Re: [hatari-devel] TT with 8 MHz crashing (with NVDI)

[Nicolas Pomarède <npomarede@xxxxxxxxxxxx>]

Le 17/09/2024 à 12:10, Uwe Seimet a écrit :
> Any chance that this will be fixed? It's probably not that relevant, but
> still an item on my list of open Hatari issues.

It's also flagged as TODO in my list of emails :)
Can't tell when I will have a look at this, I have other pending items.

Nicolas

> > Le 02/09/2024 à 14:55, Uwe Seimet a écrit :
> > > > On Montag, 2. September 2024 13:29:48 CEST Uwe Seimet wrote:
> > > > > I doubt that NVDI behaves differently with 8 MHz than it does with 16 or 32,
> > > > > so this may be an issue with Hatari.
> > > >
> > > > There is definitely a _CPU check in all NVDI versions, however depending on
> > > > the Mhz would really be strange. Which NVDI version are you using?
> > >
> > > 5.0.3
> > >
> > > I have identified the commit that causes the crash:
> > >
> > > * 9dd63033 - Revert part of "Fast 68020+ modes didn't start pending interrupt im
> > > mediately" fix (from WinUAE 5.1.0+ 2023/12/29) only SR modifications must check
> > > interrupt immediately (8 months ago) <Nicolas Pomarède>
> > >
> > > With 14bcf9b0 and f437b8fd there is a bus error sometimes (only with 8 Mhz),
> > > but not the crash (4 bombs) I see with 9dd63033.
> > >
> > > 47f3ebbe never crashes.
> > >
> > > Best regards
> > >
> > > Uwe
> >
> > ok, so that would maybe be an incomplete fix for the case in the commit
> > message and it would trigger the error more easily at 8 MHz because of a
> > race condition that makes it harder to see it usually at 32 MHz.