> Would you mind posting your 'dmesg'?  Mainly, I'm interested to know what
> your specific CPU type is; it's possible that gcc may be generating an
> instruction or two that aren't properly supported on your CPU(?).
> 
> (Note:  I've tested this on a Celery 400, and "ls -l" from the new build is
> working fine.  Hrmpf.)


Here it tis:

NetBSD 1.5Y (SCHNAUZER) #356: Thu Nov  1 02:57:11 CST 2001
    thieleke@yahoo.com:/usr/src/sys/arch/i386/compile/SCHNAUZER
cpu0: Intel Celeron (Mendocino) (686-class), 300.69 MHz
cpu0: I-cache 16 KB 32b/line 4-way, D-cache 16 KB 32b/line 2-way
cpu0: L2 cache 128 KB 32b/line 4-way
cpu0: features 183f9ff<FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,SEP,MTRR>
cpu0: features 183f9ff<PGE,MCA,CMOV,FGPAT,PSE36,MMX,FXSR>
total memory = 191 MB
avail memory = 164 MB
using 1200 buffers containing 19600 KB of memory


This is a Celeron 300A, non-overclocked, running in a generic 440LX
motherboard.  One peculiar thing about this system that may or may not
be relevant is that there is actually 256MB of RAM, but the system sees
the 128MB SDRAM stick as only 64MB.  The memory itself is fine, however,
since it was pulled out of a working system and seems to be ok when
using old toolchain binaries.

I dunno...maybe the hardware is marginal enough that gcc 2.95 exposes
some problem, but it still is completely stable with 2.91?  Has anyone
put together a recent NEW_TOOLCHAIN based i386 snapshot that I could
test?  The most recent version on ftp.netbsd.org is a couple months old,
so it was probably built with the old toolchain.



Jeff Thieleke