I have personally rewritten Getcookie() in assembly language last year for ColdFire compatibility. It is used by the startup code of every program compiled with GCC, and it works fine, in any case.
Well, I didn't inspectet it that deep - just read the kernel documentation about it and there was written that one shouldn't use it on plain TOS systems. Now I'm using a compatible method when running under classic TOS.
It's about the EddiCookie... which contains an function pointer as value. But the problems happens with or without Eddi Cookie - so I believe calling the Getcookie() function is enough to cause a crash.
The causes can be:
- stack overflow on the user stack (try to increase it using the "stack" tool, it works for TOS, too).
I already did that, but that did not help. But the problem is already "fixed". See my other mail :/
This looks wrong. A6 should point to your own stack, and that value can't be correct. You should try to see when A6 goes wrong.
It had the value $01000 after UNLNK instruction, before that it was something like $08012344...
This is what happens when the system/AES stack overflows. I had same trouble when debugging USERDEFs.
Hm, no, I doubt that, no USERDEFS called during the function. As I said - I already have found my mistake... I believe... ( altough I think a program compiled without -m68000 flag should run as good as an program compiled without the -m68000 flag..., even when it is linked against -m68000 only libs).
Greets, m