Re: [hatari-devel] 68030 and movem

[Andreas Grabher <andreas_g86@xxxxxxxxxx>]

> Am 17.09.2023 um 09:39 schrieb Andreas Grabher <andreas_g86@xxxxxxxxxx>:
> 
> 
> 
>> Am 15.09.2023 um 09:33 schrieb Nicolas Pomarède <npomarede@xxxxxxxxxxxx>:
>> 
>> Le 14/09/2023 à 23:42, Cyprian Konador a écrit :
>>> Hi Team,
>>> Agranlund has faced an issue with movem instruction and 68030 emulation under Hatari: https://atari-forum.com/viewtopic.php?t=43071 <https://atari-forum.com/viewtopic.php?t=43071>
>>> I've tested on the real TT and under the following build: http://antarctica.no/~hatari/latest/windows/ <http://antarctica.no/~hatari/latest/windows/> and Hatari shows me a different result than  the TT.
>>> Attached you can find screenshots, test program and its source code.
>>> Regards
>>> Cyprian
>> 
>> Hi
>> 
>> thanks for the test program ; I already sent a mail about this to toni, hopefully he can fix this with the help of your program's result.
>> 
>> Nicolas
>> 
> 
> I made a quick patch. Please check if this solves the problem. WARNING: It is untested and might require refinements regarding bus error handling. I think Toni has the best overview. 
> 
> The patch also includes a minor patch for softfloat that makes the code a bite more safe (no more shifting by more than the type's size).
> 
> <test.diff>
In my first patch I interpreted "size of the operation" as the total size of all data written to memory. It might as well mean size of one single transfer, which might be word or long. My second patch is for that case. I think some testing is required to find out which patch is correct.

Attachment: test2.diff
Description: Binary data