Blender Git Commits

RNA: disallow negative fcurve data-path array index

Cleanup: unused variables

Cleanup: spelling

Cleanup: doxy comments (use colon after parameter name)

Also remove colon after `note`.

Cleanup: reduce indirect DNA header inclusion

Remove DNA headers, using forward declarations where possible.

Also removed duplicate header, header including it's self
and unnecessary inclusion of libc system headers from BKE header.

Cleanup: Reduce variable scope

Cleanup: type check failure, discarded-qualifiers warnings

Caused by 7f6ffe9195c105a2ed9776b907aa2e6e34ed4bf3

Cleanup: Use LISTBASE_FOREACH macro, reduce variable scope

NLA: Action Strip Defaults Sync Length On

See T82309#1055564 {T63675} and their duplicates for how Default-off
can cause confusion.

This is just for convenience since it allows animators to keyframe
outside of the strip's bounds by default. This was likely off by
default before since Syncing Length would undesirably shift the whole
animation after leaving tweak mode (fixed by {D7602}) and the animator
wasn't able to keyframe outside the strip bounds anyways
(fixed by {D7533}). Now it's better if the flag was on by default.
While the animator is still roughly developing their animation NLA-wise
they won't have to worry about strip bound keying failures. When they
are more certain of the strip bounds, they can disable the flag to
prevent affecting the rest of the NLA system.

Reviewed By: sybren

Differential Revision: https://developer.blender.org/D9661

Nla Refactor: Blend functions explicit Div0 check

It's an explicit check to prevent division by zero if caller hasn't
done the check. Future patch {D8867} will not use the nla remap
function and thus not do the check. This patch also replaces some
float (==) equality checks with IS_EQF().

Split from {D9247}

Reviewed By: sybren

Differential Revision: https://developer.blender.org/D9694

Nla Refactor: nlaevalchan_validate_index_ex()

And removes redundant index local variable. Future patches {D8296} and
{D8867} make use of this function.

No functional changes.
Split from {D9247}

Reviewed By: sybren

Differential Revision: https://developer.blender.org/D9693

NLA: Fix context.selected_nla_strips PointerRNA

The strips were given the wrong owner ID. This only caused issues for
python based UI as far as I know. Property changes would not properly
update the viewport.

Reviewed By: lichtwerk

Differential Revision: https://developer.blender.org/D9685

Fix incorrect RNA type warning

Using the RNA type for regular modifiers instead of grease pencil
modifiers caused a warning in context.c.

Geometry Nodes: Input data type utility function

This commit adds a simple utility function for getting the data type of an
attribute or its "constant" socket counterparts. No functional changes.

Differential Revision: https://developer.blender.org/D9819

Fix several issues with handling of numpy in CMake.

Issues were:
* Abusing of `WITH_PYTHON_INSTALL_NUMPY` by both Audaspace and
Mantaflow.
- `PYTHON_INSTALL` options only decide whether we copy python (and
some extra modules) in our Blender installation. On linux it
makes much more sense to use global python installation.
- Now we have instead a proper `WITH_PYTHON_NUMPY`
* Bad assumptions regarding path of headers relative to path of python
module.
- In current Debian testing, modules are under `python3.9`
directory, while headers are under `python3` directory.
- Now we properly `find_path` for headers as well, modifying
`find_python_package` to take an optional argument for headers.

Note that the required changes done to `extern` libraries are in
blender-specific files that do not exist upstream.

Differential Revision: https://developer.blender.org/D9773

Cryptomatte: Data structure in compositor node

This changes the way how the mattes are stored in the compositor node. This used to
be a single string what was decoded/encoded when needed. The new data structure
stores all entries in `CryptomatteEntry` and is converted to the old `matte_id`
property on the fly.

This is done for some future changes in the workflow where a more structured
approach leads to less confusing and easier to read code.

Geometry Nodes: enabled supports-mapping in nodes modifier

Sometimes the geometry nodes modifier does support mapping and
sometimes it does not. We have no infrastruture to determine this ahead
of time currently. In order to support common use cases, it makes sense
to add this flag to the modifier.

One such common use case is to use the mesh as surface that other
things are distributed on. Often, the distribution is controlled by vertex
groups. Therefore, it would be helpful if the modifier is evaluated
when the object is in vertex paint mode. This allows the user to see the
distributed objects while painting.

If the nodes modifier transforms the mesh in any way, vertex painting
might not work as expected anymore, because the `deformMatrices`
callback is not implemented. I'm not sure how this can be solved nicely, yet.

Geometry Nodes: require vertex groups data in nodes modifier

Without this, the modifier evaluation code might remove any
vertex groups from the mesh for performance reasons.
We can't say for sure whether the node group will need the vertex
groups, but it is quite likely.

Ref T83357.

Cryptomatte: apply volume transmittance to Eevee.

This patch will add volumetric transmittance to the cryptomatte coverage
data of all samples when post processing the cryptomatte passes.

It was discussed with Cycles that this is desired, but tricky to
implement in Cycles.

Fix memory leak and possible other issues with custom previews

Mistake in 812ea9184221. Fixed that same one before in the branch, I may have
brought it back when resolving merge conflicts.