Re: [Gegl-developer] On gegl commit 7dcbb7321357166d36a9bf379efde7ad87fc4c17

[Jon Nordby <jononor gmail com>]

CC'ing the gegl-developer list

On 28 April 2013 19:33, Sebastian Pipping <sping gentoo org> wrote:

Hello Jon,


I noticed this change of yours to Gegl:

  -Gegl_ GEGL_MAJOR_VERSION@_ GEGL_MINOR_VERSION@_gir_INCLUDES =
GObject-2.0 GLib-2.0 Babl-0.1
  +Gegl_ GEGL_MAJOR_VERSION@_ GEGL_MINOR_VERSION@_gir_INCLUDES =
GObject-2.0 GLib-2.0


https://git.gnome.org/browse/gegl/commit/gegl/Makefile.am?id=7dcbb7321357166d36a9bf379efde7ad87fc4c17

Since we have a hard time with Babl-0.1.gir downstream at Gentoo too,
that patch could allow me to turn --enable-introspection back on with
Gegl.  So I am wondering

 * what the implications of the change really are (above all),

 * what your motivation was,

 * if that commit is going to make it into the next release of gegl, and

 * what the timing plans on the next release are.

Would be cool if you had a minute for me.  Many thanks in advance!

The change means that API in GEGL that uses Babl types will not be usable in GI bindings. That means any API used for accesses pixel data. I am not sure if this has worked in any of the past static bindings, and arguably one should (at least for the most part) do pixel manipulations in GEGL ops. Having bindings for the graph API is in any case more useful than having no bindings at all.

The reason for the change is that GI binding toolchain does not for work for Babl at the moment. GI is meant mainly for glib-based code, and has certain conventions that code is to follow to be bind-able. It may be that changing the name "Babl" to "BablObject" or "BablFish" may make it workable, though this would likely be an API break in GEGL (or perhaps it can be worked around with a typedef).

Unless someone comes up with a plan and code changes to make Babl bindable, that commit will be in effect for next release.

I do not know when the next release will be. Perhaps pippin can answer this?

--
Jon Nordby - www.jonnor.com