On Saturday 03 February 2001  9:44 am, Chris Cowdery wrote:
> > > A silly question - why do we have to manually edit rc.conf to
> > > say rc_configured=YES as until you do that it's a right pain
> > > to do anything.
> >
> > You should find that rc has changed alot though, so it's more a safety
> > thing so that people actually notice I suppose.
>
> Would be a lot easier if it was YES as default, as I made no other
> changes to it.

It follows what most unixes do which is drop you into some kind of config 
mode on first boot.

> > > I've got a kernel rolled with buggy ICSIDE driver removed, and
> > > all looks good.
> >
> > Is it buggy?  I've not got one, my rapide works pretty well though.
>
> Crashes to debugger just after probing Etherh. About 50% of the time.

Could you get a trace of it?  (You will need to select loading symbol table 
in the bootloader)

> > Where did you get it from?
>
> The ftp site.
> ftp://ftp.uk.netbsd.org/pub/NetBSD/NetBSD-1.5/arm32/installation/kernel/net
>bsd.RPCINST.tgz

hmm, I think that's the one I built.  There's a few magic things you can do 
to have it ignore a podule, eg add an option to your boot options of 
poduleX.disable (where X is which slot it's in)

> > > And why does it take so long to probe when the kernel boots.
> > > I'm sure it didn't way back on 1.3.2?
> >
> > What device is it getting stuck on?  I know that the internal wdc can
> > take a while if nothing is on it, it times out.
>
> I think that's it - (going from memory which is bad) - I think it's
> just put up pioc0, and then waits for _ages_. Then it gets a wdc, and
> then waits for _ages_ again. _ages_ is of the order of minutes.
>
> I was under the impression (obviously wrongly) that if I altered the
> kernel config file to remove all the *'s and ?'s, i.e. put in the
> dmesg output, it wouldn't probe for anything so it would boot a
> lot quicker.

Should do, mine doesn't probe for the internal interface (mainly because I 
prefer to cut out stuff I don't want from the kernel, it's just a waste of 
space :)

> Which makes me think that (for all operating systems really), there should
> be no need to probe the universe every boot time, they should remember
> what they found first time around, and instantly boot with that. There
> should be a special option to rescan hardware for changes, as after all
> you boot more often with exactly the same hardware than with different.

That's perhaps correct for a home/domestic user.  However we cater 
for ppl that want 24x7 on their machines and so only reboot on a hardware 
change.

Cheers,
Chris