Blender Git Loki

some more mutex locking for dynamic fracture, there still were crashes when interactively moving dynamic objects during running sim

added threading lock to modifier sync during rigidbody object sync, helps against crashes in dynamic mode

improved output of constraint impulse visualization a bit, included name, impulse and threshold as well as attempted to offset strings at same position

windows compile fix

added constraint impulse visualization (very basic only) and reworked the mesh constraint name system

compile fix for windows and mac

added basic physics visualization for viewport rigidbodies like in game engine (its optional and can be turned off)

ignore individual mesh island transforms in rigidbody system again and just re-apply them when converting back to objects, it just caused trouble and brought not the expected result

fixes for: did assume autohide material was always at number 1, did change to find inner material number, own verts transformation bug, shards to islands delay bug (needed to click twice before), and some crashers with dynamic and prefractured as well as reduced a memoryleak in prefracture mode

some more effort to make FM simulation results looking closer to BCB simulation results

attempt to reduce a memoryleak

exposed bullet tick callback to bpy.app.handlers, also wrapped the passed timestep into the rigidbody world, so it can be accessed too

fixes for dynamic and external mode, in external mode each simulation run should be consistent with the prior ones now

fix, small typo in appliedImpulse soft limit (read only)

added appliedImpulse() to regular constraints too

added a RNA function to read the appliedImpulse from each constraints live, during simulation (slow when done from a python frame handler over all constraints)

fix: dynamic mode didnt work at all anymore and has still some issues like flickering after baking and a crasher on exiting blender, also testwise enabled autoexecute on prefracture during sim

fix: added extra update call on hook object in custom modifier evaluation

fix: forgot to accumulate differences of multiple hook modifier changes

fix: fake parenting in testfile didnt work anymore after fake hook was added to code, now both should work depending on whether the kinematic flag is set (=parenting) or not (=hook, if vertexgroup and hookmodifiers are used)