seebs@solon.com sez this happens on m68k.  I just tried it on
NetBSD/sparc (an IPC), everything built from Feb 27 sup sources.  And I
see the bug too.  Probably something at least mostly MI....

Kinda makes me want to take all -O options out of the world builds.   I
already find that NetBSD/sun3 kernel builds fail due to an optimizer
bug; when the current build-of-the-world finishes on my /sun3 machine,
if it still happens, I'll generate a PR for that too.

					der Mouse

			    mouse@collatz.mcrcim.mcgill.edu