Blender Git Loki

insert_vert_fcurve sets active keyframe

Activate keyframe before deselecting

Removed VLA's leftover from the 2018 GSoC project.

Fixed bug when not using light tree from merge.

GPencil: Remove curve flag and replace with RNA dynamic prop

We don't need this flag because we can check the struct pointer itself.

Merge branch 'master' into greasepencil-object

Merge branch 'master' into active-fcurve-keyframe

GPencil: FIx flag GP_STROKE_CURVE_MODE

The flag was set to 2^11 and it was 2^16 before. The type of the flag is short, so it was
never actually beeing set.

GPencil: Add 'select all' operator for curve

Change the 'select all' operator so that, if the stroke
is in curve edit mode, it selects the curve handles.

GPencil: Use curve fitting write data for operator

Instead of writing test data to editcurve, use the
BKE_gpencil_stroke_curve_create function to write fitted
curve data to editcurve.

GPencil: Add function to create editcurve from stroke

This BKE function uses curve_fit_cubic_to_points_fl to fit a curve
to a grease pencil stroke.

EEVEE: Motion Blur: Use closest interpolation for sampling

This Avoid halo artifacts

Merge branch 'master' into greasepencil-object

added tetgen

EEVEE: Motion Blur: Split both direction accumulation

Merge branch 'master' into cycles_texture_cache

added tetgen

Add faster fprintf writer, remove unused headers

Changes here:

Reverted temporarily to std::vector to keep working, instead of
waiting for D7931 to get committed in master.

Adds an forced inline function for calculation of face/ polygon normals
by averaging vertex normals. I will look for an existing method.
This way, the allocated memory in `data_to_export->mvert[i].no` is
actually used.
Also, `BLI::Vector<uint> face_normal_index` is not needed anymore since
we loop over the same polygon list while writing normals, so
their indices will be the same.

Adds a writer method option, `fprintf`, which is faster than `fstream`.
With fstream, 478 MB of an ico-sphere with 8 subsurf takes 22 seconds.
with fprintf, the same takes 13 seconds.
With fstream, a 44 MB of cube with 8 subsurf take 2.3 seconds.
with fprintf, the same takes 1.4 seconds.

Adds timing info of the full export directly in console.

Removed unused and repeated headers from `wavefront_obj.cc`.

Differential Revision: https://developer.blender.org/D7959

EEVEE: Motion Blur: Do not request motion blur data in viewport

EEVEE: Motion Blur: Simplify shader code