On Sun, 11 Oct 1998, Todd Whitesel wrote:

> Today I brought over my freshly supped source tree (which works on i386) and
> tried to build it on my CATS system. Previously the CATS system was running
> the 19980805 CD distribution that came with the board. I used the -current
> CATS config file to do these builds. So far I have not gotten any new kernels
> running; other than make and config, I haven't rebuilt any of userland.
> 
> [ much build output deleted ]
> 
> rm -f netbsd
> ld -Ttext F0000000 -e start -x -o netbsd ${SYSTEM_OBJ} vers.o
> pckbc.o: Undefined symbol `_pckbc_machdep_cnattach' referenced from text segment
> *** Error code 1
> 
> Stop.
> 
> 
> I noticed that some other ports have this function in their machdep.c file,
> so I tried splicing one in based on what i386+alpha were doing. The resulting
> kernels linked, but were DOA (dead on arrival).
> 
> By this I mean the CATS firmware would load /netbsd fine but after it printed
> "Calling 0xF0000020" there would be absolutely nothing, except maybe a small
> IDE activity light flash a fraction of a second later. The CATS was dead in
> the water until reset.
> 
> I always keep /netbsd.old around because of exactly this sort of problem...
> 
> Another thing: I just noticed some messages "Failed page fault in kernel" --
> they show up during booting my 19980805 kernel -- what the heck are those?
Hi,
  Opps a couple of things here.

1. I'm missing the line

#define __HAVE_NWSCONS

from arm32/include/types.h
I'll commit that as soon as my ISP fixes their routing problems.
This should fix the undefined symbol.
Another change I forgot to fix was the default console device. The kernel
is defaulting to the diagnostic serial port for its output.
Again I will fix that ASAP.

Cheers,
				Mark