[7706] in testers
Re: My Athena session will not run.
daemon@ATHENA.MIT.EDU (Greg Hudson)
Thu Sep 18 10:52:06 2008
From: Greg Hudson <ghudson@MIT.EDU>
To: William Cattey <wdc@mit.edu>
Cc: testers@mit.edu, athena10@mit.edu
In-Reply-To: <475C46DD-CE17-44B2-B66B-5AF7CBBE1604@mit.edu>
Content-Type: text/plain
Date: Thu, 18 Sep 2008 10:49:49 -0400
Message-Id: <1221749389.7130.84.camel@error-messages.mit.edu>
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
On Wed, 2008-09-17 at 16:55 -0400, William Cattey wrote:
> I didn't see any obvious source of bourne shell code.
It's a mystery.
I've created an instrumented version of the tcsh session script which
might help to track down this problem if it happens again. You can
install it with (as root):
cd /usr/lib/init
mv xsession.tcsh xsession.tcsh.aside
cp /afs/dev.mit.edu/user/ghudson/xsession.tcsh.debug xsession.tcsh
Perhaps the problem will crop up again the next time you reinstall an
image.
> One odd thing I noticed. Apparently some time in the distant past I
> brought one of the system init files into my home dir:
>
> I have a .cshrc file that is a 1989 vintage Athena .cshrc file.
It is normal to have a .cshrc file of some vintage in your homedir.
See /usr/prototype_user for a "template" of dotfiles which are expected
to exist.
> However, I think I'm on the trail of something interesting. I *DO*
> have a .cshrc.mine that does lots of stuff. I think it blew out,
> because I just did "alias" and none of the aliases I've asked for
> were set. That .cshrc.mine is also attached.
You got rid of your .cshrc, so none of the standard C shell
intitialization is happening in your terminals. You should
restore .cshrc from either /usr/prototype_user/.cshrc or cshrc.old.
> When I click on show details it says, "/mit/wdc/.xsession-errors
> could not be opened."
Yuck, this isn't so good for debugging dotfile problems. The dialog is
run by the greeter process, which never obtains tokens. We don't really
have any way of fixing the gdm functionality (it would require a big
architectural change to gdm, so submitting an upstream bug report would
be a real long shot). We could probably duplicate the functionality
ourselves using a session wrapper, but it's not low-hanging fruit since
it would require writing GUI C code.
As a practical matter for you personally, your homedir is readable by
system:authuser but not system:anyuser. If it were readable by
system:anyuser, you could read the errors file directly and not have to
check it from elsewhere. However, that also puts your homedir on the
web, so the current state may be preferrable to you.
