I finally have my cross-compilation environment:  as, ld, ar,
ranlib and gcc.  I saw a rather frightening bug when running NetBSD
ranlib in Linux.  It was happily reading symbol table entries from an
object file in an archive when suddenly fread starts returning data from
a totally different section of the file.  Even ftell lied and said it
was still reading from the old section.  fflush fixed the problem.  I
suppose that this may be what happens when you mix low level calls
dealing with file descriptors with high level C standard library calls
as is being done in ranlib.

Anyways, it seems that I can finally cross compile stuff reliably.  I
built the jed editor and installed it on my VS2000.  It's slower than on
a 386DX20 with an ST-251 MFM drive, but it does work.  I'm pretty sure
that the MFM driver for the VS2000 is the bottleneck.  NFS actually
seems faster.

Kernel compilation seemed to be going fine too (and much faster on a
P133).  However, it failed on "arch/vax/vax/clock.c".  Many of the
definitions in chip_clkread and chip_clkwrite (eg. YR_OFF, MON_OFF,
DAY_OFF, etc.) are undefined.  I can't find them, and the VS2000 seems
to agree, so this looks like a geuine problem with the NetBSD-current
kernel sources.

-- 
|  Boris Gjenero <bgjenero@undergrad.math.uwaterloo.ca>              |
|  Home page:  http://www.undergrad.math.uwaterloo.ca/~bgjenero/     |
|  "Luke, you're going to find that many of the truths we cling to   |
|  depend greatly on our own point of view." - Obi-Wan Kenobi, ROTJ  |