Hi Timo,
 
> i started the second attempt (yes, on a clean dir and obj empty) with
> './build.sh -O /usr/obj -T /usr/tools release' (AFAIK this is an
> implicit -j 1?) and this runs up to now (the machine builds yp_stuff
> right now).

releng says macppc built overnight, so I need to cvs up soon.  I've been
playing with the 601 code and I don't want to tie up the box yet.  I'm
stuck on the mc* driver, which is part of why I thought I'd play with
this 7200 for a while. The code I've got will do 20,000 ICMP responses
without an error (or Tx interrupt), and will do about 2/3s of full
ethernet frames without errors, but I can't get it completely
error-free.  I'm going to start looking into seeing if somehow there is
bus arbitration issues or some other problem that keeps the dbdma chip
from accessing the memory. Even when all of the packets are written into
memory and I do an endless dbdma command cycle of Tx, the dbdma chip
gets jammed up.  Sometimes it'll get whacked on up to three commands in
a row; sometimes it'll Tx as many six or eight packets in a row
completely error-free.  There is a note floating around about a bit
setting in gc that will allow autonomous dbdma retries, so I will look
into this as well.

tim