I made a mistake today that (I think) may have uncovered an
inconsistency in /etc/netstart.

(skip next two paragraphs to get to meat immediately)

I accidentally installed the etc.tar.gz snapshot I made over my news
server's etc directory.  No prob for most stuff; the password file is
replicated from somewhere else and all of my sendmail.cf changes are
stored on the SDMF.  The rest of it was a no brainer.  I figured
"Serendipity" and proceed to try and reboot after fixing up the *.conf
files.

One of the changes in the new system is I'm going to reimplement all the
firewall stuff I was using under Juniper, so I enabled 'ipf' in the
rc.conf.  I didn't bother creating an /etc/ipf.conf since I didn't
actually have a clue at that point what I needed to do.

I rebooted the system and got a 'missing ld.so' error right after the
domainname report.  This also aborted the netstart, so none of the
network wasn't getting started and the network drives weren't getting 
loaded.  As it turns out, the netstart is trying to use 'logger' which 
was failing because the ipf.conf file hadn't been created, but the 
'ld.so' stuff hadn't been done yet.  It was finding logger, so I know 
the drives were mounted, it's just that the shared library stuff 
wasn't 'ready'.

If we were to mark this up as a feature, I could certainly see that - if
you expect the system to have ip filtering running and are starting the
program, but don't have a filter rules applied, there should be an error
at least this bad.

-- 
Dave Burgess                   Network Engineer - Nebraska On-Ramp, Inc.
*bsd FAQ Maintainer / SysAdmin for the NetBSD system in my spare bedroom
"Just because something is stupid doesn't mean there isn't someone that 
doesn't want to do it...."