On Thu, 3 Sep 1998 RYAN_ORDWAY@Non-HP-Vancouver-om1.om.hp.com wrote:

>      I'm running a 1.3F system right now (circa May or June) on my Q630 
>      with like a month of uptime. I've compiled various kernels with 
>      -current sources acquired via a weekly sup, and the recent kernels 
>      seem to work ok, but for some reason when I try to run X I get a 
>      'Can't run X with no screens!' error, like I used to get way back 
>      circa 1.2G. Something change in the 630 intvid driver?

intvid was broken for a short while. They symptom was that 'dmesg' would
show no intvid at all, and therefore no grf devices. Even the old X
servers work fine now that Scott Reynolds committed the fix. This is an
excerpt from dmesg with a kernel built against August 30th's sup:

intvid0 at obio0 @ f9001000: Valkyrie video subsystem
intvid0: 800 x 600, monochrome
grf0 at intvid0
ite0 at grf0 (minimal console)

If you don't see 'grf0 at intvid', you should try a newer kernel. 

>         Is anyone out there running a fairly -current setup out there on a
> Quadra 630 (or ilk)? I'd like to confirm that the reason I can't get a 
> -current system that lasts for more than a minute in multi-user mode is 
> because I screwed up somewhere.
>      
>         What I did: clean install of kernel and userland from
> /pub/NetBSD/arch/mac68k/current-19980829/* & booter 1.11.3.
>      
>         Results: kernel boots to single user OK, seems to get configuration
> info correct. I set rc_configured=YES and reboot into multi-user. From 
> there on whatever kernel I use I get a panic within a minute or so. I can 
> provide details if someone doesn't point out an obvious glaring mistake 
> (such as: -current just doesn't always work, idiot!).

It's probably nothing you did.  There have been numerous changes to
-current in the last couple of weeks, and something may well be broken.
Best thing is to report the details to the current-users mailing list.
Include the exact message that accompanies the panic. The stack trace
may also help--you get that by typing "trace" or "t" at the "db>" prompt.
Most of the developers seem to prefer just the function names here.

I've had trouble building kernels for the last couple of days on both '630
and i386. I should know later tonight if the recent fix to "ld"
changes anything.