On Sun, 19 Sep 1999, Rob Windsor wrote:

> I've installed 1.4D snapshot in my Quadra 700 and have been building 
> packagesrc-ish things for a while.
> 
> I decided to get current, so I downloaded the latest, built a kernel, and 
> typed "make build" on the userland (before rebooting).

Not a good idea. Best to always make sure the kernel boots first.

> I had to re-build make(1) and config(1) as expected, and
> everything appeared to work just fine (other than it's SLOW!).

:-)
 
> When I boot, it doesn't make it to the line that mentions "a.out ..." (a line 
> or two above the "Reagents of Calif" copyright).

Wild guess: You're the first person to test the new PMAP wrappers on
mac68k. In any case, file a PR. In general, when you have problems
with current, you should file a PR, or post to currrent-users.

> Fortunately, I can boot back to the 1.4D GENERIC kernel.

Fortunately indeed. Worst case, "init" and several daemons use
syscalls that weren't in 1.4D, and so dump core immediately.
 
> I guess I am asking if this is a known problem, or if I am missing something 
> important in my kernel config file.

It looks OK to me. You don't need FPU_EMULATE for the Q700, nor
RTC_OFFSET. The latter is for PC's. MacOS sets the real-time clock to
GMT (on my Mac's at least), as NetBSD prefers. KMEMSTATS is doc'd (in
options(4)) as incurring a performance hit; you might want to leave
that out, although I can't notice a difference either way. There's
nothing in there, though, that would account for the problem you
describe. Anytime you're not sure, you could always try to build and
boot a GENERIC kernel.

> options         NTP             # NTP phase/frequency locked loop

I'd be curious to know how that works, especially with the net/ntp4
package. On Quadra 840AV running 1.4.1, it never syncs. "ntpq -c pe"
shows the offset pegged at ~500ms most of the time, with or without
the kernel PLL. Allen Briggs made changes to the interrupt scheme in
current that could help with that.