@T428226448378@PixelCNinja Please report on the GitHub. The moves that break usually have a common issue, Naomi board / SHC compiled code behaved differently when dealing with read through nulls etc. So the code is correct, but a bug in Naomi presents different on PC due to a) memory reads, b) compiler.
@Codeserial0@PixelCNinja Naomi decomped, then DC after wards, they shared about 95% of their code. Code can generate either, and the port can use both and switch between them!
@oldmanuk@Sabre_AZ@sexymame2 So thanks for the hint, I was abstaining from heavy macro work (I used it for the sequencer which was great) etc, but your tip has helped clear up some of the cludge! I'll give you some credit in the repo as I do a pass. Thanks again
@oldmanuk@Sabre_AZ@sexymame2 Also we had the debug symbols from the real Capcom SF3: 3rd Strike, and the real names are as bad as I have them for the most part (unless I have provisional names). I use macros where the code got unwieldy.
@oldmanuk@Sabre_AZ@sexymame2 A few prompts would not have fixed it, this was not done trivially. Instruction matches, sure trivial, not exact register placement, even with compiler source code guiding me, I didn't want to divine a rules that would fix something without making sense. It's ongoing though.
@oldmanuk@Sabre_AZ@sexymame2 Also I get the impression you don't think I'm a coder! I was there during that era, coding QuakeC as my first ever projects :) Anyway yes there is some tricks in the code while we improve the compiler, but a lot of the stuff in there is just horrible table based state machines!