On Mon, Dec 10, 2001 at 09:33:16AM -0800, Tracy Nelson wrote:
> > The symptoms you've shown are typical for bad memory;
> > have you run some extensive memory checks on the hardware?
> > The BIOS POST memory-counting is often not enough, there are
> > free memory checkers available on the net, some of which also
> > tell you which bank contains a faulty SIMM.
> 
> You can sometimes also verify this by swapping [SD]IMMS and seeing if the
> failure address changes.  Not the most reliable way but hey, it's cheap!
> 

Also keep in mind that bad disks can masquerade as bad memory.  Some
years back I had a box that refused to reliably complete a large
compilation.  We swapped the RAM, but the problem didn't go away
until the SCSI data disk was replaced.  

The disk wasn't obviously bad.  The problem only manifested itself
when several programs were being compiled simultaneously.  We
attributed the problem to the on-disk cache, but I don't remember any
smoking gun.  

My experience wasn't with NetBSD, and I don't how NetBSD would deal
with a similar situation. 

--jkl