Getting started with WHDLoad development

[jotd]

Originally Posted by field3d

@Don_Adan If I hdr that change I receive the error ‘error 23: undefined symbol <Dx>’ Apparently Dx is not recognized.

This comment suggests that you need to practice a lot in asm.

Don't take it the wrong way. D0 - D7 are data registers. Dx would be a variable if declared.

[copse]

Originally Posted by jotd

Barfly has in-line opt directives (start of file). It is the historical assembler used by Bert and many others on the amiga. All slaves have a "IFD BARFLY" section with "BOPT" directives, one of them trying to add (pc) to addresses.

But that's bad practice. As you end up doing things like move.l d0,addr and hope that the optimizer adds (pc)...

Ah! Thanks for explaining that. That explains why the Pool of Radiance slave got away with this given that it has this Barfly legacy.

[field3d]

I see but for example in this case the affected parts are this ones:

_CD32_Common_Held						;Used for Pause button
		move.l		joy1(pc),d1
		lea		_held_button(pc),a1
		btst.l		d2,d1
		beq		.NotPressed
		btst.l		d2,(a1)
		bne		.clrkey
		bset.l		d2,(a1)
		move.b		d3,d0
		bra		.exit

The lines are the ones with error only when compile and I removed the .l and can compile but that part how you changed these lines form that code?

btst.l		d2,(a1)

and

bset.l		d2,(a1)

[field3d]

I made this test could figure to work?

_CD32_Common_Held                     
        move.l      joy1(pc),d1
        lea         _held_button(pc),a1
        btst.l      d2,d1
        beq         .NotPressed
        
        move.l      (a1),d0           
        btst.l      d2,d0             
        bne         .clrkey
        
        move.l      (a1),d0           
        bset.l      d2,d0             
        move.l      d0,(a1)           
        
        move.b      d3,d0
        bra         .exit

[copse]

Originally Posted by field3d

The lines are the ones with error only when compile and I removed the .l and can compile but that part how you changed these lines form that code?

btst.l		d2,(a1)

and

bset.l		d2,(a1)

I think the size is implicit and it is okay to remove it. My understanding is that the source operand is modulo 8 for byte accesses and modulo 32 for longword accesses. The size of the access is determined by the destination operand, only Dn (data register direct) destinations are longword values for purpose of testing the modulo 32 bit number. All other destination operand types are 8 bit references and modulo 8 presumably (manual states this for bset but not for btst).

If that is correct then "move.l (a1), d0; btst d2, d0" tests the longword at a1 and "btst d2, (a1)" tests the byte at a1. Hopefully someone will double check my logic.

[copse]

Since I no longer trust anything I have not seen run myself, I ran it through an m68k simulator (https://asm-editor.specy.app/):

lea $2000, a1
move.l #$FF000000, (a1)+
move.l #1, d1
lea $2000, a1
btst d1, (a1)               ; N flag zero after this line (all flags zero)
move.l (a1), d0
btst d1, d0                 ; N and Z flags set after this line (others zero)

The simulator rejected btst.l and would not compile it. This if correct shows that btst/bset "dn, (an)" is a byte access and btst/bset "dn, dm" is a longword access.

[field3d]

Oh excellent idea the simulator. I think is fine now.

[Don_Adan]

Something like this:

_CD32_Common_Held						;Used for Pause button
		move.l		joy1(pc),d1
		lea		_held_button(pc),a1
    move.l (a1),d0
		btst.l		d2,d1
		beq		.NotPressed
 ;		btst.l		d2,(a1)
    btst d2,d0
		bne		.clrkey
 ;		bset.l		d2,(a1)
 bset d2,d0
 move.l d0,(a1)
		move.b		d3,d0
		bra		.exit
.NotPressed
;	bclr		d2,(a1)
 bclr d2,d0
 move.l d0,(a1)
.clrkey		moveq		#0,d0
.exit		rts

You must remember that all bits related instructions works for 8 bits in memory, and for 32 bits in register.
Then bclr is also affected, not only btst, bset or bchg.
Of course this is next bug in original code.

[Don_Adan]

Originally Posted by jotd

This comment suggests that you need to practice a lot in asm.

Don't take it the wrong way. D0 - D7 are data registers. Dx would be a variable if declared.

Right, but seems that he can learn fast enough.
Also new look at old code can be useful.
I dont expected that many bugs exists in some WHDload slaves.
For me joypad code and second fire code cant works correctly.
If same code was reused in other slaves, then more slaves can be affected.

[jotd]

newest slaves with joypad code works properly since 2020 at least. He just picked an old one.

[Don_Adan]

For this game, no new slave, since 2011 year
I dont have joypad for tests.
But I remember some threads on EAB about strange working joypad buttons.