dc44+@andrew.cmu.edu (Donna Converse) (03/25/88)
If you have installed the Xqdss server on a micro-vax II with a GPX
display under Ultrix 2.2, please read and respond. You may reply directly;
also please send a copy to "converse@morgul.psc.edu". I will post a
summary of replies and send a bug report to mit in about a week.
I have installed X version 11 release 2, and fixes 1 through 11.
I ftp'd the sources directly from mit and the checksums matched. I followed
most defaults during build and installation, including the flags allowing
both TCP connections and Unix domain connections.
The man page for Xqdss says: "You need to make sure your kernel is
configured with DMA_BUFSIZ set to 10 in sys/data/qd_data.c, or things won't
work right." The value of DMA_BUFSIZ has been changed from (1024 * 3) to
(1024 * 10) and I am running with this new kernel.
My server, Xqdss, intermittently fails and appears to take down the
the entire machine when it goes. The failure occurs upon startup or after
the server has received a HUP (hangup) signal. The failure has occurred when
the server is started manually, when the server has been sent a HUP signal
manually, when the server is started by init through a ttys entry, when
the server receives a HUP signal after I logout, and when I have started
the server inside dbx. The failure rate is high; perhaps 30% to 50% of the
time that the server is started or receives HUP it will fail.
When the server receives a HUP due to a logout, the display flashes
entirely white, then entirely black, and then either the familiar server
grey stipple background and xterm window appear, or the server has failed.
When the server has failed, the screen is black, my machine does not respond
to ping across the net, and the machine itself does not indicate hardware
error. I have no other way to determine the state of the machine, as I do
not have any direct connect ttys, and after doing a restart and a boot,
savecore does not find a kernel core file to save. The server doesn't drop
core. There are no messages other than startup messages in /usr/adm/X0msgs.
I never see messages on the console screen just before failure. This
behavior is consistent whether I am running the server through init, manually,
or in dbx. I have compiled the entire server code with -O, then with no
flags, and finally with -g, but this doesn't change the problem.
Here is my ttys entry (split into two lines here):
ttyv0 "/usr/bin/X11/xterm -L -geometry 80x24+195+275 -display perlis:0" \
xterm on secure window="/usr/bin/X11/X perlis:0"
I have also invoked X with an argument of "0" alone, and an argument of ":0".
Both work, but the intermittent failure of the server still occurs. I don't
understand whether an argument of "perlis:0" versus "0" will have a different
effect. Would it force the server to use TCP connections in the first case
and Unix domain connections in the second, or does it make no difference?
The display always has a black, unused band about an inch wide on the
left hand side, and an unused band of about a third of an inch on the right
hand side. It is a similar problem to the qvss server, but I haven't yet
tried to correct it.
Has anyone found any unitialized pointers being used in the qdss
server? I would appreciate your help. I will be writing an X driver for
our CGM (Computer Graphics Metafile) translator.
Internet: converse@morgul.psc.edu
Bitnet: converse@cpwpsca
Snail: Donna Converse * Pittsburgh Supercomputing Center * Mellon Institute
4400 Fifth Avenue * Pittsburgh PA 15213 * 412-268-5126dc44+@andrew.cmu.edu (Donna Converse) (04/11/88)
A few weeks ago I made a post about the Xqdss server bombing under
Ulrix 2.2 on a micro-vax with a GPX and promised to summarize responses.
Three other people announced they had identical problems. I received two
useful responses from Robert Scheifler and Mary Larson, thanks go to both of
them. I haven't yet tried to rebuild my kernel, so I can't provide
confirmation...
------------------------- Robert Scheifler's reply ---------------------------
Date: Mon, 4 Apr 88 17:40 EDT
From: Robert Scheifler <RWS@zermatt.lcs.mit.edu>
I have heard that the vanilla drivers in 2.2 have bugs that cause
problems with user-mode servers (as opposed to the kernel-based Ultrix
X10 servers). You should be able to drop your 2.0 qd.o back in and
things will work, or else beat on your DEC rep to get you a later
driver (they do exist).
------------------------ portions of Mary Larson's reply --------------------
Date: Tue, 05 Apr 88 11:02:01 EDT
From: larson@decvax.dec.com
.... As one of the server implementors in Ultrix I
do know of a problem with the Ultrix2.2 driver and X11.
When the server recycles (ie. you kill the last client) the machine may hang.
The only way to get to the console is by powering down the machine.
If this is your problem,
go back to the Ultrix2.0 driver (/usr/sys/BINARY.vax/qd.o) Once you
put this qd.o in place, "cd /usr/sys/machinename" and "make loadvmunix",
reboot with the new kernel.
-----------------------------------------------------------------------------
Internet: converse@morgul.psc.edu
Bitnet: converse@cpwpsca
Snail: Donna Converse * Pittsburgh Supercomputing Center * Mellon Institute
4400 Fifth Avenue * Pittsburgh PA 15213 * 412-268-5126cetron@utah-cs.UUCP (Edward J Cetron) (07/30/88)
I have had the same sorts of problems as Donna. I have scoured the newsgroup and have isolated the following facts: 1. the Ultrix 2.2 qd.o is no good, but the 2.0 one will run under 2.2. 2. the Xqdss server runs just fine under 4.3bsd. 3. the Xqdss server runs just fine under Ultrix 2.4. Wonderful, I have Ultrix 2.2 only with no previous 2.0 qd.o..., I have access to 4.3 bsd (all my other unixen run 4.3) but no qdxxx.c/h/o files for 4.3bsd, and my sales rep hasn't yet heard of Ultrix 2.4..... So, once again, I would like to impose on the generousity of the net: can someone please send me any qdxxx.c/h/o files so that I can pick any of the above three solutions? BTW, my most serious problem occurs during string input from the keyboard - everything just hangs and sometimes just exits without a core dump or error message. Has this been seen before? thanks to all (in advance), -ed cetron@ced.utah.edu center for engineering design university of utah