Blender Git Loki

Support reading Face Map assignments from .blend, corrections to previous commit

Rework workflow for connecting Face Maps and bones

Previously, a Face Map was 'connected to a bone' by giving it the same name as the Face Map, and assigning the armature which includes the bone to the object's Face Map using an Armature modifier. This was pretty unintuitive and tiresome to set up and it also brought some technical issues (there was no nice way to find a Face Map assigned to a bone from the bone - this means we couldn't use Face Maps together with bones in pose mode as current design approach would require).

So finally I came up with a new approach:
The bone context in the Properties Editor got a new panel "Face Maps". In this panel there is a search button to choose an object. After an object is chosen, a Face Map from this object can be chosen using a second search button. That means a user can assign a Face Map to a bone directly, without having to worry about the armature.

This approach currently has the limitation that only one Face Map can be assigned to a bone (a Face Map can be used by multiple bones though). This means we can later use one Face Map to manipulate multiple bones, but we can't use multiple Face Maps to manipulate one bone.
Anyway, this could be changed easily later, but I'd rather avoid this since it complicates the UI/workflow again.

----

Old code is kept.
Also includes minor cleanup.

Merge branch 'master' into id-remap

Disabled some unused function args.

Merge branch 'master' into object_nodes

Add support for texture node trees in new dependency graph.

Cleanup: get rid of bmain in foreach ID callback.

messing with depsgraph here is not a good idea I think... and makes things simpler!

Merge branch 'master' into id-remap

Split align group code into its own file.

interface.c is becomming a bit too fat...

Fix comparison between buttons (from review).

Merge branch 'master' into ui-align-rework

Attempt to fix the 'user_one' mess...

Seems to work (though will need much more tests), but makes code more ugly. :|

Added clouds noise pattern node.

attempt to fix customdata related memoryleak... it has been reduced, but isnt completely gone yet

Merge branch 'master' into id-remap

Use the material blending function for the RGB mix node for the time being.

Merge branch 'master' into ui-align-rework

Update from reviews, mainly limiting comparison between buttons for big groups.

Current code now sorts buttons in aligngroups from top to bottom, which allows
us some early-out of sub-loop, limiting the O(n2) effect with huge groups of buttons.

Also, bunch of cleanup.

BGE: RASOffScreen object reimplemented with Renderbuffers.

RASOffScreen was previously using GPUOffScreen object from the gpu library,
which in turn uses textures as render destinations, also when multisample is
requested. This is a problem for Intel integrated as they generally don't
support multisample textures.

ImageRender doesn't need textures as render destination: Renderbuffers are
sufficient because it only read pixels from the render with glReadPixels and
glCopyTexSubImage, both of which work with Renderbuffers.

Therefore, RASOffScreen is reimplemented with Renderbuffers and multisample
offscreen renders are now supported on Intel GMA. The gpu library is no
longer used. There is no API change.

Added the RGB mix node (unfinished).