[2805] in testers
problems updating to 8.0A
daemon@ATHENA.MIT.EDU (Jonathon Weiss)
Wed May 15 08:45:05 1996
From: Jonathon Weiss <jweiss@MIT.EDU>
To: testers@MIT.EDU
Cc: matt@MIT.EDU
Date: Wed, 15 May 1996 08:44:49 EDT
or, how I lost in one easy step...
I think the most important problems are 10, 12, 13, and maybe 4, 6.
1) Why was /var/cron/log summarily removed? Is the equivalent info being
logged somewhere else now? not at all? (Ther version specific script
for 8.0A specifically removes this logfile.)
2) My machine has a /sun4 link pointing to
/afs/athena.mit.edu/system/patch.76/sun4. given some of the other
things that were removed in the version specific update you might want
to get this one too.
3) /etc/shells was made a local config file, rather than a link to the
packs. mkserv remote needs to be updated to not try to bring this
local (It currently prints an error, but doesn't bash the file.) I'll
put this on my list of things to do when I start dealing with mkserv
for 8.0.
4) named was not successfully restarted before mkserv was run. In
playing with this by hand, I was unable to start either the 7.7 or the
8.0 named successfully, before rebooting. Furthermore, I edited
resolv.conf and that lost too (ie, I pointed it at a reasonable
machine and was still unable to get queries answered). This will
cause some machines to fail to update properly. They will end up
without any mkserv services configured. I believe that machines with
the following setup will lose (I think but cannot guarantee that all
other configurations will work): any machine that has a locker:path
line in the .services file (this is caused by using this syntax
designed for testing mkserv scripts. Note that when you remove the
test service, you do not necessarily remove the locker:path line and
will still lose). I was able to work around this problem by changing
all of the relevant lines to :full-afs-path. Since there is no locker
name that must be resolved in order to be attached, I didn't lose.
(Matt, make sure you do this for mkserv paranoid.)
Other things that require DNS during the mkserv process will also
fail. I don't know of anything in the default services that falls in
this category, but 1) I may be forgetting something and 2) I have
things in this category in my .private script.
5) a number of files and/or dirs were chowned to 1047 (builder)
6) during the update the following was reported. I don't know what it
means, but these look like errors: :-)
/srvd/usr/athena/lib/update/do_update: /srvd/install/aux.devs/D10audio: not found
/srvd/usr/athena/lib/update/do_update: /srvd/install/aux.devs/D20appletalk: not found
/srvd/usr/athena/lib/update/do_update: /srvd/install/aux.devs/D10audio: not found
/srvd/usr/athena/lib/update/do_update: /srvd/install/aux.devs/D20appletalk: not found
Copying new system files...Copying new bootstraps...Installing new bootblocks
done.
Updating version...done
7) fs (at least with certain arguments) will print errors of the form
"socket: No such file or directory", but it seems to DTRT anyway. Not
a big deal.
8) /usr/athena/shar/emacs/19.30/lisp/hilit19.el has an older date than
the one on the 7.7 beta packs, therefore it wasn't synctree'd local
like my config told it to. Is there aversion skew problem here?
9) Many links in /usr/athena were removed (and there was much
rejoicing). there are three remaining emacs and zmailnotify point to
themselves in bin for reasons I more or less understand. session_gate
points to itself in /usr/athena/bi (not bin) is this intentional?
10) Many things I had synctreed local (via .private.sync and or the
.sync files for various mkserv services) were not sync'd local. The
reason for this appears to be that these files are no longer in /srvd
since they have been moved to /os. I haven't checked throughly, but I
beleive this problem affects a significant number of the mkserv
services available for Solaris. I do not know how we want to deal
wiht this. Perhaps I'll come up with a good idea when I've had some
more sleep, but I'm not counting on it. If anyone has suggestions,
I'm interested in hearing them.
11) /usr/athena/bin/kermit as a link to the modem locker seems to have
diasppeared.
12) I can't login in console mode. I hit ^P and get a console login
prompt. when I try to login I get a WArning about ttymon being unable
to allocate a controlling tty. If I tried to log in as root it prints
the passwd prompt, but immediatly fails and prints a login prompt
(without letting me enter a passwd). If I continue to try to login,
everything is the same, except that I don't get the warning until I
restart and re-kill X. If I try to log in as me the behavior is very
similar, except that it after skipping the passwd prompt it gives me
an "Enter password:" prompt, which, of course, doesn't seem to
recognize my password.
13) config_afs no longer understands the syntax: -cell.mit.edu to
disallow setuid from the cell 'cell.mit.edu'. It used to be the case
that I could put '-cell.mit.edu' in /usr/vice/etc/SuidCells/local.
Jonathon
