Re: [evolution-patches] Session manage evolution-wombat

[Rodrigo Moya <rodrigo ximian com>]

On Fri, 2003-10-24 at 09:51, Frederic Crozat wrote:
> Le ven 24/10/2003 à 00:22, Rodney Dawes a écrit :
> > I don't like this patch. It adds the need of a display for wombat. The
> > wombat is supposed to exit when all clients are disconnected from it,
> > which it seems to do here just fine. The alarm-notify daemon uses the
> > display and connects to the session manager so should be getting killed
> > on log-out and started back up on log-in. There may be a bug in the
> > version of bonobo that the evolution in redhat/fedora is running
> > against, or something similar, such that wombat isn't getting the idea
> > that all the clients are gone. Using the session manager to kill off
> > servers that don't know the clients are gone though, doesn't feel like
> > the correct workaround.
> > 
> > Changing evolution-alarm-notify to use bonobo_main_quit () though, seems
> > good to me.
> 
> Well, I had a discussion about that with rodrigo on IRC and it seems in
> Evo 1.4 branch, wombat has no way to know that it is no longer needed
> but it has been fixed (or will be fixed) in Evo HEAD.
> 
wombat does have a way to know when clients are gone if the clean up
properly. What I told you was that if the client crashes or is
terminated by some other mean, without cleaning up, wombat thinks it is
still alive.

AFAICS, if all goes well, wombat cleans up properly, at least here. It
is never left running, although I've seen recently, while trying to
reproduce this bug, that bonobo-activation-server and gconfd-2 are left
running, so it might be some reference that is not cleaned up, maybe in
evolution or maybe in some other program.

cheers