The shader compilation job assumes it can only be closed on program
exit, leaving all their materials as queued.
However, render tasks can kill it, causing drw_deferred_shader_add to
get stuck in an infinite loop.
Pull Request: https://projects.blender.org/blender/blender/pulls/119172
Previously the CUDA context was always destroyed and the module along
with it. Now that this no longer happens, the missing module free became
a memory leak.
Also fix the same issue for HIP, though this is destroying the context
so it's not a problem yet.
Fix part of #119035
Co-authored-by: Brecht Van Lommel <brecht@blender.org>
Accessing `__class__` on a removed object raised an exception, making
`isinstance(ob, cls)` unreliable. Always allow class access.
Resolves issue raised by blender/blender-addons!104958.
If we run into NULL `RegionView3D` `regiondata` [which e.g. happens if
we just set `bpy.context.area.type = 'VIEW_3D'` without further actions
in the text editor prior to calling the transform operator], we can make
it gracefully work just by using `t->spacetype = SPACE_EMPTY` in
`initTransInfo`.
Similar check is already done in ba229e3859 (marked /* running in the
text editor */).
Transform code is smart enough to have fallback code in place that sets
matrices etc.
Pull Request: https://projects.blender.org/blender/blender/pulls/119205
When restoring a temporary context, account for changes to the context
made by actions (typically operators) in the script.
There was an incorrect assumption that an override which didn't change
the current context would also be unchanged when restoring the temporary
context's original values.
"r_map" is null for the generate layers operator (rather than the modifier
evaluation). It should disable the creation of the "CD_NORMAL" layer too,
like the crease, bevel weight, sharp edge, and uv seam attributes above.
As mentioned in new code comments, the auto smooth behavior in 4.0 was
to skip sharp angle tagging when the evaluated mesh had custom normals.
There was already a check for custom normals on the original mesh (we
can't access the evaluated mesh from versioning code). But that didn't
handle cases where custom normals were created by modifiers (the normal
edit and weighted normal modifiers). Now skip adding the new modifier
when those modifiers come last in the stack. Alternatively we could
check if they existed in the stack at all, but that seems a bit more
risky.
The `cache.curves_cache.proc_attributes_buf[i]` is independent of the subdivision level.
Therefore it's not freed when the subdivision level changes currently. However, the
check assumed that when no attribute was changed, the final subdivided buffer is
non-zero, which is not true when the subdivision level changed.
Now directly check whether the attribute is cached already.
Pull Request: https://projects.blender.org/blender/blender/pulls/119216
Global built-ins appear to not work on AMD cards.
Also add a tweak to avoid a performance regression, similar
to what was done before. Disable adaptive subdivision kernel
code if not used.
Pull Request: https://projects.blender.org/blender/blender/pulls/119175
When linking a collection from library, it can be linked inside another
linked/overrided collection if it is selected in outliner. This can be
prevented by linking with editable parent collection.
Pull Request: https://projects.blender.org/blender/blender/pulls/119144
The depth pass of the Viewport Compositor has low precision if the
precision of the node tree is set to Auto. To fix this, we always
expose the depth pass in full precision.
The new operator "View in Graph Editor" (#114407) only looked at the selection
when generating the FCurve bounds.
This failed in certain cases if things are not selected, or not selectable.
The simplest example is to key the emission strength of the world settings and
try to view that.
This was mentioned by Jonathan Lampel on the PR #114407
The fix is to also resolve the bounds of the animation data on the button.
This means for a few FCurves the bounds might be calculated twice but it
doesn't change the overall result.
This has another side effect though:
The Graph Editor might zoom into an area where there is no FCurve
visible because the thing is not selected so it isn't shown.
This issue existed before if filters were applied, but is now more pronounced
since by default the Graph Editor doesn't show unselected objects.
A warning is raised when at least one of the FCurves that are viewed is not
visible with the current filter settings. The view will still zoom into that area.
Pull Request: https://projects.blender.org/blender/blender/pulls/118658
Caused by 9c1da81a4c.
This happened in the case where objects shared the same data and slots
are assigned to object and receiving object did not have a material slot
yet.
Then, `ob->matbits` may be NULL, now just check for this, too.
Pull Request: https://projects.blender.org/blender/blender/pulls/119161
The property is actually there, but was marked PROP_SKIP_SAVE |
PROP_HIDDEN in 4234cddda9
This PR removes this again (with the consequence that shortcuts/menu
entries will now remember the select setting from the last time).
