Re: Signal editor rework

On Thu, Aug 26, 2010 at 3:45 PM, Johannes Schmid <[email protected]> wrote:
> Hi Tristan!
>
>>   - Firstly I can see you went and implemented the GtkTreeIter->stamp a
>> bit more
>>     robustly than in GladeProject, additionally the code needs to assert
>> for
>>     any input iter: (iter->stamp == model->priv->stamp). I've already
>> done this for
>>     GladeProject and will commit that soon (it helped alot to debug...
>> unfortunately
>>     I could still not completely solve bug 627078 and have not
>> committed that yet,
>>     I'll commit the checking code now regardless).
>
> I filed bug https://bugzilla.gnome.org/show_bug.cgi?id=623879 for that
> some time ago to also fix it in GladeProject. But you might be right that
> we might want to check it with an assertion.
>
>>   - Can you explain why the model->priv->stamp must be incremented
>> when the overall
>>     model data changes ? (it could be that GladeProject needs this
>> treatment too... or
>>     that the model's stamp should not change... I'm not sure).
>
> IMHO the gtk+ documentation says that an iter is only valid as long as the
> model doesn't change. If you can assure that the iter is valid after a
> change you might not need it but when a node is deleted, the iter of that
> node definitly becomes invalid. I think it's good to catch those things
> with an assertion before running into a memory corruption.

Ok that makes perfect sense to me. Sorry I missed the stamp fix
for GladeProject, I'm going to go over your patch and make sure
I didnt miss anything (for instance I at least missed the incrementing
of the stamp when the project tree changes...).

>
>>   - The iter->stamp should be initialized to something random,
>> otherwise every model
>>     starts with stamp == 0 (possibly making the iter from the last
>> destroyed signal model
>>     appear "valid")
>
> I wasn't assuming that someone would use an iter on a wrong model but if
> we want to catch that case we might want to start which a random number. I
> think we should check what GtkTreeStore does.

I guess it is at least important that the initial stamp is not 0, since
0 is what we generally use to represent an invalid iter (when
the iter to return was not found for instance).

Actually I was just guessing at why treestore initializes its stamp
to a non-zero random number... I can vouch that treestore (or liststore that
I checked... not sure) does this but I'm not exactly sure why.

>
>>   - I would prefer we stick with firing Glade's generic object
>> selection dialog to select
>>     the project object for a signal user data - I dont like the idea
>> of navigating recursive menus
>>     or dropping down a combo box with a possible > 1000 objects to select.
>
> Yeah, I will change that. I just was easy to implement for now but the UI
> is satisfying at all.
>
> Just some additional note: The dnd code currently uses the old gdk_drawing
> funtions that have been removed from gtk+. I didn't have the time to
> change that as I didn't fully understood how to do that in cairo. But it
> shouldn't be a big deal.

Extra note... I went ahead and copied out some GTK+ internals:
   glade_utils_cairo_draw_rectangle(), glade_utils_cairo_draw_line()

... in the case they may be helpful to you... also feel free to add some
convenience drawing functions to glade-utils.[ch] if you think that makes
sense for what needs drawing.

Cheers,
      -Tristan
_______________________________________________
Glade-devel maillist  -  [email protected]
http://lists.ximian.com/mailman/listinfo/glade-devel