About Gtk3/Gtk2 in 4.12

[Natanael Copa]

On Fri, 25 May 2012 09:36:42 +0200
Nick Schermer <nick at xfce.org> wrote:

> Folks,
> 
> We've ported a couple of parts of Xfce to Gtk3 and IMO it is better
> to stick with Gtk2 for 4.12. Beside that fact that it is an awful
> amount of work, the resulting applications are visibly slower (large
> treeviews mostly), consume more resources and hardly any advantages
> (api technically).

Are the speed and resource issues reported upstream? (I think your
findings might explain why ubuntu have been close to useless with 1MB
ram latest releases)

Do you have specific testcases?

Did you test different gtk3 versions (i.e is the current 3.4 better
than the first ones? Is there a tendency that those things are getting
better or worse in gtk?

> So for me porting the core of Xfce in the state Gtk3 currently is, is 
> for me a no-go.

:-(

gnome-disk-utiltity, needed for GIO support in thunar is gtk3 only
nowdays. Maybe someone should write a xfce-disk-utility (gtk2) with the
minimal things needed to mount USB disks in thunar?

Are there many/any of the dependencies that are gtk3 only? Might be
nice to have a list of those some place.

> That said we can still prepare for this to make the switch easier in 
> the future.
> 
> - Depend on Gtk 2.24.
> - Use new API (compile with GSEAL) as much as possible and drop 
> deprecated API.
> - In a number of places we use Gtk widgets instead of our own
> versions.
> - Libraries (libxfce4ui and exo) can provide a gtk2 and gtk3 version
> of the lib.
> - Some core modules (xfce4-appfinder for example) can optionally 
> compile against gtk3. This allows us to track Gtk3 progress and
> possibly help finding issues.
> 
> This will give us more time to port parts to Gtk3, because with the 
> complexity it brings, it will consume a lot of time, leaving not much 
> room for adding new features. Not to mention the fact porting (and
> only porting) takes to fun out of the project, for me at least.

Yes. The "fun" factor should not be underestimated.

> It might even be possible to run gtk3 panel plugins in a gtk2 panel 
> (externally), so also for goodies it will give a better migration
> path.
> 
> We've already discussed this a bit on IRC/private, and Jannis and
> Peter agree with this path. Hopefully others do as well, because then
> we can assign the porting parts and work on the 4.12 roadmap.

I was really hoping for gtk3 and fixing (or helping upstream fix) the
issues in gtk itself.

That said, to my understanding the only "payment" you get for your work
on xfce is the "fun" part so I think you should do whater you prefer.

Thanks!

-nc