exxos's Atari, Amiga & retro forum

agranlund wrote: 15 Mar 2022 03:46 Being able to debug it in Hatari is surely going to help :)

No problem.

It's right at the very start, before the framebuffer is intiialised but after the mouse pointer is hidden. You'll still be able to access menus and click on the desktop after the crash, but can't see the pointer (pressing ALT-P or some other drive letter that doesn't exist to open a window on that drive restores the mouse pointer as Teradesk opens an alert box).

Running it from bash in toswin2 just prints 'Bus Error'. Logging doesn't seem to do anything so perhaps isn't even initialsied yet either.

This might be a very small and immaterial bug, or it might lead you to something else, but hopefully it's reproducable now. :)

BW

Badwolf wrote: 15 Mar 2022 10:09 This might be a very small and immaterial bug, or it might lead you to something else, but hopefully it's reproducable now. :)

That was a great catch! It would cause all kinds of seemingly random issues, or no issues at all if in luck...

uint32 cookie;
...
(getting some other cookie in the correct way)
...
isMagic = (Getcookie('MagX', cookie) == 0);
isMint = (Getcookie('MiNT', cookie) == 0);

Getcookie wants a pointer. It should of course be Getcookie('MagX', &cookie).

This was trashing random memory based on the value of previously correctly gotten cookies.
In our case the last one was a Getcookie('_FPU', &cookie) so what memory location it was corrupting depended on which type of FPU you got :)

That error was so easy to glance over when looking at the code, but ended up obvious thanks to your repro package.

It boots now with that fixed. It may fix other hidden issues too depending on what memory it had corrupted :)

Great work, Anders.

I love an easy win. :thumbup:

BW

agranlund wrote: 15 Mar 2022 11:13Getcookie wants a pointer. It should of course be Getcookie('MagX', &cookie).

Somebody ignored warnings and/or -Wall I presume? ;-)

mikro wrote: 15 Mar 2022 17:46

agranlund wrote: 15 Mar 2022 11:13Getcookie wants a pointer. It should of course be Getcookie('MagX', &cookie).

Somebody ignored warnings and/or -Wall I presume? ;-)

Haha, guilty as charged :lol: :oops:

Works well with my PAK machine ...

030 /50Mhz (32kB L2 Cache) + FPU + 64MB Fastram


With my MACH64 Graphiccard and Nova driver it do not work !

frank.lukas wrote: 15 Mar 2022 19:03 Works well with my PAK machine ...
030 /50Mhz (32kB L2 Cache) + FPU + 64MB Fastram
With my MACH64 Graphiccard and Nova driver it do not work !

Nice!
I'll give those Nova drivers a try on my machine as well (though with an ET4000 on an STGA interface).

I remember making an attempt with these on my computer a long time ago but remember having some trouble getting it set up correctly. That could have been because of my particular hardware, or just me not understanding it correctly.

I test only 1280x1024 at 16,7M Colors. Today I test mono and 256 Colors ...

Black and White are not possible with the Nova MACH64 driver. 256 Colors works well but the MacOS make 256 greyscale from the 256 Colors !?!

Can you not switch between monochrome and colour using the Monitors control panel?