Blender Git Loki

Greatly improve speed of Delaunay when have a lot of holes.

Using part of a patch from Erik Abrahamsson, this replaces the
use of linked lists for original id tracking by Sets.
I had thought that the lists were unlikely to grow to more than
a few elements, but when the mesh has a lot of holes (whose
original ids go *outside* the hole, and therefore, most of the
mesh), this assumption can be very wrong.
On a Text regression test, the time went from 11.67s to 0.16s
with this fix. I also tested to make sure that Boolean didn't
slow down with this, and found it actually had a very slight speedup.

Using Sets exposed a dependency on the ordering of the items
in the id lists, luckily caught by a mesh intersect regression test,
so fixed that.

Fix: Incorrect logic in spline lookup function

This section of code deals with evaluated points,
so that is the size it should use.

Versioning: fix vertex group name loss in linked duplicates.

After rB3b6ee8cee708 by @HooglyBoogly vertex groups were moved
to mesh data, and versioning code was provided to upgrade old
files. However, it fails to consider the case of linked duplicates
having different name lists, and dependent on the object order
can cause some of the names to be lost. This can even be all of
them, if there is a duplicate without any names, which can be
easily created by lazy Python code.

To fix this, change the code to use the longest available name list.

Differential Revision: https://developer.blender.org/D11958

Speed up Delaunay raycast.

From Erik Abrahamsson, this uses parallel loops for raycasting.
It speeds up one example with many crossings of a bezier curve,
from 0.68s to 0.28s.

Cleanup: remove redundant parentheses

Cleanup: sort struct declarations

Fix error using uninitialized state in BLI_task_parallel_mempool

Single threaded operation used the state before it had variables
written into it.

Error in 15cdcb4e9085c3cf35528c2f7e559955b4ff531a.

Cleanup: reduce variable scope in task_iterator.c

Would have prevented the error in
15cdcb4e9085c3cf35528c2f7e559955b4ff531a.

Fix T70356: Scaling up 1x1 pixel image reads past buffer bounds

Also resolve a crash when when displaying thumbnails, see T89868.

Fix T89861: Checking face selection breaks UV stitch operator

Rename UI error message on attributes panel for name collisions

* Name Collisions > Name collisions
* Info icon > Error icon

Cleanup: preview rendering, update assumptions in comment

The `action_preview_render()` function used to just render, but now it
also temporarily applies the pose. Its comment is now updated for this.

No functional changes.

Fix T89782: Segfault populating popup menu with dimensions above the opengl limit

The crash happens because `GPU_offscreen_create` is called with `err_out` `NULL`.

This patch proposes a solution within the `GPU_offscreen_create` itself
and raises an error report in the interface if a menu is called with
dimensions beyond what is supported.

Ref T89782

Maniphest Tasks: T89782

Differential Revision: https://developer.blender.org/D11927

Fix T89875: False dependency cycle on particle systems

`POINT_CACHE_RESET` pointed to `GEOMETRY_EVAL_INIT` while
`GEOMETRY_EVAL_INIT` pointed to `POINT_CACHE_RESET`.

Now `POINT_CACHE_RESET` points to the same nodes pointed to by
`GEOMETRY_EVAL_INIT` thus avoiding the dependency cycle.

Surface Deform: support sparse binding mode for improving performance.

When a vertex group is used to limit the influence of the modifier
to a subset of vertices, binding data for vertices with zero weight
is not needed. This wastes memory, disk space and CPU cycles.

If the vertex group contents is known to be final and constant,
it is reasonable to optimize by only storing data group vertices.
This has to be an option in case the group can change.

Supporting this requires adding a vertex index field and spliting
the vertex count into mesh and bind variants, but both happen to
fit in available padding. The old numverts field is renamed to the
new bound vertex count field to maintain the array length invariant.
Versioning is used to initialize the other new fields.

If a file with sparse binding is opened in an old blender version,
it is corrupted into a non-sparse bind with vertex count mismatch,
preventing the modifier from working until rebind.

Differential Revision: https://developer.blender.org/D11924

Cleanup: Remove redundant logic

The object type was checked twice unnecessarily. Also use a function
for the check to be more explicit.

Cleanup: Get vertex group names directly from grease pencil data

Fix T89899: Crashes when accessing vertex groups from objects

We need to be more strict about trying to retrieve a list of vertex group
names from objects now, as only three object types support them.
This commit adds a check for vertex group support in a few places, the
data transfer operator/modifier, copying vertex groups to selected
objects, and the vertex group remove and clear functions.

Differential Revision: https://developer.blender.org/D11947

Cleanup: Move gpencil_geom.c to C++

This will help enable development on optimizations to the perimeter
calculation here. Using C++ data structures like Array can make the
code easier to read as well.

Longer term, this can help improve integration with attributes
and possibly the new curve code (since strokes and curves are
quite similar in theory).

Differential Revision: https://developer.blender.org/D11941

Fix T49944: Compositor ID Mask Anti-Aliasing not working

Replaces current ID Mask node Anti-Aliasing operation by SMAA
operations with default settings as proposed by Jeroen Bakker.
SMAA produces smoother edges.

Reviewed By: manzanilla

Differential Revision: https://developer.blender.org/D11881