Re: [hatari-devel] Possible bug in 060 exception stack frames

[Thorsten Otto <admin@xxxxxxxxxxx>]

On Sonntag, 17. Januar 2021 20:10:36 CET Toni Wilen wrote:

 

Hm, i now pulled current Hatari again (including the fix you posted earlier), and the crash went away. Last debug output had shown that the SSP that was used during the exception was in the range of the user stackpointer, so this fix seemed to be needed even when the instruction was executed in user-mode.

 

The integer tests now run as expected. The floating point tests still fail, but don't crash. If you still get crashes there (on amiga), that might be because of assembler bugs, and you should try using the original *.sa (either from my repo, or from somewhere else, they should be identical, all that i changed in some of them was to replace the "long" directives by .dc.l)

 

Now trying to figure out why the floating point tests fail. Its a bit difficult because the tests only print the first failing test number, then abort. They don't even tell which values were expected, and which are produced.