On Sun, 29 Feb 2004, Simon Burge wrote:

> On Fri, Feb 27, 2004 at 12:09:54AM +1100, Simon Burge wrote:
>
> It looks like this problem has been around since November 2002(!).  I've
> fixed this with rev 1.102 of sys/arch/mips/mips/vm_machdep.c.  The fork
> pthread regression test now works.  If some folks can try more complex
> programs and report the results I'd be appreciative.
>
> Simon.

Hi,

I compiled new kernel with 1.102 version of the file (I did _NOT_ build
userland. JUST kernel. Stupid, since version has changed, I know. I'll
rebuild all tomorrow and try again).

My first results are:

- My test program runs fine _IF_ the threads don't have to much work
(some calculation loops with heavy memory usage in the threads, reducing
loop counters makes program work). If each tread runs longer, the program
suddenly stops, system goes 100% idle and program's state in "top"
switches  to "sawait". ("sa" == "scheduler activations" ??).
If I start a second instance of the program while the first is running,
I *sometimes* get "Abort trap".

- The MozillaFirebird now installs. But some helper program dring install
crashes (Segmenation fault). The MozillaFirebird executable itself also
crashes with "Segmentation fault".

But I never got a hard lock or reboot with the new kernel.

That's all for the first test. Another post follows as soon as I rebuilt
everything.

Florian