Re: g-s-d and SM

[jacob berkman <jacob ximian com>]

On Thu, 2002-06-13 at 13:05, Jonathan Blandford wrote:
> jacob berkman <jacob ximian com> writes:
> 
> > as reported in bug #85030, if g-s-d gets lost from the session, and
> > activated by one of the capplets, it won't register with the session
> > manager, so won't be started on login next time.
> > 
> > at this point, the user can't even manually start g-s-d to add it to
> > their session, since there's already one running.
> > 
> > a couple of options:
> > 
> >         1.  add something to tell g-s-d the SM
> 
> Is this thought incomplete?

no, just poorly communicated.

"add something to tell the g-s-d where the SM is" maybe?

basically tell it what's in $SESSION_MANAGER.

> 
> >         2.  have gnome-session start g-s-d via bonobo-activation and
> >         remove it from the session
> 
> I might vote for this one.

yeah this seems more fool-proof, except that we don't get the
restart-on-exit feature of having it be in the SM.  and i'm not sure how
to fix this?  an n-second timer which tries to make sure it's active?

> >         3.  punt it, since this shouldn't be happening
> >         4.  something else?
> 
> Along these lines, we were going to write a sanity-checker to make sure
> that your environment (mostly GConf and ORBit related) was sane when
> starting GNOME.  Did that ever happen, Havoc?

gnome-login-check has already supposed to checking the sanity of the
orbit stuff (although reports from wipro indicate that it's not) and
last week i committed a patch from havoc to run gconf-sanity-check there
as well.

i'd almost want to have something run after login to check that
nautilus, the panel, etc. are running...

 - jacob