Blender Git Loki

Take cache library items into account when baking caches.

This means we don't bluntly store all the data in a group, but use the
selection from the cache library. It also helps to avoid issues with
object visibility which is not yet stored in the cache.

Some improvements/tweaks from review.

Image node, small fix, though it doesn't fix the main issue I'm facing

Working implementation of cache reading for dupli data during renders.

Merge branch 'master' into temp-ghash-setops

From review: UI re-arrangement for Camera Stereoscopy panel

From review: redraw window when scene.use_multiview changes

Remove public GHash API for union/inter/diff/symmdiff.

This is dead easy to add back when we want (or use other naming if we really
like them better), core code remains exactly the same anyway.

Note we keep checks (isdisjoint, isequal, etc.) for ghashes though, those are
pure-keys non-modifying operations anyway, so here set ops for ghash are
definitively valid (equivalent of set ops on dict.keys() views in py).

Fix for updates in master...

Revert "Remove gtests."

This reverts commit 2b8fac2a153e135d0f6cc7621a3755901477fd3d.

Remove gtests.

Squashed commit of temp-ghash-experiments, minus the 'hash storage' part.

Merge remote-tracking branch 'origin/master' into multiview

Conflicts:
source/blender/makesrna/intern/rna_scene.c

Use the temporary 'is_dupli' flag to prevent objects from freeing
overriden data that is owned by the cache.

This is not at all nice ... Hopefully it doesn't get too complicated to
work around all these drawing code and depsgraph issues, so the code
can be understood and replaced at some point in the future.

Replacing boundbox temporarily during drawing is dangerous, since it
can be freed at any point.

Luckily we only need the boundbox locally for testing dupli visibility.

Fully reset the OB_FROM_DUPLI flag for each dupli instance, in case
later duplis can not be found in the cache.

Distinguish realtime/render setting in the Object writer.

Now the derived_render mesh version is created locally when writing
render results.

From review: fix Cycles tiles udpate not working *

* note: I tried passing b_combined_pass to the do_write_update_render_result() function
but b_rlay is created in do_write_update_render_tile, thus it's not the same that we have on run().

fix: another possible fix for loading correct vertexbased constraints

fix: vertexbased constraints should be rebuilt correctly now after loading