Re: [Tracker] Refactored the src/plugins & (new) Evolution support

[Martyn Russell <martyn imendio com>]

Philip Van Hoof wrote:
> On Wed, 2009-02-11 at 17:55 +0000, Martyn Russell wrote
> You can also just use git on this (branch is called push):
>
> http://git.codethink.co.uk/?p=tracker;a=shortlog;h=push

OK thanks.

> > #2, Before we go on, can you explain ALL the libraries and what they do. 
> >  From the names I don't really have any concept of what they are 
> > supposed to be doing. As far as I can see, we have:
> >
> >   - libtracker_evolution_indexer.la (now renamed)
> >   - libtracker-evolution-daemon-module.la
> >   - libtracker-evolution-indexer-module.la
> >   - ...
>
> *-indexer-module.la is an 'indexer' module which runs in
> tracker-indexer. It will pump the data actually into the database.
>
> *-daemon-module.la is a 'deamon' module which runs in trackerd. It will
> receive the data from the external application (presumably over DBus)
> and it will 'pass it on' to the *-indexer-module.la.

The first thing that I think here is, why do we need both? Especially if 
we are going to be moving the crawling and queueing all in to the 
indexer anyway (at a later date). Is it just so we have a permanent 
entity (trackerd) always available to receive data? Is there some other 
reason?

> > #13, For the cases where we reference another object here, we need to 
> > make sure we reference the new object first. We do this to protect 
> > ourselves from unreferencing the same object we are about to reference, 
> > (in the unlikely event that we set the object to the same object we 
> > already have in priv->object):
> >
> > +void
> > +tracker_push_registrar_set_object (TrackerPushRegistrar *registrar,
> > +                                  GObject              *object)
> > +{
> > +       TrackerPushRegistrarPrivate *priv;
> > +
> > +       priv = TRACKER_PUSH_REGISTRAR_GET_PRIVATE (registrar);
> > +
> > +       if (priv->object)
> > +               g_object_unref (priv->object);
> > +
> > +       if (object)
> > +               priv->object = g_object_ref (object);
> > +       else
> > +               priv->object = NULL;
> > +}
> > +
>
> So, what is the problem here exactly?

Say priv->object = 0xdeadbeef. Say object (passed in as an argument to 
the function) is also 0xdeadbeef, both are the same object. The object 
has 1 reference. We then remove that reference (which finalizes it) and 
then we try to ref it again. It is better in these conditions to always 
ref the new object just in case we are silly enough somewhere to set the 
object to the same thing we already have set.

--
Regards,
Martyn