[39] in athena10
ATHENA_SESSION_TMPDIR
daemon@ATHENA.MIT.EDU (Tim Abbott)
Sat Jan 12 01:07:23 2008
Date: Sat, 12 Jan 2008 01:07:11 -0500 (EST)
From: Tim Abbott <tabbott@MIT.EDU>
To: athena10@mit.edu
Message-ID: <Pine.LNX.4.64L.0801120050590.26368@mega-man.mit.edu>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed
I'm wondering what the plan for ATHENA_SESSION_TMPDIR is in the future.
Currently, ATHENA_SESSION_TMPDIR is created by some code in the system
dotfiles:
# Create a unique temporary directory for the session, and remember
# the directory name in the environment.
sessiondir="`$initdir/mksessiondir -s $$`"
if [ -n "$sessiondir" ]; then
ATHENA_SESSION_TMPDIR="$sessiondir"
export ATHENA_SESSION_TMPDIR
fi
One of the annoying things about this system is that it leaks a file in
/tmp every time someone logs in. Its other problem is that
ATHENA_SESSION_TMPDIR is not guaranteed to be set if the user is not using
the system dotfiles for whatever reason (the common case for this on
Linerva is people logging in with GSSAPIDelegateCredentials=no).
Currently, the only thing Debathena has used ATHENA_SESSION_TMPDIR for is
working around the fact that the elinks package (the version of links
installed on Linerva) creates a named pipe in $HOME, and the package which
configures this is only installed on Linerva and mega-man at the moment
(though it might be deployed more widely in the future if we can't get
elinks fixed upstream).
I recently generalized pam_xauthority.so to pam_mktemp.so (currently
available in /mit/debathena/packages/debathena/), which makes it
convenient to set an environment variable based on the results of
mkstemp/mkdtemp at login, and remove the resulting file at logout.
So, one option would be to use this new PAM module for this purpose.
We're currently planning to modify Debathena to use this new PAM module to
handle ATHENA_SESSION_TMPDIR, but figure it's worth discussing the
long-term fate of ATHENA_SESSION_TMPDIR first (I don't know what its
original purpose was). In particular, I'm wondering whether there's a
reason ATHENA_SESSION_TMPDIR shouldn't just go away (we can probably get
the elinks bug fixed upstream).
-Tim Abbott
