Re: [hatari-devel] WinUAE core freeze with ST emulation

[Laurent Sallafranque <laurent.sallafranque@xxxxxxx>]

OK, I've finaly found it : it's in m68000.c

/**
 * Reset CPU 68000 variables
 */
void M68000_Reset(bool bCold)
{
    /* Clear registers */
    if (bCold)
    {
        memset(&regs, 0, sizeof(regs));
    }


The memset clears the regs (and spcflags is into this structure).

It's because we come from change.c which calls Reset_Cold();

Should I change the reset cold by a reset warm or add some ifs to avoid 
the problem ? (This may complexify the understanding of the code).

Laurent





Le 16/01/2012 22:00, Thomas Huth a écrit :
> Am Mon, 16 Jan 2012 00:18:43 +0100
> schrieb Laurent Sallafranque<laurent.sallafranque@xxxxxxx>:
>
> > > If I boot with "--machine falcon --tos etos512k.img" and in
> > > the settings select ST / 68000 / 8Mhz / TOS 1.02, them the old TOS
> > > boots up fine too without saving settings, even with prefetch
> > > enabled.
> > It's only by luck, as the wrong CPU is running after the reboot.
> >
> > As I explained before, changing of CPU in the GUI makes the code run
> > into newcpu.c:check_prefs_changed_cpu (void)
> > and at the end of the function, it goes to
> >
> >       if (changed)
> >           set_special (SPCFLAG_BRK);
> >
> >
> > But when the code returns to the running CPU (like m68k_run_1 (void),
> > m68k_run_2 (void), ...) the value of SPCFLAG_BRK is back to 0 and the
> > program doesn't return to m68k_go (that's where the change of CPU
> > occurs).
> >
> > I haven't managed to find this problem, and it's quite annoying. Some
> > help would be appreciate on this one.
> I remember having some similar problems a couple of years ago with the
> uae-cpu core. Some quick observations:
>
> 1)
> - check_prefs_changed_cpu() in uae-cpu/hatari-glue.c sets SPCFLAG_MODE_CHANGE
> - check_prefs_changed_cpu() in cpu/newcpu.c sets SPCFLAG_BRK
>
> 2)
> - uae-cpu/newcpu.c m68k_reset() does:
>    regs.spcflags&= ( SPCFLAG_MODE_CHANGE | SPCFLAG_DEBUGGER );
> - cpu/newcpu.c m68k_reset() does:
>    regs.spcflags = 0;
>
> ==>  I guess one of your problems is that m68k_reset() (which is
> certainly called when changing the CPU level) clears SPCFLAG_BREAK so
> that the m68k_run-loop is never left.
>
>   Thomas