Hello,

>>which appears to be a macro: INTERLOCK_RELEASE(lkp, LK_SPIN, *s);
> 
> The compiler obviously shouldn't crash.  It would be helpful if you
> could report that as a gcc bug at http://gcc.gnu.org/bugzilla/.
> Attach the preprocessed source code--the .i file generated by the gcc
> command line with the --save-temps option.

No problem, I'll do that

> These look to me like a minor, unimportant, bug in the NetBSD
> sources--if the variables are static in sysv_msg.c, then it doesn't
> make sense to declare them as extern in msg.h.

It seems strange that gcc 3.4.2 does not complain about it.
Also, there are those message about 'signedness'.

GCC tends to be stricter as time passes. Also, I was a bit disappointed, 
even if that doesn't regards the kernel : bladeenc appears to be twice 
slower when compiled with 4.0 than with 3.3.3. Go figure :)

Vincent