pmax getting back to being buildable

Hi all,

back in April 2010 I reported problems in using a DS5000 to rebuild
its NetBSD-current.

I had some feedback and in several steps (including matt-nb5-mips64
being merged to HEAD as well as other kernel updates) the situation
improved during Spring this year to a state where I could boot the
machine to single-user state and transfer a crossbuilt kernel, libc,
and /bin/sh to the pmin and then recover the machine to a bootable
state.

Installing furher parts from the cross-built NetBSD improved the
situation to a state where I could update the sources on the DS5000
using cvs and rebuilding the kernel locally.

I still had some severe problems in running build.sh -- after a
shorter or longer period of working through the build, it would hang
silently. Ctrl-T on the console indicated that no CPU time was
consumed in user-mode, but the load average reaised steadily.

If caught early enough, Ctrl-C would stop the build, and I could do
another attempt. At other times the only way out was a hardware reset.

This way I never got it to complete a full 'sh build.sh -u tools' even
if I commented the parts completed out of tools/Makefile to reach
further ahead.

This has been the case for kernels based on sources up to about 17th
July. But now on a kernel build from sources of 27th July, it suddenly