Topic: dink's FBN Development & Fixes thread

[barbudreadmon]

While testing games on fba standalone to see if they had the same input issues as the libretro port, i came upon those input issues on standalone :

aceattac : non working parent ? clone seems fine
fsoccerb : can't move or shout
fsoccerba : can't pass select screen
sbm : can't pass select screen

[Gab75]

- aceattac: I confirm the issue in SEGA Ace Attacker (aceattac) with the input controls, the clone (aceattaca) seems to work fine.
- fsoccerb: IIRC the input controls bug in SNK Fighting Soccer (all versions) is a known issue.
- sbm: you can press the key "2" (P2 Start) to pass the select screen; in any case, Sonic Blast Man is a "punching-ball arcade game", then it's very difficult to emulate with a keyboard and/or a joypad.

[barbudreadmon]

Player 2 can't join game in asterix (parent and clones), by looking at the driver file, i can say there is no "p2 start" defined.

[dink]

add this under the P2 Coin line:
   {"P2 Start",      BIT_DIGITAL,   DrvJoy2 + 7,   "p2 start"   },

[barbudreadmon]

Thanks

[Pityesz]

There is a little problem with G.I. Joe on Arcade. Seems like the sprites placed wrong.

[Gab75]

There is a little problem with G.I. Joe on Arcade. Seems like the sprites placed wrong.

I can confirm... there seems to be a wrong vertical alignment between the sprites/objects and the background (all Konami G.I. Joe versions). In MAME the issue doesn't exist (for a better explanation I attached some snapshots).

[barbudreadmon]

Could be related to an issue i didn't look into yet : G.I.Joe seems to freeze on fbalpha-libretro (whatever the platform) on title screen, and it tends to be quite sensitive to small bugs that don't break standalone.

[dink]

Pitysz, thanks, thats an easy one to fix (sprite offsets)

Unfortunately the problems are a lot deeper, and gi.joe is in a state of disrepair at the moment.  I spent several hours on it last night but no luck. 

I noticed was the tilemaps go completely bonkers in later levels, like clouds covering up half the top of the screen on an elevator level, and on the next attract sequence (like the 3rd of 4th) the tilemaps become even more broken.

barbudreadmon: Can you explain these little bugs a bit more, because maybe they will help me track down the issue?  is anything else happening besides the freeze on title?

best regards,
- dink

[barbudreadmon]

I'll try to get you as much information as i can. I didn't look into it yet, it was reported as an input issue to me (because people couldn't press start) but when i tried it on both standalone and libretro, i noticed it seemed more like some kind of freeze, also i remember being able to press start on one of the clone (but it would freeze on next screen). I suppose putting a lot of bprintf in d_gijoe.cpp to see when and why it stops (if it stops) could be a good start.

Edit : the clone thing seems like a fluke, actually the game seems to freeze after some time, and i can go to next screen only if i'm quick on the "press start".
Edit2 : gdb says nothing, game keep running, it just seems to draw nothing new, i don't know where to start, any suggestion ?

[dink]

re: g.i.joe: ugh.  I spent nearly all day one this one.  and a few hours the day before... gonna can it for now before I get burnt out. 

What I learned though:
I think the problem is in k056832.cpp, and it seems like the same or similar problem to mtlchamp, where certain tilemaps are covering up things that shouldn't be covered up.  like they aren't properly disabling, or being transparent.  in gijoe, (K056832ReadRegister(0x14) == 2) when things get screwy.  Seems like leftover tilemaps from the previous scene are being erraneously enabled on the next scene.  (fast forward to the elevator scene and the 2nd one after it to see)

Regarding the lockups in gi.joe, there's probably a memory overwrite in k056832.cpp or something or one of the other konami customs, and it may even be related to the problem above.  arm gets a freezeup, x86 gets screwy tilemaps.  seems fair, eh?

If you have any tools to validate variables/memory etc, and think they might help, please try.   I know this old p4 running winxp is probably no use for that kinda stuff.  iirc there is a commandline for gcc on linux that does this, but I honestly can't remember what the name is/was.  it had to do with checking memory overruns/out of bounds/mem validation.  anyone remember this?

best regards,
- dink

[Haze]

fwiw the elevator scene is broken in MAME too because old versions of MAME contained some truly horrific code to write directly to the tilemap rendering cache to implement some weird mode they use (which once the code was cleaned up to proper c++ and that bitmap made non-accessible outside the tilemap system couldn't be done for obvious reasons - it should never have been done like that)

the code in MAME is still very much not safe tho, makes lots of assumptions, so I wouldn't use that as a reference here ;-)

[dink]

wow, the amount of time into just fixing g.i.joe is astronomical at this point, and I'm going to need a good break now to ward off burn-out.

best regards,
- dink

[barbudreadmon]

If you have any tools to validate variables/memory etc, and think they might help, please try.

By the way, do you mean this ? http://www.gnu.org/software/libc/manual/html_node/Heap-Consistency-Checking.html
Or perhaps this ? https://gcc.gnu.org/onlinedocs/gcc/Code-Gen-Options.html

[Arcadez]

wow, the amount of time into just fixing g.i.joe is astronomical at this point, and I'm going to need a good break now to ward off burn-out.

best regards,
- dink

Nice work!!!