[4779] in testers
Re: Most strange panel behavior today.
daemon@ATHENA.MIT.EDU (Christopher D. Beland)
Wed Jun 13 18:06:34 2001
Message-Id: <200106132206.SAA47421@whack-a-mole.mit.edu>
To: "Christopher D. Beland" <beland@MIT.EDU>
cc: testers@MIT.EDU
In-reply-to: The events that comprise the history of the universe.
Date: Wed, 13 Jun 2001 18:06:29 -0400
From: "Christopher D. Beland" <beland@MIT.EDU>
Okay, I did a series of tests using the gu account on the various
machines in the test cluster running 9.0.6, logging on them in various
orders at the same time, trying to reproduce the behavior I was
getting yesterday. I succeeded. I also determined that Bill was
experiencing the same problem; it had nothing to do with telnetting in
and whatnot.
I then narrowed down the cause to two things. One, when you log in
under Linux you get an instant panel crash no matter what. (I believe
this may have been contaminating the logins on the other machines in
some way, which only experience crashes after you've logged in under
Linux.) Two, there's some sort of intermittent bug in panel-wrapper
that is tickled by panel crashing almost immediately after it starts.
After I upgraded the machines to 9.0.7, I ran the full test sequence
again, and panel never crashed on me. (Yay!)
However, I was still able to reproduce the panel-wrapper bug by
manually killing the panel process, and then killing the new one that
panel-wrapper respawns sometime around when it's drawn in the gray
background but not had a chance to draw in any of the launchers and
whatnot. (It doesn't happen every time; I was also doing this on a
Sparc 5, which is relatively slow; maybe that helps.)
-B.
===============================================================
Christopher Beland - http://web.mit.edu/beland/www/contact.html
MIT STS/Course 6 (EECS) - MIT Athena User Interface Project
===============================================================
