Re: Removing global variables

[Alexandre Bique <bique alexandre gmail com>]

On Wed, Nov 26, 2014 at 3:31 PM, Emmanuele Bassi <ebassi gmail com> wrote:
> hi;
>
> On 26 November 2014 at 14:22, Alexandre Bique <bique alexandre gmail com> wrote:
> > On Wed, Nov 26, 2014 at 3:17 PM, Emmanuele Bassi <ebassi gmail com> wrote:
> > > you mean apart from the major breakage of API, the massive
> > > complication of the internals, and the portability issues?
> >
> > It is about adding a pointer to a context.
> >
> > GtkCtx *gtk_ctx = gtk_ctx_new();
> > GtkWindow *window = gtk_window_new();
> > gtk_widget_set_ctx(window, gtk_ctx);
> > gtk_main_loop_ctx(gtk_ctx);
> >
> > Would that work without breaking the actual interface?
> > Then yes the internals would not an update pass.
>
> it's really not "just that".
>
> it means locking down every single static variable in GTK; it means
> locking down every single GObject property, and signal emission. it
> basically means doing the same things GStreamer does — without
> breaking API, and without breaking existing code. GStreamer has been
> engineered in that way, GTK+ hasn't. ensuring that every single
> initialization happens in a separate thread local storage or
> thread-safe context would be a fair bit of engineering, and would also
> likely result in an API break — at least, to avoid existing code from
> subtly breaking.
>
> I suspect you'd also want the GObject type system to be loadable on
> demand, and unloadable as well, which is not really supported these
> days.
>
> I strongly suspect you want another toolkit for your use case.

I see.
Thanks to all of you for your answers and advice.
I wish you a good day :)

Cheers,
-- 
Alexandre Bique