[5654] in SIPB bug reports
Re: LaTeX
daemon@ATHENA.MIT.EDU (Albert Dvornik)
Mon Nov 20 17:37:30 1995
From: "Albert Dvornik" <bert@MIT.EDU>
To: Jered J Floyd <jered@MIT.EDU>
Cc: bug-sipb@MIT.EDU, bug-newtex@MIT.EDU
In-Reply-To: jered@mit.edu's message of 20 Nov 1995 (11:06:20 EST).
<199511201606.LAA30789@vorlon.mit.edu>
Date: Mon, 20 Nov 1995 17:36:45 EST
> Regarding xdvi, does that mean that nobody has used xdvi since it had
> last been installed? Or had they just not used it on documents with those
> fonts? Anyhow, thanks for fixing this.
As far as I can tell, xdvi shouldn't have been finding *any* fonts on
Linux machines with no /usr/athena/lib/tex.
> The TeX problem is more confusing. I compared the straces on two
> machines, and they are nearly identical except for:
[strace gunk, with one machine referencing "arch/linux/bin", deleted.]
> Both machines have sysnames of 'i386_linux1', and machtypes of
> 'linux'. However, the machtype on keesh is an actual binary that
> responds appropriatel y to the different options (I think -S is
> being used), which the machtype I have is:
> #!/bin/sh
> echo linux
> exit 0
Ah! This makes perfect sense now.
The sysname used by the stuff in newtex is overridden by the value of
the environment variable ATHENA_SYS, if it is set. (If it isn't,
"@sys" is used. To do otherwise, one would have to either compile the
sysname into the binary, or (shudder) teach TeX about AFS.) Since you
were using a machine where "machtype -S" was always returning "linux",
the value of ATHENA_SYS was being set wrong by the login procedure
and/or the "add" command; therefore, TeX was getting confused.
While this particular problem is IMO a Linux installation bug, I'm
getting a bit worried about whether the above scheme for determining
sysname is appropriate. It would be trivial to change to always using
@sys (no recompilng is involved!), so while I'm still inclined to
think the current scheme is preferable, I'd like to hear comments on
it from anyone who'd like to offer some.
--bert
