Blender Git Commits

probably it is not necessary to clear the cache on every rna property change

like changing autohide / automerge parameters which are considered live postprocess steps caused the mesh to disappear when not on cache startframe.

temporarily bringing back the legacy subsurf code for performance reasons

opensubdivs subdivision code has currently an abysmal performance and is thus not yet
really usable, especially with many ngons and variable topology in animations.
After opensubdivs performance issues have been fixed, this legacy code can go again.
Not before.

re-added metaball remesher and deform motionblur in cycles for it and FM

added testwise threaded fracture via openmp

after more testing this might be converted to task, but works ok already, and added only
trivial changes to the code

submodule pointers.. yawn

fixed some memory leaks, removed unused "reset" parameter in fracture operator

also for now disabled "Geometry limitation tree", intent was to accelerate voronoi + bisect
by only processing "necessary" new geometry (and omitting unchanged), but doesnt work correctly

Merge remote-tracking branch 'origin/blender2.8' into fracture_modifier-2.8

# Conflicts:
# source/blender/modifiers/intern/MOD_boolean.c

forgot to display rigidbody settings subpanel with opengl render engine

crash fix attempt when adding / removing rigidbody worlds

still issues when appending FM objects, need to re-add rigidbodies and refracture

added shape-based clusters

clustergroup object with meshes will now cut out clusters according to their mesh,
make sure mesh is subdivided enough for better results

fix attempt for having proper constraints with dynamic fracture again

"mixed" and "all" settings were basically not properly taken into account before

added new cluster trigger propagation feature and a couple of fixes

fixes for example for split to island crashs

forgot to free duplilist after usage, and re-added previously removed depsgraph relation

relation is necessary to smoothly looking continued motion after dynamic fracture

attempt to add more detail for dynamic fracturing

fix for external constraints (centroid)

crash fix attempts for split to island function

try to avoid crashes due to excessive usage of immediate draw buffer

since the buffersize is limited to 4 MB (but it effectively crashed above usage of 3 MB,
probably due to 4 bit padding), there is a hard limit of usage ratio to be taken into account. The debug drawer counts the bytes used and increases an epsilon value as softlimit
first. This filters out short lines / very close verts and reduces the buffer usage. If that is not enough, there is a hard limit at 95% buffer usage. Then line drawing is skipped.

got physics visualization operational again

kudos to hypersomniac for pointing me to the debug draw api

attempt to fix pack group and export to objects

attempt for fixing fake cloth + dynamic fracture