Blender Git Commits

Depsgraph: Fix memory leak in root graph node

We'd better switch to OBJECT_GUARDED_NEW() so we see memory leaks more instant.

Depsgraph: Workaround to be able to compile with Clang

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

Depsgraph: Allow transforming animated objects

The issue was solved by skipping flush to animation component
when tagging ID block for an update. This seems to be reasonable
thing to do: animation should only be tagged directly when editing
in the animation editor or via the time source node update.

Depagraph: proper relations for lattice modifier

Depsgraph: Cleanup, move helper functions to an anonymous namespace

Depsgraph: De-duplicate root ik solver bone find code

Depsgraph: Don't evaluate drivers/nla from action evaluation callback

Merge branch 'master' into depsgraph_refactor

Merge branch 'master' into depsgraph_refactor

Depsgraph: Experiment with bettr handling of parents when there are IK solvers

The main issue here is that pchans from the same IK solver chain should use pre-solver
position of the parent pchan. But in cases when pchan's parent is in different solver
it's child should depend on the solved result of the bone transform.

The last rule was violated in previous implementation. Now added quite quick and not
so quick code to check if pchan and it's parent are in the same IK solver and made
dependencies built based on this information/

This isn't final at all, committing just for the reference. Next step would be to gather
all collected knowledge from this week and implement rig dependencies in a really clean
fashion.

Depsgraph: Workaround-ish hook of on_visible_update to the new depsgraph

It should be totally re-implemented actually, but current change is good
enough for testing purposes. So this way opening the file with new DEG
enabled from the command line wouldn't end up in some unknown state.

Merge branch 'master' into depsgraph_refactor

Depsgraph: Use proper dependency for the armature modifier

Depsgraph: Correction to the previous commit

For until granular updates uses proper intermediate storage need to handle
dependencies to bones from self a bit different.

Depsgraph: Fix wrong dependency from the bone target in IK solver

Depsgraph: Corrected dependencies for hook modifier and bone parent

Victor rig is still broken tho..

Depsgraph: Hook up spline IK solver to the granular update

A bit of test of prove-of-concept of the design made in the previous commit.
still needs a bit of cleanup (de-duplicate some logic, move functions to their
final location etc).

But currently it's good enough for testing i guess.

Merge branch 'master' into depsgraph_refactor

Depsgraph: Initial move towards more granular armatures update

Main goal is to be able to do interleaved updates of bones from different armatures,
so it'll be possible to drive some bones of armature B with bones of armature B and
drive other bones in an opposite direction. This also beans interleaved update of
bones from the same armature should be possible.

This commit adds an initial support of this but have some limitation still, for
example drivers are not supported yet at all. Spline IK is also not supported in the
granular update. IK solver is so called "seems to work" state, but might need some
extra work.

Some technical details of the changes:

- Basically just filled in existing callbacks which so far were doing nothing,
also needed to fix some typos in the code.

- Changed the Bone component a bit, so now it does have "Final Transform" NOOP
node which would guarantee Bone component depends on the IK solver if it exists.

So now all the external dependencies would use bone transform after the solver,
plus bones child which are not in the solver would use solved transform of the
bone.

This required changes in the relations of the IK solver as well, so now it
depends on the bone local transform and optionally on the bone constraint stack.

- Currently scene is being passed for the current time, in the future we need to
pass time source node instead.

This would likely break some setups which used to work before going granular, but
even at that time update wasn't finished at all.

There are some known TODOs about making code more clean in the change, will be
solved later.