http://qa.mandrakesoft.com/show_bug.cgi?id=6140





------- Additional Comments From [EMAIL PROTECTED]  2003-15-10 17:04 -------
Hmm, there is something I still don't understand : killing X using
Ctrl-Alt-backspace should not affect gconfd which is not using X at all..
Moreover, if you are trying to login from the SAME computer at killing X, it
should work (well, it works here, when I tried).. But if you are trying to log
from ANOTHER system using NFS home, it will fail.

I've fixed this particular problem, which is not a NFS lock problem (well, it is
needed to workaround broken NFS lock system) :
in GNOME 2.4, GConf is using "local" lock which are stored in a temporary
directory, which is, by default, /tmp.. Unfortunately, Mdk Linux is setting
TMPDIR to $HOME/tmp, so the local lock is NOT local, when $HOME is NFS mounted..

So, if you are trying to run on a single workstation at a time, just grab test
packages from http://people.mandrakesoft.com/~fcrozat/gconf/ , they should fix
the problem I was explaining just above.

If you are trying to use log twice using the same NFS account on two different
workstations, you'll have to :
-allow ORBit2 to accept incoming TCP/IP connection by adding "ORBIIOPIPv4=1" to
/etc/orbitrc
-tel GConf to use global lock (this time, shared over NFS), by setting
GCONF_GLOBAL_LOCKS=1 (in /etc/profile.d/gconf.sh, for instance)..

So, could you try test packages I've uploaded ?


-- 
Configure bugmail: http://qa.mandrakesoft.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.


------- Reminder: -------
assigned_to: [EMAIL PROTECTED]
status: NEW
creation_date: 
description: 
Environment :

/home NFS mount ( vers=3, tcp, rsize=8192, wsize=8192, nosuid, soft, intr,
timeo=600 )

serveur = mdk 9.1 + soft RAID1 ( ext3 partition ), PIV

client = current cooker ( 2.4.22-10mdk, GConf-1.0.9-11mdk, GConf2-2.4.0.1-1mdk,
nfs-utils-clients-1.0.5-1mdk, kdebase-3.1.92-2mdk, metacity-2.6.1-1mdk ), AMD

Symptoms :
if for example I quit abruptly gnome - CTRL+ALT+BACKSPACE ), next time I want to
log, I can't because gnome can't remove the lock ( or find a lock ) in
~/tmp/gconf-username/ ( this location change between mdk 9.1 and mdk 9.2, before
this was ~/.gconfd/... ). 
Sometimes you can read the message saying that you have a locking problem ( and
thus read the path to the lock ) but sometimes you see nothing ( just some
windows with dash instead of text ... ).
A bad combinaisaison is when you log in and have the gnome preference daemon bug
( see #6138 ), then you decide to stop everything before it mess up your session
config ( as it asks to remove some applets ... ) by hitting CTRL+ALT+BACKSPACE.
You reboot the computer to resolve the problem ( argh )and when you try to log
in again ... gconf locking problem ( so you need to remove manually the lock )
and then you can log in ( so 3 attempts and 1 reboot just to log in in Gnome ! ).
This is really painfull when you have several computers to administer under
gnome. the locking proble can be solve by erasing by hand the lock at each gnome
start ( i put this in /etc/gnome/gnomerc ), but this is just a hack. Gnome is
really too fragile in an NFS environment ...

Reply via email to