Mark Newton wrote:
 
> I sucked down the tarballs for -current yesterday and installed them last
> night.  My first step was to attempt to build a -current kernel, but
> that was an abysmal failure:  /sys/nfs/nfs_vnops.c failed to compile
> with the -1.1 compiler (complained about an internal compiler error about
> 3500-or-so lines into the file).  At that point I resigned myself to
> writing-off the machine for a couple of days as it rebuilt the entire
> system from scratch:  cd /usr/src; make build.

The internal compiler error is a bug in gcc packed with 1.1.  You may get
a newer gcc binary in ftp://ftp.netbsd.org/pub/NetBSD/arch/sun3

> I'd like to be able to login to the machine and give more precise
> error messages, line numbers, etc, but unfortunately I'm at work at the
> moment and attempting to connect to the box yields:
> 
> newton@psychotic> telnet squish.apana.org.au.
> Trying 203.14.159.12 ...
> Connected to squish.apana.org.au.
> Escape character is '^]'.
> Connection closed by foreign host.

Did you make the pseudo-terminals by running 'sh MAKEDEV all' in your /dev
directory?

> ... from which I surmise that exec() has stopped working again :-)  The
> system will just have to stay like that until I get home from work and
> reboot it (oh well:  At least I'll get some work today, instead of
> telnet'ing back home and hacking NetBSD when nobody's looking -- "Larval
> stage" is bad for the work ethic <grin>).

Be carefull when upgrading: keep your libraries compatible with your
kernel and main executables.

> Can anyone offer any pointers to info sources which deal with upgrading
> from 1.1 to -current?  Could this be related to the rumored flakiness
> of the ncr_si driver when si_options = 7 ?  Has anyone had more success
> than I have?  Should the build be carried out in single-user or multi-user
> mode?  Am I wasting my time?

si_options=7 works well.  Be sure to be using the latest bug-free gcc for
your builds.

---
	Pappires