ggm@dstc.edu.au said:
> I replaced the PMAGD with a PMAGB-B which is detected as an sfb
> instance.
> the console still switches over to serial port.
> ach well. maybe tc alphas are best as headless servers anyway. 

I wouldn't give up that early. Almost the same code works on
pmax, so it can only be a matter of some small bugs.
We should first find out in which subsystem -keyboard or display-
the error is generated. Can you print out the return values
of zs_ioasic_lk201_cnattach() and tc_3000_500_fb_cnattach()
in dec_3000_500.c and find out which one causes the error?
In the current version, zs_ioasic_lk201_cnattach() should return
zero on success, and tc_3000_500_fb_cnattach() zero on error.
Then you can track down further along the call path.

For the keyboard, it is
zs_ioasic_lk201_cnattach() in alpha/tc/zs_ioasic.c
	zskbd_cnattach() in dev/dec/zskbd.c
The latter does always return 0, ie success.
Oh - there is one possible bug: alpha/tc/zs_ioasic.c doesn't
#include "zskbd.h" (at least not explicitely) before it tests for
"NZSKBD > 0". It should. Can you try this first?

The display call path is:
tc_3000_500_fb_cnattach() in alpha/tc/tc_3000_500.c
	tc_fb_cnattach() in alpha/tc/tcasic.c
		perhaps tc_checkslot() in dev/tc/tc.c


best regards
Matthias