Re: [hatari-devel] Possible bug in CPU emulation?

[Nicolas Pomarède <npomarede@xxxxxxxxxxxx>]

Le 07/10/2019 à 21:51, Christian Zietz a écrit :
> Thorsten Otto schrieb:
>
> > As reported here <https://github.com/freemint/freemint/issues/127>, a
> > user created a disk image that works with Hatari, but crashes on real
> > hardware. The culprit seems to be the CBHD driver, but i would have
> > expected that Hatari shows the same error.
>
> I'd rather bet on a slightly different behavior of the emulated ACSI
> drive compared to the real HW the people in the bug report are using.
>
> Also, I would still investigate the original crash (regardless of
> whether it is reproducible in Hatari). Might still be a bug in XaAES or
> MiNT, that is just triggered by CBHD.

Hi

I agree that cpu bugs are always possible, but given the "complexity" of 
the reported setup, it's quite likely that the combination of several 
parts might be creating this bug, cpu emulation itself is really close 
to real HW (except for cycle accuracy, which might explain the differences)

As you say, maybe the bug is in one of the driver, but it never shows on 
real hardware "by luck" because of slightly different timing compared to 
a less accurate emulation.

I second your opinino that if it crashes on real ST but not under 
Hatari, the path to follow is indeed to find why it crashes on real HW, 
and then we can see why emulation doesn't create the same context for 
the crash. But doing it the other way around, by trying to guess where 
Hatari should crash instead of working correctly will just be too 
tedious and nearly impossible to solve.


Nicolas