On Mon, Nov 11, 2002 at 12:19:19AM -0600, Dan Debertin wrote:
> I'm seeing completely random compiler deaths while building a kernel
> with a new 1.6 install on my ELC. It tends to die on the
> high-memory-usage bits like nfs_vnops.c, though it is not at all
> predictable. Even the signal is sometimes SIGSEGV and other times
> SIGBUS.
> 
> Common wisdom says to suspect my hardware. I have 64MB of ram in 4
> SIMMs, each of which I have swapped with a new one. Strangely, the
> problem does not occur -- or at least I can complete the entire kernel
> compilation -- if I back my memory down to 16MB, but I don't want to
> run in that configuration forever.
> 
> I'm running completely diskless, mounting from a Solaris machine and
> swapping to a file (tried a range of swap sizes; 16, 64 and 128MB).
> 
> There are a few posts in the archive about issues with sun4c
> machines. Is this a known issue (hopefully with a known workaround?)?

There are problems with *some* sun4c machines. For now it's known to
happen on SS1, SS1+ and IPC machine, and seems related to the type of
CPU cache.
What does the dmesg or your machine looks like ?

-- 
Manuel Bouyer <bouyer@antioche.eu.org>
     NetBSD: 23 ans d'experience feront toujours la difference
--