:melee:cobj-match

7badb2f

Normally I'm not picky about how floats are written, and I usually prefer to leave off as much as I can like f32 x = 0;. But since we're trying to match exactly, I think it would be useful to standardize all of our literals to be suffixed with either F or L. I know 0.0 is explicitly a double in C, but having the extra L should make it easier to scan the code.

Some existing literals use lowercase f so I think we should agree on one or the other.

So I don't forget.

dolphin

gx

GXFifo

• melee/src/dolphin/gx/GXFifo.c

  Line 133 in 346f267

  // todo: This is weird; try to match without it.

GXLight

• melee/src/dolphin/gx/GXLight.c

  Line 3 in 346f267

  #include <functions.h> // todo

sysdolphin

baselib

mtx

• melee/src/sysdolphin/baselib/mtx.c

  Line 70 in 346f267

  f32 new_var; // TODO: try to get rid of this

shadow

• melee/src/sysdolphin/baselib/shadow.c

  Line 38 in 346f267

  f32 hack; // TODO: remove once the whole file is converted and we can use float constants

melee

ft

fighter

• melee/src/melee/ft/fighter.c

  Line 47 in 346f267

  // TODO: verify that this is really a spawn number counter, then rename this var globally

• melee/src/melee/ft/fighter.c

  Line 89 in 346f267

  // inline auto preferring 3 levels of depth. TODO: Determine if the Setup MTX

• melee/src/melee/ft/fighter.c

  Line 123 in 346f267

  // @TODO: &lbl_80458FD0+2, +3, +4 are not defined in the fighter.s data section, how does this work?

• melee/src/melee/ft/fighter.c

  Line 140 in 346f267

  // TODO: size parameter too large to be only the size. maybe this contains flags. In other function calls, size only looks plausible. Or is this virtual memory?

• melee/src/melee/ft/fighter.c

  Line 824 in 346f267

  Fighter* fp = (void*)getFighter(fighterObj); // bit of a fake void* cast, but a sacrifice well worth getting rid of an fp_temp and filler. TODO: Maybe we can still do better?

• melee/src/melee/ft/fighter.c

  Line 1198 in 346f267

  // TODO: There has to be some way to get rid of this while maintaining the

• melee/src/melee/ft/fighter.c

  Line 2222 in 346f267

  // TODO: these double_lower_32bit variables are probably integer counters that get decremented each frame,

• melee/src/melee/ft/fighter.c

  Line 2269 in 346f267

  // TODO: we set this velocity to 0 after applying it -> Is this SDI or ASDI?

• melee/src/melee/ft/fighter.c

  Line 2313 in 346f267

  // TODO: do the bitflag tests here tell us if the player is dead?

ftbosslib

• melee/src/melee/ft/ftbosslib.c

  Line 11 in 346f267

  }* lbl_804D782C; //TODO: unify this struct

• melee/src/melee/ft/ftbosslib.c

  Line 36 in 346f267

  { // TODO: assert macro

ftdata

• melee/asm/melee/ft/ftdata.s

  Line 2176 in 346f267

  # TODO: What is this table?

chara/ftIceClimber/fticeclimber2_nana

• melee/src/melee/ft/chara/ftIceClimber/fticeclimber2_nana.c

  Line 3 in 346f267

  /// TODO remove float externs when file is finished and float data is removed from asm

chara/ftKirby/ftkirby

• melee/src/melee/ft/chara/ftKirby/ftkirby.c

  Line 3 in 346f267

  //// TODO remove externs

• melee/src/melee/ft/chara/ftKirby/ftkirby.c

  Line 7 in 346f267

  //// TODO find real struct / array

chara/ftMasterHand/ftMasterHand_2

• melee/src/melee/ft/chara/ftMasterHand/ftMasterHand_2.c

  Line 104 in 346f267

  Fighter* fp_1; // todo: dup?

chara/ftMasterHand/ftMasterHand_8

• melee/src/melee/ft/chara/ftMasterHand/ftMasterHand_8.c

  Line 54 in 346f267

  temp_r31->xB0_pos.y = temp_r30->x50.x; // TODO: The code matches, but is this right?

chara/ftSamus/ftsamus3

• melee/src/melee/ft/chara/ftSamus/ftsamus3.c

  Line 142 in 346f267

  extern s32 lbl_803CE6B8[]; // TODO

gr

ground

• melee/src/melee/gr/ground.c

  Line 428 in 346f267

  // TODO: what stages are these?

