Blender Git Loki

Sanity checks for Alembic archive validity to avoid crashes when trying
an invalid file operation (e.g. overwriting files).

Alembic caching support for particle paths.

This hooks into the particle path update functions and reads cache data
if available instead of constructing expensive child paths every time.

Currently the cache grows quite rapidly due to lack of optimized sample
storage: constant values such as numer of segments, times, colors etc
only need to be stored in the first sample. "Velocities" (actually
position deltas) could also be calculated on-the-fly, which is
relatively cheap compared to full child calculation. In any case, disk
space it cheap ...

Fix for constructing cache reader/writer based on RNA pointer.

Because of the awkward internal cloth modifier inside the particle
system, we have to look at the psys to determine if we actually want to
use the particle system cache or the internal cloth modifier cache.

Sanity NULL check when freeing caches.

Merge branch 'master' into alembic_pointcache

Merge branch 'master' into asset-experiments

Depagraph: Partial fix for missing updates when changing action

This doesn't totally work because of the missing tag re-scheduling
described in the previous commit. However, it at least ensures relations
are correct which hopefully will prevent some crashes.

Depsgraph: Fix missing animation update when tweaking fcurves in graph editor

Not totally happy with this solution, mainly:

- It's a bit weird to explicitly tag ID for update when updating animation data,
thinking maybe we can have function like DAG_anim_data_tag_update(). But it is
also a bit tricky to get back from animation data to ID it seems.

- Re-scheduling time update tags is not supported at this moment. This means if
both relations are tagged for update and animation data is tagged for update
the animation update will be missing.

Anyway, it's an interesting issue and think it's good starting patch for it.

Preview generation on Windows: attempt to fix mysterious error by using even more ID's names.

Passing around ID's themselves is not safe - they may change their mem addresses.

Yet, I had no issue on linux, but on win build this is still not enough,
have to use a super strange hack (add an 'import sys' line just before where error happens)
to get things running...

Store particle index and randomized value from particle instance
modifiers, for use in shaders.

The particle instance modifier can now store 2 custom data layers:
- Index: the integer index of the particle
- Value: a randomized floating point number for shader variation

These layers have user-defined names, so multiple particle instance
modifiers can be used without overwriting customdata layers. In Cycles
the data can be accessed using an Attribute node with the same names.

Merge branch 'master' into asset-experiments

Depsgraph: Make it more clear in TODO that space check is needed in previous constraint commit

Depsgraph: Fix missing curve path in certain situations

Basically this is a quick re-implementation of evaluation flags we
had in the old dependency graph. I know it is not the real solution
but it's not so bad as well and would allow us continue nailing all
the other bugs with current granularity level before we go more
advanced.

Merge branch 'master' into gooseberry

Conflicts:
source/blender/editors/space_sequencer/sequencer_draw.c

Depsgraph: Fix flickering of copy rot/loc/scale/transform constraints

Those constraints in general needs world's matricies, which wasn't reflected
in the relations.

Merge branch 'master' into depsgraph_refactor

Merge branch 'master' into depsgraph_refactor

revert bad activation fix for triggering, caused crashes in some cases (like on non-kinematic triggers)

Merge branch 'master' into asset-experiments

Conflicts:
source/blender/editors/space_file/file_draw.c
source/blender/editors/space_file/filelist.c

Cleanup of ugly allocation stuff (cast of void pointer, not-so-safe sizeof()'s, etc.).