Blender Git Loki

Merge branch 'master' into mesh-transfer-data

Merge branch 'master' into mesh-transfer-data

Finalize transfer datalayout for vgroups.

As best as possible I think, given the specificities of vgroups (mix of object and mesh data).

Merge branch 'master' into mesh-transfer-data

Transfer operators: do not run several times on same mesh, nor on linked ones.

Fix for last merge.

Merge branch 'master' into mesh-transfer-data

Conflicts:
source/blender/blenkernel/BKE_object_deform.h
source/blender/blenkernel/intern/object_deform.c
source/blender/editors/include/ED_object.h
source/blender/editors/object/object_vgroup.c
source/blender/editors/space_view3d/view3d_buttons.c

Merge branch 'master' into mesh-transfer-data

Cleanup: more 'fromlayers/tolayers' -> 'layers_src/layers_dst' renames.

Add standalone part of 'data layout transfer' op.

'Transfer DataLayout' operator (WIP).

Functionnal, but not complete yet. Should be mostly usable though from modifier.

Notes:
* The logic of current 'transferdata' pre-process on data layout is reused here,
avoids duplicating code and ensure consistency in behavior, and does not add
*too* much complexity.
* Also fixed some issues and mistakes in that part that were affecting
TransferData feature itself too. Will need at least one more review pass,
though, this is complex logical code.
* VGroup: The 'delete' part is not yet implemented (option of plain operator,
to get an exact same data layout as source).
Need to first transfer some vgroup stuff from ED to BKE (esp. delete logic),
will do that in master as separate patch.
* Standalone part of the op is empty right now (but this should be mostly monkey coding).

Merge branch 'master' into mesh-transfer-data

Grmll, typo-space fix.

Project (raycast) in both directions by default.

As suggested by Campbell, this option makes more sense most of the time.
We can always add back an option to only cast along normals if needs arise.

thin arrow looks a little nicer

use modifier attribute name 'object' from python/rna

use const floats & define initial buffer size

rename vars

don't store index as float (can fail for large values & conversion isn't free)

avoid sqrt when nearest values are under the threshold