On Fri, Nov 20, 1998 at 04:22:50PM -0500, Ken Hornstein wrote:
> >For the last 3-5 days (since about the time of detachable devices ?),
> >current kernels on my Quadra 610 have been unusable. They die,
> >typically with 'Panic: MMU fault trap" during rc file processing.
> 
> A co-worker has seem a similar thing; he believes it to be a optimization
> bug.  Can you try compiling the kernel without optimization?

Doing this produces a kernel that hangs early in boot, just after printing
the copyright string, but before printing the NetBSD 1.3I line. Maybe something
about locore doesn't like non-optimized code ?

As I must admit that I use egcs, I did another try: I reinstalled the
non-egcs gcc (src/gnu/usr.bin/gcc) and used that to completely rebuild
the kernel (with the default -O2 again). This kernel fails just like
the  egcs-compiled kernel, ie. panic: MMU fault trap.

So, it may be optimization related, but then it applies to both the old
gcc and egcs. I'm really more inclined to believe in a problem in the
kernel sources, possibly m68k specific, as I don't see this problem on
neither my i386 nor my pmax.

I may try some combination of mixed optimized/non-optimized compilations :-(

- Erik