Re: [hatari-devel] Moongame, question

[Laurent Sallafranque <laurent.sallafranque@xxxxxxx>]

Hi,

> If that was the case, the game would crash on falcon too.
> The fact that the crash happens at the vbl start is not a coincidence 
for me. Maybe due to some timings issue the VBL happens *before* the 
program set dd21a when running under Hatari (there're some similar 
issues on STF).
> For me there's no hint to an ikbd problem in the traces you sent.

OK, thanks for the analisys.

I've added the DSP (16 bits) --> crossbar (8 bits) transfer in 
synchronous mode.
It doesn't fix the problem, but at least, it may solve some other demos 
with bad sound.

I'll upload this patch when finished, and return to moongame to try to 
fix it.

Regards

Laurent




Le 27/01/2012 21:01, Nicolas Pomarède a écrit :
> On 27/01/2012 18:42, Laurent Sallafranque wrote:
> > Hi,
> >
> > I've continued to investigate on moongame :
> >
> >
> > > The dbf at 5e71a is interrupted by a new VBL (video_cyc=0), but the
> > exception executes code at RAM $0, this is highly wrong and is the cause
> > for the crash.
> >
> > > For me, the problem is in
> > > $05e6ec : 21f9 000d d21a 0070 move.l $dd21a,$0070.w
> >
> > > I think $dd21a is not correctly initialized and contains $0 instead
> > of the address of the VBL handler.
> >
> >
> > I've logged 12M instructions (2 Go of datas) for the beginning of
> > moongame to the ILLEGAL instruction.
> >
> > Then, I've used a grep "dd21a" to find where it is initialized.
> > In fact, it is not. I guess it does an ILLEGAL on the falcon too.
>
> If that was the case, the game would crash on falcon too.
>
> The fact that the crash happens at the vbl start is not a coincidence 
> for me. Maybe due to some timings issue the VBL happens *before* the 
> program set dd21a when running under Hatari (there're some similar 
> issues on STF).
>
> For me there's no hint to an ikbd problem in the traces you sent.