Blender Git Loki

Fix panorama render crash with split kernel, due to incorrect buffer pointer.

Also some refactoring to clarify variable usage scope.

Cycles: reorganize Performance panel layout, move viewport BVH type to debug.

Fix for recent fix in fc890cd, cstdint is c++11 only so don't use it yet.

Cleanup: rename manipulator API functions

- WM_manipulatorgrouptype_remove- > free
- WM_manipulator_group -> WM_manipulator_group_type

Naming here is still a bit confusing,
now at least free/remove are differentiated.

Fix crash re-registering manipulators

Duplicating strings caused problems using strings in
both struct hash and manipulator group types own hash.

Fix T52479: Regression: Motion Tracking no longer works

Merge branch 'master' into blender2.8

Cleanup: uneven/double indentation

Modify menu from last commit

Also correct tool-tip.

Fix T52434: Restore mesh center of mass calculation

The new method while improved for solid objects
doesn't work for non-manifold meshes, keep both.

Eevee: Offset the for each AA sample.

This means we have less overall noise for rendered image.
SSR, AO, and Refraction are affected by this change.

SSR still exhibit artifacts because the reconstruction pattern needs to change every frame (TODO).

Fix T52473: blender internal Fresnel and Layer Weight only work with linked normal.

Please backport this to 2.79.

Revert "Cycles: remove square samples option."

This reverts commit 757c24b6bceaeeae95f743b72b6a7040880a0ebf.

We'll revisit this when doing deeper sampling changes.

Fix T52470: cycles OpenCL hair rendering not working after recent changes.

Cycles: attempt to recover from crashing CUDA/OpenCL drivers on Windows.

I don't know if this will actually work, needs testing. Ref T52064.

Fix Windows build error after recent Python changes in a10a7f42.

Cleanup: move gawain header out of GPU_matrix

Avoids requiring glew headers in Python API.

Code cleanup: deduplicate some bsdf node methods.

Fix T52439: Crash after adjusting lenght of hair particles.

Regression from rBfed853ea78221, calling this inside thread worker was
not really good idea anyway, and we already have all the code we need in
pre-threading init function, was just disabled for vertex particles
before.

To be backported to 2.79.

CMake version bump: 3.5