Re: [hatari-devel] Pure debugger, illegal opcodes, exceptions

[Thomas Huth <th.huth@xxxxxx>]

Am Tue, 9 Oct 2012 22:26:31 +0200
schrieb Uwe Seimet <Uwe.Seimet@xxxxxxxxx>:

> > As PD works even with WinUAE core when using TOS v1.04 (but not TOS
> > v1.62 which still implies 68000 i.e. similar CPU exception frames),
> > do you know whether there is some difference in how that handles
> > exceptions compared to newer TOS versions?
> 
> I'm afraid I don't know.
> 
> > Interestingly even the output with TOS vv1.04 differs between
> > Hatari versions.  With WinUAE core:
> > 	Autostart file 'DESKTOP.INF' for 'PD.PRG' matched.
> > 	Autostart file removed.
> > 	Exception 2 (0) at fc0ee4 -> fc0ee6!
> > 	Exception 2 (0) at 2cd76 -> 2cd86!
> > 
> > Whereas with old UAE core:
> > 	Autostart file 'DESKTOP.INF' for 'PD.PRG' matched.
> > 	Autostart file removed.
> > 	Bus Error at address $fffa42, PC=$2cd76
> > 
> > I.e. there was no bus error at fc0ee4.  According to Compedium:
> > 	"On pre-TOS v2.00, this location marked the beginning of
> >      the operating system ROMs."
> 
> So there are more bus errors (the one at $fc0ee4) with the new core.
> Looks as if there is a general difference wrt the handling of bus
> errors (when reading data from write protected memory areas?) in the
> two cores.

I remember that we had to fix bus error handling in the old UAE core a
couple of years ago ... maybe the WinUAE core also does not use the
right stack frames yet (at least the 68030 MMU mode uses the wrong
stack frames as Andreas already discovered).

I'm quite busy this week, but if somebody wants to have a look, check
the exception functions in cpu/newcpu.c for the right bus error stack
layout...

 Thomas