Re: [hatari-devel] Wrong bus error stack frame with 68000 and JSR

[Nicolas Pomarède <npomarede@xxxxxxxxxxxx>]

Le 19/03/2024 à 21:45, Christian Zietz a écrit :
> Hi,
>
> the following bug report is probably something for Toni, but I don't
> know how to reach him.
>
> I investigated why Spectre (the Mac emulator) crashes under Hatari. This
> has been reported before, but never resolved or investigated in detail
> as far as I can see:
> https://www.atari-forum.com/viewtopic.php?p=451553#p451553
>
> I have found out that it crashes while handling an expected bus error,
> because the stack frame is different on Hatari compared to a real 68000.
>
> Consider the case where a JSR is executed to a target address that
> causes a bus error. This happens on the emulated Macintosh, when
> something tries to jump into the Mac ROM above 0x400000. Spectre has
> code to handle this, but it chokes on the incorrect stack frame.
>
> A test program is attached, as well as results from a real 68000 (.jpg)
> and from Hatari's 68000 emulation (.png). As you can see, on Hatari the
> stacked PC is the address that causes the exception (0x401234 in my test
> program), i.e., *after* the JSR. Whereas on the real 68000 the stacked
> PC is still in ST-RAM, i.e., *before* the JSR.
>
> Note that this is not a regression. A similar crash is observed with
> Hatari 2.4. Nevertheless, I think it should be fixed.
>
> Regards
> Christian


Hi

thanks for your test, this look indeed like an error.

Toni has written a rather extensive "cpu validator" program that runs on 
real amiga and on WinUAE and by comparing real HW vs emulation this 
helped in the past finding non correctly emulated features (such as 
undocumented behaviour for ZCNV bits, wrong value on the stack and so on)

But as Amiga doesn't really use bus error (except for some external 
board), it's quite possible some cases were missed.

Anyway, I'm cc'ing Toni, but I'm sure the subject of your mail alone 
would catch his attention as he's subscribed to this mailing list :)

Nicolas