I have "tcp/some.host:7100" as the only element in my FontPath. When the fontserver on "some.host" is restarted, X server starts behaving very badly - CPU utilization goes to 100%, it would not allow any new application to start (application would just freeze), etc. If I change the FontPath (using xset fp ...) to something that does not include tcp/some.host:7100 and back to "tcp/some.host:7100", everything comes back to normal. This was true for RH 5.0 and 5.1 and is still true for XFree86*3.3.3.1-1
Exactly how did you restart your font server? You should not use SIGHUP instead use SIGUSR1. See man page.
I restarted xfs by rebooting the computer it was running on.
make sure xfs is enabled. As root, do a 'chkconfig --add xfs'. This should solve your problems.
Scratch that, I thought you were using 5.9 Beta. You are still using 5.2. We do not enable/configure/run xfs by default in Red Hat Linux 5.2 -- how are you starting it? See the man page for configuration information, it is most likely you have misconfigured the font server. We cannot locally duplicate the problem.
I think you misunderstood me. Xfs is restarted fine when computer is rebooted, the problem is that X server still tries to use the old TCP connection and does not try to open a new one util I explicitly tell it to do so (using xset). P.S. I start xfs from init. In /etc/inittab I have: mojave|~>grep xfs /etc/inittab xf:345:respawn:/usr/bin/X11/xfs -config /etc/X11/fs/config -port 7100 xc:345:respawn:/usr/bin/X11/xfs -config /etc/X11/fs/config-cyrfirst -port 7200
I wanted to add that this problem is 100% reproducable on several different machines and that it is very annoying.
I need to get this straight: Are you experiencing errors when you already have X up and running and you "restart" the font server? i.e. kill it and restart it, or kill -HUP it? If this is the case, this is a known deficiency in the font server, but there is a work-around. If you make changes to the config file and you want the server to pick up on them, send it the SIGUSR1 signal, i.e. killall -USR1 xfs. I'm not sure what you mean when you say 'old tcp connection'. If you reboot the machine, all sockets get recycled. There shouldn't be any 'old tcp connections' lying around at all.
Yes, I am referring to that "known deficiency in the font server" and I am saying that your workaround is not enough for me. I am running X and xfs on _different_ machines. In fact, a single xfs serves several X servers on several different machines. When the xfs computer is rebooted, all those X servers go in a really bad state. My question is: if I can make X recover by hands (using xset), why can not X be programmed to detect that xfs connection went away and recover on its own?
This bug does not exist in RedHat 6.1 - when xfs is restarted, X now properly reconnects.
yay.