Update Player Animation During Overworld Load

Submission Details

Name:	Update Player Animation During Overworld Load
Authors:	Kevin, zuccha
Added:	2023-11-27 09:49:47 PM
Type:	Game Mode
Includes GFX:	No
Includes Hijack:	No
Featured:	No
Description:	Update the current player animation while loading the overworld. In practice, this code fixes an issue where if Mario is placed on a water level tile on the overworld as his starting position, he has the walking animation instead of the swimming animation right after starting a new game (you need to move to start using the correct animation). This UberASM checks if the current tile the player is standing on is water, in which case we set the swimming animation, otherwise we set the walking animation. Known issue: in two-players mode, Luigi's animation will only be updated after he enters and exits a level.
Tags:	animation bug fix game mode 0d lorom overworld sa-1 swim
Comments:	5 (jump to comments)
Rating:	0.0 (0 ratings) No rating

Download 2.24 KiB | 86 downloads

Screenshots

View all

Comments (5)

MarkAlarm 2023-11-27 09:52:44 PM Link

Moderated with:
- Lunar Magic 3.40
- UberASMTool 1.6
- SA-1 Pack 1.40 (also tested without)
- Snes9x v1.60, bsnes-plus v05

The known issue is minor enough to where I don't think it warrants a rejection (though it would obviously be nice to have it be fixed in the future). Due to the original bug (incorrect animation) being minor as it is as well as the issue only appearing in a mode that rarely gets usage, I believe this is fine as is to accept.

Kevin Author 2023-06-28 08:03:40 AM Link

To fix the animation being wrong during fadein you can call the Mario drawing routine in your code right after .return:

Code

.return:
    phb
    lda.b #$04|!bank8 : pha : plb
    phk : pea.w (+)-1
    pea.w $048414-1
    jml $04862E|!bank
+   plb
    rtl

zuccha Author 2023-07-01 10:59:21 AM Link

Thanks!
I will update the submission in a couple of days when I find the time.
I still need to learn how to deal with different banks and invoking (jumping to?) original routines. From what I understand, instead of invoking $04862E (which ends with RTS, so you cannot JSL), you jump long to it? But it's still obscure to me how the routine actually returns to your code, does the entire bank setup you did serves that purpose? I don't actually expect an answer, since I suspect it's quite complex. I will read the ASM workshop document an hope to find my answers ;)

Kevin Author 2023-07-07 07:52:08 AM Link

Basically you trick the processor into RTSing to an RTL instruction, so that it can then return to your code.

Code

    phb
    lda.b #$04|!bank8 : pha : plb

This preserves the current DB and then sets it to $04 ($84 on lorom/fastrom), just because the routine likely uses tables in the same bank (it this wasn't the case, you wouldn't need to do this).

Code

phk : pea.w (+)-1

This pushes the (long) return address to the stack, which is what the RTL will use.

Code

pea.w $048414-1

This pushes the (word) address of an RTL in bank 4 to the stack, which is where the RTS will return to (needs to be in bank 4 since RTS can only jump to the same bank).

Code

jml $04862E|!bank
+

Now you jump to the routine (not JSL, as that would also push the long return address to the stack but we needed to setup the stack ourselves). Now when the routine ends it will do: RTS -> return to $048414 -> RTL -> return to "+" label in our code.

Code

plb

Now just restore the DB to the original value that was pushed at the beginning.

zuccha Author 2023-07-09 09:10:43 AM Link

Oof, I completely overlooked the fact that when you do JSR/JSL the return address is automatically pushed on the stack (which in retrospect is obvious, RTS/RTL need to retrieve it from somewhere). That clarifies a lot, thanks again!