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

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

Source: tar_files, 6th April 1994, from sun-lamp.cs.berkeley.edu
-and- i386 binary snapshot, 3rd April 1994, from sun-lamp.cs.berkeley.edu

Base version of NetBSD: 0.9.

Upgrade from previous -current: yes, 27 March tar_files from
sun-lamp.cs.berkeley.edu

Machine specifics: 486DX2/66, 16MB RAM, 340 MB IDE (Conner), VGA card

Other software: None, as yet...

Tar files integrity: good.

Additional things to do during upgrade: see below

Any warnings during compilation: again, see below

Any problems during make: 

Observations:

Changes from 27 March tar_files:
(This is taken from the CHANGES file on sun-lamp - last update on 4 Apr 94)

        change uid_t and gid_t to 32-bit quantities, and off_t to a 64-bit
                quantity.  Lots of associated changes. (cgd)
        __initdb() should only warn about spwd.db errors once, fix from
                <nils@merlin.ims.uni-stuttgart.de> (deraadt)
        i386: attempt to get the cursor correct on MDA/Hercules cards. If
                this still doesn't work, use FAT_CURSOR in your kernel
                config file. (deraadt)
        repair host map alias in ypmatch (deraadt)
        various sparc fixes (deraadt)
        take disklib.c out of libkern, put it back in the kernel
                proper (deraadt)
        fix sh/mkinit.c to not generate garbage in init.c (deraadt)
        fix hang in rcmd.c if either socket was >= 32 (deraadt)
        reorganize mmap() flags, and squish MAP_FILE. (cgd)
        upgrade 'vi' to the latest version of nvi (1.11) by Keith Bostic,
                available from ftp.cs.berkeley.edu. (cgd)
        add an 'ac' command, by Simon J. Gerraty and myself. (cgd)
        add 'ttyflags' to set device-specific tty flags. (cgd)
        i386: add -r flag to boot block, to select RB_DFLTROOT. (cgd)
        lots of man page fixes from Mark Weaver. (cgd)
        fix tn3270's idea of where map3270 lives, and fix the map3270 defaults
                to work with vt200's and xterms.  from Mark Weaver. (cgd)

+ others in the April 6th source snapshot...


Notes:

Well, I don't know if any of you have noticed, but the off_t typedef has
changed from 32 bits to 64 bits. It seems to have gone quite smoothly,
with no disruption for anyone.

"NOT!" - W.Campbell

I decided to leave things until the heat had died down, but I
obviously didn't let it go quite cold enough.  I upgraded with Mark
Weaver's instructions, and that went fine - make and config built and
installed, include installed, new kernel config-ed and built, src/lib
rebuilt and installed, libgcc.a rebuilt and installed, ld rebuilt and
installed, ranlib rebuilt and installed and...  ranlib died on me,
with ld.so giving error 10.

I tried all kinds of incantations.  I couldn't get Alan Barrett's
ranlib shell script (both of them) to work, I just kept getting that
good old error 10 from ld.so, and so...

In the end, I have to admit to downloading the 3rd April binary
snapshot from sun-lamp, extracting usr/bin, usr/lib and /usr/libexec
into a temporary directory, renaming /usr/lib to /usr/lib-old,
/usr/bin to /usr/bin-old, and /usr/libexec to /usr/libexec-old, and
then renaming the binary snapshot directories to /usr/bin, /usr/lib
and /usr/libexec.  I'm not proud.  It was after that that I found my
backups of the statically-linked as, ld, ld.so etc.  Oh well.

That was enough to get me a working C compiler, ld, ranlib and ld.so.

I see that Mark Weaver's instructions didn't include ar - I don't
know if that had anything to do with it.

However, I'm now up and running enough to re-compile /usr/src/lib,
which I couldn't do beforehand.
[Stop press - I've just recompiled everything, and am a tremendously
happy bunny once again - agc.]

Other Things
============

1.  I see that Matthieu Herrb's compiled XFree86 2.1 for the new off_t
changes, they can be got from

        ftp.laas.fr:/pub/NetBSD/XFree86-2.1.

WARNING: the shared libraries version numbers are now 3.0.

2.  config(8) has changed again - so remember to change your kernel
config files, and re-make and install config before making your new
kernel.

3.  pccons has gone away - pcvt is now in the kernel. Some say that
the keyboard locking problem (TM) has now gone away for them. Charles
Hannum says that pcvt has some of the problems that the old pccons
had. Stay tuned.

4.  Thanks to Charles Hannum - does he ever sleep? - the i386 swtch()
has been turbo-charged.

5.  lofs is going away - you're advised to use symlinks.

6.  mfs has been affected by the off_t change - it should be fixed
in the binary snapshot (3rd April) and tar_files (6th April).

7.  sun-lamp would only let me download one file at a time (i.e.  ftpd
would drop the control connection after one file had been transferred
on an mget - aren't our firewalls wonderful, such quaint technology as
ftp - but the files weren't on iastate, so I had to use sun-lamp)

8. Mark Weaver gives some instructions for pcvt -

"Ignore the instructions in the pcvt docs to change the default
gettytab entry to use "np" instead of "ap".  I've found this screws up
other things (such as hard-wired terminals).  Instead, leave the
default alone, and add an entry for Pcvt:

Pcvt|Pcvt console:\
        :np:ap:sp#9600:

And then make your ttys file contain lines like this for each vty:

ttyv0   "/usr/libexec/getty Pcvt"       pcvt25h on secure

This also assumes that you've added the pcvt termcap entries from
isa/pcvt/Etc to src/share/termcap/termcap.src, and rebuild the termcap
database.

Note: this is all just a hack I use to make pcvt work for now; it's
not pretty.  I actually think pcvt should use the standard vt220
termcap entry if at all possible."

9.  I believe that the new version of Julian Elischer's SCSI drivers
are now in the source tree.  Certainly, I'm seeing interesting new
diagnostics when I boot.  Thanks once again to Charles Hannum.

10.  A final word of thanks from me to Chris for running around and
getting things sorted out - without the April 3rd binary snapshot, I'd
have been stumped.

That's enough fun for one week...

Alistair

[I'm planning to be in Sunnyvale/Santa Clara/San Jose area for most of
May.  If any of you bay area dudes fancy getting together to talk
about Desert Island VM Faults or "What facets of the kitchen sink need
to go into NetBSD", - not that I can do anything about it, but
opinions are free- please drop me a line.  Just bear in mind that I
really suffer from jet-lag when flying west.]

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



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