- `_PySet_NextEntry` has been removed, use generic iterator access
which will has some additional overhead as it needs to create
an iterator to access the values.
- Add v3.13 compatibility defines to account for renaming:
_PyObject_LookupAttr -> PyObject_GetOptionalAttr
_PyLong_AsInt -> PyLong_AsInt
- Unfortunately use of Python's internal API needs to be used to
inspect op-codes in `bpy_driver.cc`.
Testing GLTF/FBX IO there isn't any significant performance impact
from these changes.
Resolves#123871.
PR #128051 made some improvements to code by cleaning up properties and
simplifying logic.
However, the default tool in most spaces is box select. By switching to
box select default, there should be more front-facing consistency across
spaces while still retaining near-identical behavior.
This patch would fix the small bug listed in #128671 while keeping the
simplified code benefits from #128051.
Also:
- Move selection keymap items up in the Sequencer (Global) keymap so it
is more visible (similar to Node Editor ordering).
- Split `side_of_frame` property for ctrl press keymap item into a
separate keymap item on ctrl click for both LCS and RCS to avoid
clashing with ctrl+drag for box selects in RCS.
Pull Request: https://projects.blender.org/blender/blender/pulls/129028
When starting a docking operation from an assigned shortcut, the source
area will be incorrect if the Status Bar is not currently being shown.
This PR just adds defaults that are never valid window coordinates.
Pull Request: https://projects.blender.org/blender/blender/pulls/129187
This property was available for `layers` but was missed for layer groups.
Adds a `parent_group` property for layer groups that returns the
parent group if the group is inside another group.
For groups that are inside the root, returns `None`.
Pull Request: https://projects.blender.org/blender/blender/pulls/129168
Essentially, our current code would not properly remove (dereference)
its python objects matching various RNA data created during execution.
Some cases are fairly trivial to understand (like the lack of handling
of unregstering for our 'startup' operators and UI), other were more
subtle (like unregistered PropertyGroups who would remove/free their RNA
struct definition, without releasing first the potential matching python
object).
Co-authored-by: Campbell Barton <campbell@blender.org>
Pull Request: https://projects.blender.org/blender/blender/pulls/128899
If ifdef for `WITH_ANIM_BAKLAVA` accidentally removed the `else { ... }` branch
around `name = RNA_struct_name_get_alloc(...)`. For data-block names,
`BKE_id_full_name_ui_prefix_get` is used above which adds 3 extra letters before
the data-block name. This are removed again by the bug. Later code removes them
again leaving potentially an empty string.
Pull Request: https://projects.blender.org/blender/blender/pulls/129153
During the painting operation, the `GreasePencilDrawing` is updated in-place.
Previously, we simply tagged the whole drawing for a topology update.
This meant that the triangulation and curve normals were recomputed
for (pretty much) every new input sample.
Since the draw tool only creates one new stroke, we can copy the
triangulation of all the other strokes and only recompute the triangulation
of the newly drawn stroke(s).
First, a new `triangle_offsets_cache` is added that is lazily computed.
The computation for this should be pretty cheap.
Then, a new function `Drawing::tag_topology_changed(const IndexMask &curves_to_update)`
is added. This function takes an index mask of curves and only updates
their triangle cache. The cache for the other curves is copied.
In a test file with around 400k points, recomputing the triangle cache
(every input sample) took around 45ms. This meant that there was a noticable
drop in frame rate. With this patch, the time for each sample goes down to
less than 0.1ms. The frame rate feels much better.
But, drawing a longer and longer stroke still makes this time go up,
because the triangulation for the single stroke becomes more and
more expensive. This will have to be improved further.
Part of #124149.
Pull Request: https://projects.blender.org/blender/blender/pulls/129115
This patch adds support for passes in the new CPU compositor. This
involves rewriting the get_input_texture method into a get_pass methods
that returns a result as opposed to a texture. The result wraps the
cached GPU texture or image buffer depending on the execution device.
The Render Layers node was implemented for CPU execution and a new
utility constructor for the result class was added to determine type and
precision based on GPU texture format. The fallback depth pass that was
retrieved from the viewport frame buffer was removed, as it was a hack
that can no longer be supported due to the use of stencil format.
Pull Request: https://projects.blender.org/blender/blender/pulls/129154
Changes would be lost (or worse things can happen, see below), so this
is now prevented by polishing the polls for these operations.
Behavior of Materials in override objects was considered shaky as well (lost
on reload, no undo or crash on undo, see #127605, #127606, #101552 in
general) , so conclusion was to prevent this on linked as well as liboverride
objects.
Pull Request: https://projects.blender.org/blender/blender/pulls/129064
From feedback: "Update Asset" sounds more like receiving a new version
of the brush, while this actually saves the changed brush to the asset
library. "Save Changes to Asset" is more clear and fits well with the
labels of other menu items in the menu.
The leak happens when there is a muted group node in a shader node tree which is
rendered. For rendering, the drawing code "localizes" the node tree which also
means duplicating the node groups used by each group node (might even duplicate
the same group more than once if it's used by multiple nodes). Generally, this
works fine, because all of these duplicates are freed when the tree is
flattened.
However, there is a preprocessing step which deletes all muted nodes from the
tree. This code path did not free the groups recursively.
Pull Request: https://projects.blender.org/blender/blender/pulls/129124
Design of the essentials asset library is to treat it as part of
Blender (as if it were compiled into the binary), so the location and
state of these assets is clear and can be assumed in code. User edits
are not expected.
Because of that we should only look for this asset library in the
installation ("system") location, not in the user configuration which
would take priority if present. On some devs machines this location
would actually be present, making the essentials unavailable and causing
confusing/misleading warning prints, see #128420.
When trying to select bones that are inside other bones which are not selectable,
the selection would fail. This is inconsistent with object mode, where unselectable
things are just ignored.
This patch fixes it for pose mode and edit mode.
Note that there are two areas for edit mode that were modified.
The one in `ed_armature_pick_bone_from_selectbuffer/202` is never actually reached,
but I added the code there for completeness.
The code in `get_nearest_editbonepoint` had to be modified some more to
implement a similar logic as `armature_select.cc/220`
Pull Request: https://projects.blender.org/blender/blender/pulls/129120
The "Set Handle Type" operator didn't check the curve types
(not if the curves are editable for that matter).
This fix retrieves the edtiable bézier curves and only changes
the handle types of those.
Running in a headless weston session was asserting, then crashing,
causing WITH_UI_TESTS to fail.
Resolve by accounting for a wayland sessions without a seat.
