SEGA CD Mode 1 (via MSU) is not supported

[RobiTheGit]

This is the code I use to test if there is a SEGA CD and a Disk inserted

    tst.b    (SegaCD_Mode).w ; Test if there's a SEGA CD
    bne.w    .nocd ; if not, don't do the sega cd thing
    btst    Clownacy/clownmdemu-frontend#5,($A10001).l ; Test for a lack of disk (This bit is 0 when there is a disk)
    bne.w   .nocd ;If there is no disk, don't do the sega cd thing
   <CD FUNCTION HERE> 
.nocd:
    ...

Normally, it should do what ever function is after this code if there is no CD, or run the command to the SEGA CD

Instead, it black-screens the emulator.

[RobiTheGit]

A more specific code example

MCD_Unpause Macro
    tst.b    (SegaCD_Mode).w ;IS the SEGA CD attached
    bne.s    .nocd ; if not, return
    btst    #5,($A10001).l ; IS there a disk
    bne.w   .nocd ; if not, return
.wait
    tst.b    (MCD_Status).l ; Are we able to talk to the SEGA CD
    bne.s    .wait ;if not, wait
    move.b    #_MCD_UnPauseTrack,(MCD_Command).l ; Send Unpause command
    addq.b #1,(MCD_Command_Clock).l ; Keep the SEGA CD going
.wait2
    tst.b    (MCD_Status).l        ; Waiting for the first command to be executed
    beq.s    .wait2 ; if it hasn't been executed, wait
.nocd:
    endm ; return 

This code works on Genesis Plus GX and BlastEM Nightly

[Clownacy]

Could it be that .wait2 is being branched to infinitely? The 68k debuggers should be able to show you what each CPU is doing.

[RobiTheGit]

PC Flips between 430C and 4312, nothing else changes

I bring this issue up, because this same MSU handler crashes in other situations (Try building Sonic Clean Engine with MSU Mode set to 1)

Removing the test for $A10001 will hang BlastEM when there is no disk inserted; but with the check in place, (as I've said) BlastEM and Genesis Plus GX (both of which are quite accurate emulators) work fine (Kega does too, but that's not really saying much)

The main MCD module being used is krikzz's msu-md, and adapted to Sonic Clean Engine

[RealMalachi]

PC Flips between 430C and 4312, nothing else changes

Knowing what code those addresses correspond to in rom would help, but given the short distance of the addresses it looks like an infinite loop

[Clownacy]

That sounds like .wait2 is being branched to infinitely. Since .wait2 is waiting for the Sub-CPU, that would mean that it is the Sub-CPU which has hanged, not the Main-CPU.

If I remember correctly, MSU is one of those things that bypasses the Mega CD's BIOS and interfaces with the underlying hardware (such as the CDC) directly. My emulator does not emulate the Mega CD's underlying hardware, instead performing high-level emulation of its BIOS. Because of this, MSU is not supported by my emulator; this would only change if low-level emulation of the Mega CD was added.

This problem could be worked around by providing a way to disable Mega CD emulation altogether, so that the ROM-hack does not try to use MSU in the first place.

[RobiTheGit]

Currently, I need to find an emulator detector that detects ClownMDEmu (and not show up as something that actually does support the SEGA CD's hardware) so that I can make it automatically disable the SEGA CD.

[Clownacy]

That would not be ideal, since that would cause Sega CD functionality to be disabled even in future versions of ClownMDEmu which do support MSU. An alternative is to detect when low-level emulation is unavailable, such as by attempting to perform a handshake with the CDC.

Another, most robust, way would be to implement a timeout in your code, so that after an arbitrary amount of time has passed, it stops waiting for the command to register and just assumes that Sega CD support is unavailable, like this:

MCD_Unpause Macro
    tst.b	(SegaCD_Mode).w	; IS the SEGA CD attached
    bne.s	.nocd		; if not, return
    btst	#5,($A10001).l	; IS there a disk
    bne.w	.nocd		; if not, return
.wait
    tst.b	(MCD_Status).l	; Are we able to talk to the SEGA CD
    bne.s	.wait		; if not, wait
    move.b	#_MCD_UnPauseTrack,(MCD_Command).l	; Send Unpause command
    addq.b	#1,(MCD_Command_Clock).l		; Keep the SEGA CD going
    move.w	#$8000,d0
.wait2
    tst.b	(MCD_Status).l	; Waiting for the first command to be executed
    dbne	d0,.wait2	; if it hasn't been executed, wait
    
.nocd:
    endm ; return 

You could also try to submit a command when the game boots, and set a flag to disable Sega CD support if the timeout occurs.

[RobiTheGit]

That's a good idea, I'll implement that when I can get to it.

[RobiTheGit]

I went and tried this, but got the same result as not having the check. I think it's probably halting when the MCD command is sent, as if there is no CD, and you send a command, it halts the system

[Clownacy]

Then it may be getting stuck in a the .wait loop instead. If you use the emulator's 68k disassembler, or your listing file, you can check what code is at 0x430C to see where exactly it is getting stuck.

[RobiTheGit]

It was getting stuck in that first loop, and so the game now advances now, just without CD audio. I just need to figure out how to skip to the end of the macro if it doesn't pass the dbne.

[Clownacy]

You could check if the register's lower word contains $FFFF, since that is what the register is set to when the timeout expires.

An efficient way to do this would be to increment the register, and then branch depending on if the result was equal to 0:

.wait
    tst.b	(MCD_Status).l
    dbeq	d0,.wait
    addq.w  #1,d0
    beq.s   .nocd