• melee/src/melee/gr/ground.c

  Line 1499 in 346f267

  // TODO: attempt decomp once param types are known

mp

mpcoll

• melee/src/melee/mp/mpcoll.c

  Line 8 in 346f267

  // TODO: proper signatures, mplib.h

• melee/src/melee/mp/mpcoll.c

  Line 103 in 346f267

  // TODO: float order hack

• melee/src/melee/mp/mpcoll.c

  Line 634 in 346f267

  s32 dummy = 0; // TODO: fake, breaks func_80043558

• melee/src/melee/mp/mpcoll.c

  Line 686 in 346f267

  #ifdef NON_MATCHING // TODO: dummy stack in func_80043324_inline2 breaks this function

• melee/src/melee/mp/mpcoll.c

  Line 745 in 346f267

  // TODO: float order hack

pl

player

• melee/src/melee/pl/player.c

  Line 25 in 346f267

  /// TODO probably the same struct as above, figure out how to make them work as one

• melee/src/melee/pl/player.c

  Line 43 in 346f267

  /// TODO delete after fixing functions that use this

• melee/src/melee/pl/player.c

  Line 134 in 346f267

  inline hasExtraFighterID(ftMapping* data) { return data->extra_internal_id != -1; } ///TODO this can be used in more places when functions are fixed to use correct structs

• melee/src/melee/pl/player.c

  Line 368 in 346f267

  /// new match decomp.me/scratch/oHfiV TODO

Author: @VetriTheRetri

• ftcoll
  • #708
    • func_80078A2C
  • #917
    • func_80078C70
  • #916
    • func_8007B1B8
    • ftColl_CreateReflectHit
    • ftColl_CreateAbsorbHit
• lbcollision
  • #724
    • func_8000ACFC
• ftaction
  • #868
    • func_80073240

• func_802207F0
• lbl_802207F4
• lbl_8022086C
• lbl_80220870
• lbl_80220894
• func_8022089C
• lbl_80220984
• lbl_802209B0
• lbl_802209B8
• lbl_802209BC
• lbl_802209C0
• lbl_80220A10
• lbl_80220A18
• lbl_80220A4C
• lbl_80220A50
• lbl_80220AA0
• lbl_80220AA8
• lbl_80220AC8
• lbl_80220ACC
• lbl_80220B78

• func_80221C10
• lbl_80221C14
• lbl_80221C84
• lbl_80221C88
• lbl_80221CAC
• func_80221CB4
• lbl_80221D9C
• lbl_80221DC8
• lbl_80221DD0
• lbl_80221DD4
• lbl_80221DD8
• lbl_80221E28
• lbl_80221E30
• lbl_80221E64
• lbl_80221E68
• lbl_80221EB8
• lbl_80221EC0
• lbl_80221EE0
• lbl_80221EE4
• lbl_80221EEC

• #1013
  • https://decomp.me/scratch/aaw2L
  • https://decomp.me/scratch/VBgbX
  • https://decomp.me/scratch/qplEf
  • https://decomp.me/scratch/iiE2X
  • https://decomp.me/scratch/1kgI1
  • https://decomp.me/scratch/kGOOu
  • https://decomp.me/scratch/NfBMZ
  • https://decomp.me/scratch/Oh4S7
  • https://decomp.me/scratch/nxe3p
  • https://decomp.me/scratch/f9rqG
  • https://decomp.me/scratch/vUpCo
  • https://decomp.me/scratch/uhINa
• #1014
  • https://decomp.me/scratch/ngexB
  • https://decomp.me/scratch/pwnH5
  • https://decomp.me/scratch/kLgW7
  • https://decomp.me/scratch/Pkj91 (This has not been completed and is awating on the peach completion)
  • https://decomp.me/scratch/vRngr
  • https://decomp.me/scratch/HbHIu
  • https://decomp.me/scratch/68eWF
  • https://decomp.me/scratch/a6CUZ
  • https://decomp.me/scratch/OzbvZ

