From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.0.1) Gecko/20021003
Description of problem:
on remote machines with addresses in the address part of the display var,
metacity fails to load.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1.enable xdmcp gdm
2.Xnest :2 -query ***MACHINENAME***
Actual Results: gnome-session crashed. wm crashed
Expected Results: normal login
Logging in failsafe (where blip==machinename)
and running metacity by hand I see the error message:
_X11TransSocketINETConnect: not INET hostblip
Window manager error: Unable to open X display blip:2.0
(Then I have to ^C metacity as it loops)
I ran these to make sure that DNS, etc was functional:
$ host blip
blip.internal.datastacks.com has address 10.1.2.6
$ grep blip /etc/hosts
$ dig +search blip
blip.internal.datastacks.com. 300 IN A 10.1.2.6
When manually setting the display to "10.1.2.6:2.0", it runs Fine.
Looks like an Xlib problem to me - does xterm work? or xclock?
Please reopen on adding info.
xterm, mwm, gnome-panel, and nautilus all work.
More looking into it seems that this condition only happens if the name in
address *IS* in /etc/hosts.
In this case, "blip".
I really have no initial guess how metacity is different from other apps in this
area, but will have a look.
That error message occurs when gethostbyname() thinks the host is not a TCP/IP
host; can you attach your /etc/hosts file? I don't know how it would happen or
how it could be metacity-only.
I tried reproducing the problem with Xnest/xdmcp and it works for me,
tried both with and without the hostname in /etc/hosts. I am trying with
the latest beta though, not with 8.0.
Created attachment 90098 [details]
/etc/hosts file used in problem
Please reopen if still present in Fedora Core 2