> It seems *very* likely to me that you're either using a very old boot
> block, or you have an inconsistent source tree.

I don't see how the boot blocks would affect things, since the kernel
is loaded and autoconfiged. But they are from NetBSD-1.0 (Rev 1.21.1.1,
I think).

It may well be an inconsistent source tree; but if so, then the tree has
been inconsistent for over a month (since this happened with source trees
from early December and last weekend). And Mark Weaver's kernel
has the same effects, and it does boot successfully for him.