[comp.windows.x] Actual theories on GPX's R3 server crashes

allynh@marvax.berkeley.edu (Allyn Hardyck) (02/23/89)

Rather than saying "It crashes, what can I do?" I have attempted to
come up with some possible reasons.  I would greatly appreciate it if
someone would say if any of these are likely or if the problem lies
elsewhere.  Constants: a GPX running Ultrix 2.3, R3, the crashes (hangs,
more accurately, as there is no vmcore dump) always occurring on logout.

One hang occurred, with no one else on the machine, but after an xphoon
that had been running had been killed.  Is it possible that some monochrome
routines put the machine in a state that makes xterms fail to start up?
(It has baffled me for a while why monochrome images should take so much
longer to start up than color images.)

Another hang took place in the midst of attempting to kill the server
and all associated xdm's.  Apparently one xdm restarted itself in time,
but hung instead of printing the login window.  Is there some sane way
of stopping X dead deliberately?  It seems that the X server will start
up any necessary xdm's it needs, and vice versa, even occasionally when
sent signal 9.  So perhaps there is a problem with the server's or xdm's
handling of signals?

I've also thought that the server may assume total priority on startup.
There are several other terminals attached to the system, running
compiles virtually all day.  If, at the end of one X session, a compile
commandeers CPU time, is there some part of X that breaks?

Any replies, flames would probably be helpful, as we're dying at least
once a day here, necessitating total restart.  Thanks.

Allyn Hardyck
allynh@marvax.berkeley.edu