[ port-i386 added to the groups in case anyone there recognises this ]

In message <9604260007.AA03250@ gsd.harvard.edu>, Hacksaw writes:
>> Next problem!
>> 
>> I am trying to compile a new kernel, having editted up a new config
>> based on GENERIC minus a bunch of pointless devices. I even added
>> patch001.
>> 
>> I get two errors sporadically, Segmentation fault, and Internal
>> compiler error numbr 11
>> 
>> Generally, I get rid of whatever .o file it was working on, and try it
>> again, and generally it goes on farther, and then dies. Any ideas?

I get this as well.  This kind of error generally indicates some kind
of physical memory problem with the machine that is causing gcc's data
set to get corrupted.  However, I'm beginning to be less sure in this
case. Looks like you're running 1.1 (as is the machine I see it on),
and I have another 1.1 machine which periodically crashes in curious
ways, like in ufs_islocked() (from a sync() system call) by 
dereferencing a duff pointer in a vnode structure.  Again it looks a
bit like a physical RAM problem, but both machines only exhibit it
under heavy VM loads.  And on my own machine I never see windows
and processes randomly dying as I might expect if I had a duff SIMM.

So I wonder is there some VM bug which just occaisionally screws up
and maps pages wrongly?  And whether it still occurs in -current (for
various reasons I can't run a full current release on either of the
affected machines)

Pete.