Hi all,

I've finally completetd a complete rebuild of my userland from -current
source (cheaper for me than pulling down the latest snapshot ;-)) and
have recompiled the kernel from exactly the same source that used to
hang somewhere right after entering _splnone().

What can I say? IT WORKS ;-) 

Somehow a 1.4.1 user-land seems not be appropriate to compile a working
-current kernel... And I did a 'rm -rf <compile dir>; config; make
depend; make' raindance several times before with the old user-land to
no avail :-(

Sorry for the (half-)false alarm, but I'm still curious if it really was
the compiler (or maybe the assembler) that generated some code that the
poor R3000 was unable to comprehend ;-)

bye for now,
Marc