Re: [gdm-list] gdm > 2.20 and SunRay

[Meik Hellmund <Meik Hellmund math uni-leipzig de>]

On Wed, 05 Jan 2011 09:43:21 -0600
Brian Cameron <brian cameron oracle com> wrote:

> 
> Meik:
> 
> > This bug&patch is about the $HOME/.gconf directory. After updating
> > my gconfd from 2.28 to 2.32 this works for me and I have now a
> > per-session .gconf directory
> 
> This should make GConf behave better with GDM.
> 
> > $HOME/$ENV_GDM_SEAT_ID/.gconf
> > That's good.
> > But there is still only one $HOME/.gconfd directory and only one
> > $HOME/.gconfd/saved_state cache file.
> 
> I assume you mean /var/lib/gdm/.gconfd/saved_state cache file. 

Yes, I mean  the home of the "gdm" user.


> It
> isn't a problem for the /var/lib/gdm/.gconfd directory and other files
> to exist, but the cache may be an issue if it grows too big. 
> I'd delete the file, and see if it grows too big again.  If you were
> experiencing problems before, it may have been related to the fact
> that GConf and GDM weren't integrated well.  Now that you've updated
> your system so it is saving separate .gconf directories for each
> display, you may find the issue with the cache goes away if you
> remove the version that got too big.  Does this help at all?

At the moment I can only test with 2 sunrays and then everything is
fine. A new test with 50 sunrays - where I saw the problem - is only
possible at the weekend.

But I would feel more relaxed if every gconfd had its own
saved_state file. 
Anyway, your bug report 
    https://bugzilla.gnome.org/show_bug.cgi?id=617017
already includes such a patch for gconfd from Ray Strode. 
It changes the file name "saved_state" to something like
"saved_state_machine_id-session_guid".

The problem with this patch is that these files are never deleted and
will accumulate over time. So perhaps putting .gconfd/saved_state into
the same directory as .gconf is a better idea?
I think  I will apply Ray's patch and let the gdm startup script remove
all the saved_state_* files. 

  Meik

-- 
Meik Hellmund
Mathematisches Institut, Uni Leipzig
e-mail: Meik Hellmund math uni-leipzig de
http://www.math.uni-leipzig.de/~hellmund