Quoting Tino Keitel <[EMAIL PROTECTED]>:

  I suggest compiling XFree86 with the optimization flags, but also with
-g, and run it under the custom gdb at ftp://ftp.xfree86.org/pub/xpert/gdb
  Do it from a remote computer.

  It is dependant on your compiler version, snapshots may have bugs. But
most times I found this sort of problem and could figure out what was wrong
(sometimes I just find the chunk of code that is breaking and rewrite it...),
it was some program error, like writting past the start/end of a stack
buffer, uninitialized data that happens to have the same contents after
some time (anything that was left in the heap after malloc/free calls), etc.

> > Seem strange to me that a software like XFree86 cannot use
> optimization 
> > FLAGS durring compile time. Here is what I used as optimization -->
> "-O2 
> > -march=i686 -funroll-loops" and XFree86 fail. Again without the
> previous 
> > optimization FLAGS, all run ok.
> > 
> > For me, this mean that XFree86 cannot be compiled to get advantage of
> i686 
> > CPU architecture :((
> 
> I would think about this in a different way. In my point of view, this
> is a code generation bug in the compiler (gcc?). In a perfect world,
> any piece of source code should compile and work perfectly even with
> the most exotic optimization flags turned on.
> 
> I also had this problem with other software. Most of the time I could
> determine _one_ source file that will be broken with _one_ certain
> optimization option. I would at first try to build without the
> -funroll-loops option.
> 
> Regards,
> Tino
> 
> -- 
> [EMAIL PROTECTED]
> dipl.-inf.                        Innominate Security Technologies AG
> software engineer                                   networking people
> tel: +49.30.6392-3308                      http://www.innominate.com/
> _______________________________________________
> Xpert mailing list
> [EMAIL PROTECTED]
> http://XFree86.Org/mailman/listinfo/xpert

Paulo
_______________________________________________
Xpert mailing list
[EMAIL PROTECTED]
http://XFree86.Org/mailman/listinfo/xpert

Reply via email to