Reinhard Katzmann wrote:

> endless "looping" problem. If you use startx instead of g/k/xdm, you
Just in case. Has anyone thought about limiting this loop? When it
happens on your machine, it may not be such a hassle. But when this
thing happens massively on several machines and not on your working
place, this turns to a headache. One big problem is that tools like ssh
can't login into such machines. This loop eats a lot of CPU time. The
only solution is to make a 500 meter sprint and reboot each machine by
hand.

> > Some chances of what migh be wrong:
> > "clone-self = on" on /etc/X11/fs/config - some time ago I had this
> > problem with one X. Don't know exactly why but xfs broke. I had to put
> > it off.
> 
> This certainly is not the problem. I have turned this option on all the
> times and never had any problem. I also set client-limit to 20.
> Oh yes for error logging check out:
> # how to log errors
> use-syslog = on
It may not be the problem. But, on one build, I noted that this option
made xfs segfault. Two compilations on different machines (one
Coppermine, other a dual-Celleron) showed that this option was broken.
In the meantime I should note that gcc's differed at build level and one
machine had a "pure" 7.1, while the other was a cooker hybrid. It was
not long ago so I noted it "just in case".

> > > Using "xfs -port 7100 -daemon" as seen on forum.mandrakesoft.com, xfs
> > > does start and I can also start X!
> 
> Strange, never had this Problem. 7100 should be the default port, if
> the fs settings are not changed.
Not quite correct. On the /etc/init.d/xfs script, Mandrake launches xfs
with a unix domain socket (xfs --port -1), not through a tcp port. This
use protects xfs from direct access from the net. However it creates
some troubles. If I am not mistaken (please correct me) this socket use
demands the existence of a FIFO. On the case of xfs, this FIFO is
located on /tmp/.font-unix. Any problems with permissions on /tmp (it
happens sometimes) and xfs doesn't load. 

> 
> Best regards,
> 
> Reinhard Katzmann
> --
> E-Mail: [EMAIL PROTECTED]

Ektanoor

Reply via email to