>> Okay, after seeing how -O2 broke on what was it, nfs_vfsops.c I
>> think, I took it completely out of sys/arch/sun3/conf/Makefile.sun3
>> and built a kernel completely without optimization.  [...]
> I'm confused.  You took -O2 out, or you took nfs_vfsops.c out?

The former (as I suppose one could argue was implied by the "completely
without optimization" bit).

> What happened?  I don't understand what happened.

Neither do I. :-)  The coredump on the NFS file is obviously a compiler
bug, and the differing misbehavior indicates there's some other bug
somewhere.

I'm hoping all this goes away when the sun3 port is switched to the new
gcc.  Once I can accord to crash the machine with experiments (I have a
big data transfer in progress at the moment), I'll probably try
frobbing gnu/usr.bin/Makefile to use the new gcc instead of the old and
see how it goes.

					der Mouse

			    mouse@collatz.mcrcim.mcgill.edu