> Three times this week now my machine has frozen up completely, requiring a
> stop-A and boot to get back to life. This happens roughly one in four times
> when I try to link octave, a rather huge programm for mathematical simulations.
> With 80704 k swap, in use for 78%, the linker has repeatedly hung my system.
> I admit that a linking stage that takes about 64M of virtual memory and 20
> minutes to process is a bit huge, but still, it's userland, and should not
> be able to crash my machine. Does anyone have suggestions on how to isolate,
> test, document, patch or whatever this weakness? 

It might be thrashing in paging/swapping/mmu area..

If you compile a kernel with DDB you can do `stop-A' a couple of times to
find out what the kernel thinks it should be doing with its time (use
the DDB `trace' command).

-pk