[ucbvax seemed to bounce this yesterday, and as I haven't seen it
appear on the list yet, I'm sending this again.  I apologise if you
see this twice. - agc]

Snapshot report.
================

Me: agc@uts.amdahl.com (Alistair G. Crooks)

Source: tar_files, 29th January 1994, from agate.berkeley.edu

Base version of NetBSD: 0.9.

Upgrade from previous -current: yes, 24 Jan tar_files from
sun-lamp.cs.berkeley.edu

Machine specifics: 386DX/40, 8MB RAM, 340 MB IDE (Conner), VGA card

Other software: Mark Weaver's shared XFree86 2.0, from 1 December 1994.
(from ftp.iastate.edu I think).

Tar files integrity: good.

Additional things to do during upgrade: none

Any warnings during compilation: various, not noted

Any problems during make: 
1. gnu/usr.bin/gcc2 - produces thousands of 
        "RRS text relocation at 0x.....  for "function""
messages.  Fix is to ignore messages until you rebuild and install
/usr/src/lib again (at least that's what did it for me) - remember to
do a "make clean" first. (Dynamically-linking a program still produces
a runnable executable, so there's no real problem).

Observations:
Note 3 below means that you have to rebuild everything in the right
order this week (i.e.  include, share/mk, lib/libc, gnu/lib/libmalloc,
gnu/usr.bin/gas, gnu/usr.bin/ld, gnu/usr.bin/gcc2 - ignore RRS gripes
from here, lib - should stop any RRS gripes, gnu/lib, bin, libexec,
sbin, usr.bin, usr.sbin, share, games, gnu, sys)

Notes:

1.  New gdb version (4.11) added, with support for shared libraries. 
Thank you, Paul Kranenburg.

2.  Keyboard problem gone away?  Mike Long reports that his problems
have gone away, and his keyboard is a happier beast without all the
hammering it took during autoconfig.  Mike says that it's something
that changed between Jan 15 and Jan 24.

3.  ld has been changed, and so has crt0.c - you've got to make the
new ld.so before the new crt0.c (in lib/csu/i386). (If you follow
the usual instructions to upgrade, you'll be OK, bar a few gripes
when building gcc2, until you build and install /usr/src/lib after
a make clean)

4. There's a multicast panic fix from John Brezak.

5. Havard Eidnes reports that groff doesn't build. Apparently g++
can't hack the /usr/include/stdio.h prototypes with the __attribute__
keyword in the stdio header files. Happens here too. (I haven't had
the time to look into it.)

6.  James da Silva's crunch program looks really useful.  FreeBSD look
like they'll be using it for FreeBSD 1.1 install floppies in March. 
It basically makes a whole lot of different user programs into one
executable (doctoring symbol tables on the way), and actually runs the
correct executable depending on the value of *argv. cgd doesn't
consider it neat, but you might like to think about it for a fixit
floppy, if you don't have a tape to back up and aren't network-connected.
Available from: ftp.cs.umd.edu:pub/bsd/crunch-0.1.tar.gz
Quoting from the README:
"As an extreme example, I have created an 868K crunched "kitchen-sink"
binary containing the following programs in their entirety:

        cat chmod cp date dd df echo ed expr hostname kill ln ls mkdir mt
        mv pwd rcp rm rmdir sh sleep stty sync test [ badsect chown
        clri disklabel dump dmesg fdisk fsck halt ifconfig init mknod mount
        newfs ping reboot restore swapon umount rdump rrestore ftp rsh
        telnet rlogin elvis vi cpio gzip sed tar gunzip gzcat

Note carefully: vi, cpio, tar, gzip, ed, sed, dump/restore, some networking
utilities, and the disk management utilities, all in a binary small enough
to fit on a 1.2 MB root filesystem floppy (albeit with the kernel on its
own boot floppy).  A more reasonable subset can be made to fit easily with
a kernel for a decent one-disk fixit filesystem."

7. SCSI boards - what exactly is supported? Michael Van Loon's been
having trouble with a BusLogic 747 (someone suggested the drive may
be at fault, not the controller, but not confirmed as yet, and not
much comfort to Michael either, I should think).

8. resolv.conf flag for bind/yp/flags in gethostby* has been added.
i.e. host names and addresses can be resolved by resolver, and don't
have to be in /etc/hosts. Thank you, TDR.

9.  Charles Hannum emphasises again that if_ne, if_ec and if_we are
deprecated - if_ed is the one to use. Michael Graff has reported problems
with if_ed, which worked fine with an older -current, although he's
since added hardware to his setup. Indeed, the if_ne driver has now
gone away - bit me when I compiled using an older config file. (I did
a diff first, but didn't look closely enough at the output).

10.  I saw some things go past on the CVS logs that suggested sendmail
and libm had been updated, but I couldn't see anything in the -current
CHANGES file that related to them.

General verdict: (Perhaps) there's pain this week during the upgrade
(unless you like seeing lots of RRS gripes flash by on your screen),
but I'd say it's worth it, with a multi-cast panic fix, the new gdb,
and the keyboard problems gone.

And the usual afterthoughts:

1. Calamity in the Crooks household as Mandy, my wife, gets the 486DX2/66
   to run Windows - the machine was my vanilla 0.9 box. (All complaints to
   a.hockley@bra0125.wins.icl.co.uk)
2. I put the sam binaries (for shared libs), source and docs in
        agate.berkeley.edu:pub/incoming/sam-sep93.{bin,src,doc}.tar.gz
3. What are the current ftp sites for -current binaries?
        ftp.iastate.edu - up and running fine
        ftp.cursci.com - I haven't been able to ftp in for c. 1 week
   Please send me mail if there are others, and I'll add them to this list.
4. We still need some more up-to-date XFree86 2.0 binaries and libraries
   for -current. Can someone look into this as a matter of urgency please?
5. Did anyone get the Mach microkernel mk83 to compile under NetBSD?
   I tried, but was getting errors because the de Boor make in -current
   gets directories muddled during the "config" stage. I reported this
   to the address given in the package, but haven't heard anything back
   from it.
6. I've got the new version (1.20) of fvwm running on my box. Nice.
   (Linking the executable produced the usual RRS messages with the
   old shared lib X libraries - see Afterthought 4 above).
   Has anyone got rxvt 1.62 running?
7. Thanks for the positive feedback for these snapshot reports.

Alistair
--
Alistair G. Crooks (agc@uts.amdahl.com)                      +44 252 346377
Amdahl European HQ, Dogmersfield Park, Hartley Wintney, Hants RG27 8TE, UK.


------------------------------------------------------------------------------