930d31b:

Progress:
    Code sections: 731049 / 3882272 bytes in src (18.830443%)
    Data sections: 182474 / 1223369 bytes in src (14.915696%)
You have 55 of 293 Trophies and completed 7 of 51 Event Matches.
Code bytes to go for next trophy: 10956

b0d2f6a (directly after):

Progress:
    Code sections: 724977 / 3882272 bytes in src (18.674039%)
    Data sections: 182393 / 1223369 bytes in src (14.909075%)
You have 54 of 293 Trophies and completed 7 of 51 Event Matches.
Code bytes to go for next trophy: 3778

This is the source permuter we usually use for helping us finish decompilations. The idea is that we have C code that is logically equivalent to the solution -- however, there are optimizations that the compiler will do or heuristics to change the output assembly in a way that is hard to guess every time. The permuter just does random operations and hopes the score will go down. I think that this is actually a good use case for ML if we're able to express the problem correctly. Working on expressing the problem -- once we do that, we can train on the corpus of this repo and self-generated variations via the permuter :).

WIP Branch: https://github.com/JohnnyUrosevic/melee/tree/ftyoshi

• lbl_8012C7A4
• lbl_8012C80C
• func_8012C850
• func_8012CC1C
• func_8012CC6C
• func_8012CC94
• func_8012CCC4
• func_8012CCD0
• func_8012CCE0
• func_8012CCF0
• func_8012CD00

7ee0968

https://discord.com/channels/727908905392275526/727909624342380615/1114668619377414225

https://decomp.me/scratch/uCVeJ

Author: @Ka-lua
Co-authors: @ribbanya, @r-burns, @ProjectRevoTPP

• #471
• #473

https://decomp.me/scratch/QRCuA

• #1060
  • it_802F2BFC
  • it_802F2CCC
  • it_802F2CD4
  • it_802F2CD8
  • it_802F2CE0
• #1061
  • it_802F2D98
  • it_802F2E68
  • it_802F2E70
  • it_802F2E74
  • it_802F2E7C

You do not need to make a formal PR to get your matching decompilation merged, and you don't need to finish a whole file. Simply create a new issue with any number of scratches and a maintainer will integrate your code into master.

See our Getting Started guide for more information.

Author: @EstexNT

• func_8015DB2C

Such as pi, degrees-to-radians, etc. GAS supports predefined floats via .define so those should also be replaced.

Ideally also remove "private" module functions from the "public" headers and use __module_name.h for those instead.

In particular at the time of writing:

• ftpikachu.h
• ftzelda.h

• #768
  func_8007C114: https://decomp.me/scratch/QP6Eg
  Checks if the fighter has an item, if it does and it's the hammer it checks if the head should fall off (by calling func_800C555C). Then it sets a bit flag.

• #1026

  • func_800C555C: https://decomp.me/scratch/XPrfZ
    Checks if the hammer the HSD_GObj is holding should have its head fall off by calling func_8026C16C.

https://decomp.me/scratch/5YZak

In cm/cmsnap.s

void func_80031640(HSD_GObj* arg0, s32 arg1) {
    switch (lbl_80453060[0]) {
        case 1:
            func_800122C8(&lbl_80453060[1], 0, 0, 0);
            lbl_80453060[0] = 2;
            return;
        case 2:
        case 3:
        case 4:
            lbl_80453060[0] += 1;
        case 0:
        default:
            return;
    }
}

This looks like union_type_dobj or some type of inline, given it shows up multiple times.

Originally posted by @PsiLupan in #528 (comment)

• it_802C01AC
• it_802C027C
• it_802C02E4
• it_802C031C
• it_802C0368 <- rtburns & Revo, 99.78%
• it_802C0490
• it_802C0498
• it_802C04B8
• it_802C04C0
• it_802C04C8
• it_802C04D0
• it_802C04F0

os

• #481

db

• db.c (#320)

mtx

• #505

dvd

• #506
• fstload.c (#320)

dsp

• dsp.c (#209, #507)
• dsp_debug.c (#320)
• dsp_task.c

hio

• hio.c

gx

• #462

PowerPC EABI Support

• __mem.c
• __va_arg.c
• global_destructor_chain.c
• ExceptionPPC.cp
• __init_cpp_exceptions.cpp
• abort_exit.c
• buffer_io.c
• critical_regions.ppc_eabi.c
• ctype.c
• mbstring.c
• mem.c
• misc_io.c
• rand.c
• wchar_io.c
• s_fabs.c
• scalbn.c
• e_rem_pio2.c
• fminmaxdim.c
• k_cos.c
• k_rem_pio2.c
• k_sin.c
• s_copysign.c
• s_cos.c
• s_floor.c
• s_frexp.c
• s_ldexp.c
• s_sin.c

