It might be nice to keep the pinning but that's a bit more tricky technically
currently because the pinned node tree is overridden by the tool node tree.
Just disabling pinning solves the main issue right now which is that the
node editor shows the wrong node tree.
Pull Request: https://projects.blender.org/blender/blender/pulls/118811
When `Delete Light Bake` operation was executed, the light bake got deleted,
but visible viewports didn't update and required a tweak to be redrawn.
The reason is that it an event was fired, where the viewport didn't listen to.
This PR fixes this by using the `ND_DRAW` event.
Pull Request: https://projects.blender.org/blender/blender/pulls/118803
This PR adds a new method to BMLog to explicitly make the intent of
searching for an entry clear and adjusts the calling code in `pbvh_bmesh`
to utilize it.
Pull Request: https://projects.blender.org/blender/blender/pulls/118543
When an asset is heavily modified, all production files having
liboverrides of it will still try to link all their known linked
reference IDs, leading to potentially thousands of not-really-useful
warnings about missing IDs in the console.
Now that liboverrides resync cleans up better these left-over data, it's
better to report missing linked data _after_ the liboverride resync
process.
Note that the original place can still report all effectively missing
linked data if needed, but this is now a logging info, so it won't be
displayed anywhere unless explicitely requested.
Pull Request: https://projects.blender.org/blender/blender/pulls/118577
Often when the reference linked data is significantly modified, a lot of
'ghost' linked data remain referenced by liboverrides, even after
resync. This is due to the fact that missing data is ignored (skipped)
during resync process, to avoid potential destruction of data in case
the linked data is actually missing.
However, after all resync has been done, we can consider that missing
linked references and their liboverrides can be safely deleted, if the
later are not user-edited or hierarchy roots.
There are cases where the brush would not have any
`gpencil_settings` allocated.
In this case we could crash in the eraser tool (and also in the draw tool)
This fixes the issue by calling `BKE_brush_init_gpencil_settings`
if the `gpencil_settings` on the brush are `nullptr`.
Nested panels are not supported currently, and this parameter serves no
actual purposes. Only the root panel supports adding child panels and it
is not user-accessible (adding the root panel is done using a nullptr
for the parent).
Pull Request: https://projects.blender.org/blender/blender/pulls/118792
Internal link is not created for a node when it has rotation output
socket. This is because the missing case for `sock_rotation` in
`get_internal_link_type_priority()` which fails to find input_socket
priority in this case.
Pull Request: https://projects.blender.org/blender/blender/pulls/118735
In separate layer and possibly with other operators, execution stuck in
infinite loop when there is no material to remove in `BKE_object_material_slot_remove`.
Pull Request: https://projects.blender.org/blender/blender/pulls/118796
This is to accommodate Position and Normal attributes.
The normal used to be optional but isn't nowadays.
So the limit is actually 14 attributes until we do some big refactoring of
the attribute fetching.
Pull Request: https://projects.blender.org/blender/blender/pulls/118441
Performing an off-screen draw call while drawing the viewport isn't
supported, add a check that raises an exception when called from Python
instead of crashing.
Ref: !118780
This was reported as #93410 & fixed with [0] however the fix didn't
help for animated logic (with region overlap enabled).
Resolve by adding a function ED_region_visibility_change_update_ex which
is called by the animated and non-animated region hiding functions.
[0]: 8f69c91408
Blender would crash if the input of the Classic Kuwahara node is
translated. This is due to an out of bound access due to the miss-use of
IndexRange, where it was assumed to have a start-end constructor, while
it was in fact a start-size one. Fix this by computing the size from the
area and supplying it to the constructor.
This patch unifies the Keying node between the CPU and GPU. The
difference was due to imprecision when computing pixel saturation in the
CPU code. This typically happens when the pixel is gray-scale, and
should have a zero saturation, but it came out negative due to
imprecision. To fix this, reorder the mix expression so that the minimum
and maximum values are differenced first.
Ideally, the code shouldn't have such abrupt conditionals and should be
more robust to tiny imprecisions, but that would break the behavior and
should be evaluated separately.
Pull Request: https://projects.blender.org/blender/blender/pulls/118625
The GPU compositor Blur node truncates the filter kernel if the blur
radius is lower than the base radius. That's because it evaluated the
filter kernel only up to the desired radius. To fix this, we evaluate
at the entire kernel for all radii using a texture sampler for the
weights.
Enums are stored as uints, but due to a missing implementation they
where stored in shaders as ints. As draw manager now supports uints
as specialization constants we can update these constants to be
stored as uints on the shader side as well.
Pull Request: https://projects.blender.org/blender/blender/pulls/118788
