On Mar 30, 2011, at 5:01 06PM, AxiNat wrote:

> Hi Henrik and list, I was just asking how can I do the research by myself to find out where the bug comes from.
>
> Anyway, here's my config:
>
> Debian GNU/Linux wheezy/sid (testing) - Kernel version 2.6.32-5-686
>
> CPU: Intel(R) Pentium(R) Dual ��CPU ��T3400 ��@ 2.16GHz
>
> Smalltalk version 'Pharo1.2'
> Smalltalk vmVersion ��'Croquet Closure Cog VM [CoInterpreter VMMaker-oscog.35]'
>
>

Then I'd say there's probably a bug in code generation of the Cog JIT'er, seeing as how it doesn't crash right away.
Use of SSE2 instructions would be my first bet, but it looks like your processor supports those.

��First I'd try upgrading to a recent VM, .35 is really old, and there have been plenty of crash fixes since then.

If that doesn't help, try to identify which piece of code gets JIT'ed right before the crash (using logs/a debug-built VM/gdb etc.), and try to create a reproducible test-case of the code that generates invalid instructions when JIT'ed.

Then, submit testcase/details you just provided here to the Cog issue tracker, cross your fingers, and hope for the best :)

Cheers,
Henry
��