MetroTRK

• mem_TRK.c

Author: @ProjectRevoTPP

• https://decomp.me/scratch/PxWX7

I'm thinking something like

// calcprogress force
asm static void __init_registers(void)
{ // clang-format off
	nofralloc

	lis r1,  _stack_addr@h
	ori r1, r1,  _stack_addr@l
	lis r2, _SDA2_BASE_@h
	ori r2, r2, _SDA2_BASE_@l
	lis r13, _SDA_BASE_@h
	ori r13, r13, _SDA_BASE_@l
	blr
} // clang-format on
// calcprogress restore

Discussed a bit over Discord. Creating this as a placeholder for further discussion.

Hi, this is a really great start. I have a couple questions, though.

1. Is this still being worked on?
2. Would you accept contributions which just add comments/pseudocode to assembly, or which rename symbols. to make the code more readable?
3. Is there any consideration for porting parts of the code to C, like UnclePunch has been doing for his own modifications?

@PsiLupan

Left these comments on functions I don't know the names of, which exist in KAR but are stripped in melee because their only uses are auto-inlined. Maybe Psi knows what to call them.

Originally posted by @r-burns in #475 (comment)

Every function of math has been fully matched and decompiled except for func_803265A8. (which I'll do very soon, unless one of you wants to pick it up for themselves)

The currently linked math.o file is from the asm so this has to be changed as well as soon as func_803265A8 is matched.

• Match func_803265A8
• Rename variables to what they do
• Reformat function bodies/Remove comments
• Rename functions to what they do
• Link math.o from "src"

Originally posted by @FluentCoding in #235 (comment)

Author: @pdeblois

• func_8013CD34
• lbl_8013C94C
• ftPurin_SpecialHi_StartAction
• ftPurin_SpecialAirHi_StartAction

And also give that enum a name.

The reason it's mixed usage is because of me copy-pasting from structs in FRAY, which I think was done here. It's been something I've tried to cleanup as I actually decompile files themselves.

I was using this style because of Intellisense at the time and changed to match how other projects handle it with /* 0xA */, but we should settle on which needs to be done if that's the case.

Originally posted by @PsiLupan in #556 (comment)

Authors: @jeffmao99, @VetriTheRetri
https://decomp.me/scratch/fpAap

• #707
  • func_8011F16C
  • func_8011F190
  • ftIceClimber_SpecialN_StartAction
  • ftIceClimber_SpecialAirN_StartAction
  • lbl_8011F318
  • lbl_8011F354
  • lbl_8011F390
  • lbl_8011F394
  • lbl_8011F398
  • lbl_8011F3B8
• #715
  • lbl_8011F3D8
  • func_8011F6FC

• #704
  • func_80220F10
  • lbl_80220F14
  • lbl_80220F84
  • lbl_80220F88
  • lbl_80220FAC
  • func_80220FB4
  • lbl_8022109C
  • lbl_802210C8
  • lbl_802210D0
  • lbl_802210D4
  • lbl_802210D8
  • lbl_8022114C
  • lbl_80221154
  • lbl_80221188
  • lbl_8022118C
  • lbl_802211DC
  • lbl_802211E4
  • lbl_80221204
• #734
  • lbl_80221354
  • lbl_8022135C
  • func_80221288

https://github.com/doldecomp/melee/pull/587/files/cc5a9c7d8bbf9bcf6af73446c0d92e5b4ab9d6bb..ac21c20a18be1a57be5646934fcb6182fc038522

Author: @pdeblois

• #709
  • func_8013CD34
  • lbl_8013C94C
  • ftPurin_SpecialHi_StartAction
  • ftPurin_SpecialAirHi_StartAction