Blender Git Loki

API doc: add new pure-rsync py script to update API doc on Blender server.

The other one (sphinx_doc_gen.sh) needs an ssh account on the server to work - and I hate bash, too! :P

Fix T49220: Vertex paint doesn't work with OpenSubdiv

It's a bit tricky to align vertex color data between Blender and OpenSubdiv
so for now we simply disable OpenSubdiv in the paint modes.

Safe for 2.78.

Fix a few OpenCL compiler warnings.

Fix T49229: ID user decrement error when deleting group instance twice.

Another great example of inconsistency in usercount handling - dupli_group was considered
as refcounted by readfile.c code (and hence by library_query.c one, which is based on it),
but not by editor/BKE_object code, which never increased group's usercount when creating
an instance of it etc.

To be backported to 2.78.

fix: not initialised variable can crash blender (related to modifiers with cage editing enabled)

Fix T49228: Separate by material, materials dissappear after reload.

Yet another mismatch where code would decrease usercount (of Material here) but never increase it again
when re-assigning the datablock...

windows compile fix, use portable BLI_snprintf instead of snprintf

added global keep shards mode for cutter groups

fixes: activating clusters via percentage, add meshisland id as name in fracture too, load crash fix

2.78 release: Point locales to a proper updated branch

fix, enabled threading for autoexecute

added threaded checkbox next to execute fracture operator

made threading optional via a checkbox

fallback to singlethreaded for dynamic, mousebased and fast bisect.. also changed back fast bisect orderding to have more uniform shapes

first attempt to multithread fracturing, works only for prefractured for now

some more optimizations at bisect, bit faster and less memory consumption, and removing bakes now when refracturing

fixes for bisect and bisect+fill, should look correct again and be a bit faster as well

using a kdtree based preselection method to speed up bisecting

small fix for loading possibly different jack shared library versions on different linux distributions, thanks to sergey for the fix

small enhancement for voronoi + bisect and solidify after FM use case

if you use voronoi+bisect and solidify after FM, you can automerge now without having autohideable inner faces. This should have the same effect as autohide in regular use case has.

memory optimization for smoke compressibility fix, dont use recursion but a queue to avoid potential stack overflows, thanks to angavrilov for this optimization