[8175] in Hotline Meeting
atalanta
daemon@ATHENA.MIT.EDU (Joe Harrington)
Mon Mar 16 10:09:22 1992
Date: Mon, 16 Mar 92 05:01:30 -1000
From: jh@oobleck.IFA.HAWAII.EDU (Joe Harrington)
To: hotline@MIT.EDU
Cc: jh@galileo.ifa.hawaii.edu, wisdom@poincare.mit.edu
Reply-To: jh@galileo.ifa.hawaii.edu
(For readers who don't know me and wonder why someone in Hawaii is
sending to hotline, I'm at Mauna Kea Observatory getting data for my
thesis and remotely system managing for the astronomy lab, which has
an Athena cluster and which uses the afs-nfs translator to get
/usr/athena, /usr/sipb, and several other AFS filesystems.)
Since Friday, we've experienced several problems (see below) with the
afs-nfs translator. Certain programs, like latex and rn, seem to have
difficulty reading configuration or macro files. The machine logs nfs
timeout errors. However, a user can read the file just fine. The
problems seem to be fairly regular, but one set of tests shows about a
10% success rate. I understand that supporting atalanta is secondary
to your supporting the clusters, but we'd appreciate greatly its
return to ordinary operation. We haven't changed anything on our
system, and Matt Power's tests seem to indicate that the problems are
not restricted to us (see below). Since the problem was originally
seen in the SIPB version of latex, I reported it first to them. That
message is below (and includes error output from the Athena version as
well), along with Matt Power's response.
--jh--
Date: Sun, 15 Mar 92 06:27:56 -1000
From: jh@oobleck.ifa.hawaii.edu (Joe Harrington)
To: bug-sipb@mit.edu
Subject: The hundred thousand dollar question: What made latex die suddenly?
Cc: jh@galileo.ifa.hawaii.edu, wisdom@poincare.mit.edu, amehta@mit.edu
Reply-To: jh@galileo.ifa.hawaii.edu
As of sometime Friday evening or Saturday morning, latex (both sipb
biglatex and athena latex) stopped working on our machine
(astron.mit.edu). It's a decstation running ultrix 4.2 that mounts
the afs-nfs translator and gets /usr/sipb and /usr/athena through it.
Both programs start running, but the version 2.93 biglatex gives:
/usr/sipb/decmipsbin/biglatex test
This is TeX, C Version 2.93 (no format preloaded)
(Fatal format file error; I'm stymied)
(Fatal format file error; I'm stymied)
...and produces no output files at all.
Version 3.00 latex produces:
This is TeX, C Version 3.0 (format=lplain 91.7.14) 15 MAR 1992 10:00
**&latex test
(test.tex
LaTeX Version 2.09 <4 Aug 1988>
Prepared for MIT Project Athena. (/usr/athena/lib/tex/macros/article.sty)
! Undefined control sequence.
\@floatplacement ...global \@toproom \topfraction
\@colht \global \@botnum \...
...
l.3 \begin{document}
? q
OK, entering \batchmode...
! Undefined control sequence.
\@floatplacement ...bal \@botroom \bottomfraction
\@colht \global \@colnum \...
...
l.3 \begin{document}
[this repeats for a LONG time, as each tiny thing is discovered not to
be defined]
The file test.tex is:
\documentstyle{article}
\begin{document}
Hello.
\end{document}
In my tests, TEXINPUTS was unset. I also tried it with my usual
.:/mit/jh/lib/tex:/usr/sipb/lib/tex/macros
and got the same results. No other TEX environment variables were
set.
Since SIPB shut off remote access to hodge and podge, I can't rlogin
to any machine that gets the decmips AFS packs to check if the same
binaries and such work on an Athena machine, but it works on our
Athena 7.3 vaxen (where, interestingly, there is a /usr/bin/X11/*tex).
I could not find any @sys entries in the paths to the programs or
files, and could read every byte of virtex, plain.tex, latex.tex, and
article.sty. The links we use are:
lrwxr-xr-x 1 root 54 Feb 1 03:22 /usr/athena -> /afs/athena.mit.edu/system/pmax_ul4/srvd.73/usr/athena
lrwxr-xr-x 1 root 9 Nov 23 01:31 /usr/sipb -> /mit/sipb
lrwxr-xr-x 1 jh 30 Nov 23 06:39 /mit/sipb -> /afs/sipb.mit.edu/project/sipb
Could someone run latex on hodge, podge, or some other Athena decmips,
and let me know if it succeeds? Could a TeXnician or TeXpert decode
the error message to see just what file it's barfing on? Does anyone
have any idea what changed between Friday and Saturday that could
cause this?
Of course, this was discovered by a professor writing a grant
application due Monday. We have changed nothing on our machine. I
know that doesn't mean the problem doesn't lie there. Any help will
be very much appreciated.
--jh--
From: mhpower@Athena.MIT.EDU
To: jh@galileo.ifa.hawaii.edu
Cc: bug-sipb@Athena.MIT.EDU
Subject: Re: The hundred thousand dollar question: What made latex die suddenly?
In-Reply-To: [2593] in SIPB bug reports
Date: Sun, 15 Mar 92 17:25:09 EST
>As of sometime Friday evening or Saturday morning, latex (both sipb
>biglatex and athena latex) stopped working on our machine
...
>LaTeX Version 2.09 <4 Aug 1988>
>Prepared for MIT Project Athena. (/usr/athena/lib/tex/macros/article.sty)
>! Undefined control sequence.
I was able to repeat this problem, although I don't understand it
completely.
I tried running latex on test.tex several times, using different
DECstation configurations. What I found was:
1) Running latex from a machine that accessed latex directly via its
local AFS software worked consistently.
2) Running latex from a machine that accessed latex via an NFS mount
of atalanta:/afs worked successfully about 10% of the time,
otherwise it failed with one of a variety of "Undefined control
sequence" errors, such as the one quoted above.
3) The behavior in (2) was also found when using an NFS mount of
charon:/afs.
4) Whenever latex failed, ULTRIX logged two errors, which showed up
in uerf like:
EVENT CLASS OPERATIONAL EVENT
OS EVENT TYPE 250. ASCII MSG
...
MESSAGE NFS read failed for server atalanta:
_Timed out
(or "server charon", obviously).
My guess is that the problem isn't on astron, and isn't on the afs-nfs
translator (I believe you're using atalanta). Perhaps there's some
type of afs problem (?) or some type of network problem. All of the
fileservers for decmips versions of latex (athena or sipb) are in W20,
so problems in network traffic to the 18.70 subnet might be relevant.
Matt