This is now in line to how editmode does this (the "Reveal Hidden"
operator).
Pull Request: https://projects.blender.org/blender/blender/pulls/119153
When in mesh editmode, skin_vertices point to bmesh customdata, the
skin_vertices data is empty since custom data is stored per element
instead of a single array there (same es UVs etc.).
Fix similar to 3e2619b3e7, 5179b8236b (basically disallowing the
access in editmode).
Pull Request: https://projects.blender.org/blender/blender/pulls/119149
In future files from 4.2, `object_to_world` is moved to a runtime struct
and therefore isn't saved in files. It was always considered runtime information
though, since it's built by depsgraph evaluation and copied back to the original
object. Therefore to have the proper data in `object_to_world` we need to
evaluate the depsgraph.
This partially solves #118694, but there is still an issue with the snap cursor
system never running before using the `plane_omat` value.
For example, creating the "position" attribute with the wrong name or type
could crash Blender when exiting edit mode. This is because some data isn't
stored as attributes in Blender, and the attribute API doesn't work very well
with BMesh.
Two parts to the solution:
- Remove builtin attributes with incorrect domains or names when
converting from BMesh to Mesh.
- Add error messages when creating builtin attributes in edit mode. It's still
possible to create name-convention attributes, because Blender should be
able to handle different types and domains for them.
Pull Request: https://projects.blender.org/blender/blender/pulls/119110
Regression in 427eed292d.
Root of the issue was that animation system was using a single same
check to decide if an F-Curve/driver was valid to use to animate some
data, and whether user can create/edit animation for that data.
Both cases are actually different, since e.g. linked data is not
user-editable, but it can still be animated (either by related linked
Actions, drivers defined in the linked data, or some more hackish
changes like py API/RNA scripting).
This commit now defines two checks:
* `RNA_property_animateable`: whether a RNA pointer & propoerty is
animateable, based on their types and definition.
* `RNA_property_anim_editable`: whether a specific data referenced by
the RNA pointer and property is effectively user-editable.
The new `driveable` check added by 427eed292d is also renamed to
`RNA_property_driver_editable` (since the basic type-based
`RNA_property_animateable` is also valid for drivers currently).
Pull Request: https://projects.blender.org/blender/blender/pulls/119089
The Realtime Compositor crashes with intricate node setups. That's
due to hardware limitations where shaders can't have as many output
images as needed by the compositor shader operations. To fix this, we
recursively split shader operations until their output count fits the
hardware limitation.
Pull Request: https://projects.blender.org/blender/blender/pulls/119075
Typically nodes that edit instances only change the top level of
instances. As an oversight, that wasn't done in the sort elements node.
This causes problems as the top level of instances are reordered and
nested instances are edited. As the top-level is replaced, its
contained geometry sets (the nested instances) are deleted.
To fix, only reoder the top-level instances in this node.
The "capture field on geometry" utility is used in many places to store
the result of field evaluation as an attribute. There is a special case
for when an attribute already exists with the same name, data type, and
domain. But when exactly the same attribute was stored, it would assert
because it ended up copying an array to itself. Arguments for those copy
functions aren't supposed to alias each other.
As a fix, clarify the logic a bit to return earlier in this case.
Also remove a redundant case handling the same thing later on
in the function.
Pull Request: https://projects.blender.org/blender/blender/pulls/119063
This PR adds an extra check to exit from sculpt gesture code early to
avoid attempting to allocate a 3 * (UINT_MAX - 1) amount of elements
inside the trim gesture code.
Pull Request: https://projects.blender.org/blender/blender/pulls/119073
The armature reported had two bones with a `bMotionPath` but its lengths
are zero (which causes trouble in `motion_path_cache` drawing code due
to 0 allocations).
Not exactly sure how we got there, something like
`animviz_verify_motionpaths` should take care of this already in current
code, but this might be from a time where there were not enough sanity
checks.
So now early out in `motion_path_cache` if we encounter such a
"corrupted" motion path.
Pull Request: https://projects.blender.org/blender/blender/pulls/119081
This can happen e.g. when relative "//" is in Preferences > File Paths >
Temporary Files is used.
Now catch the corresponding COLLADASW::StreamWriterException
and cancel export then.
NOTE: 51126fab33 might have prevented the most common case
to run into this crash (but there might be other cases still so being safe
here does not hurt)
Pull Request: https://projects.blender.org/blender/blender/pulls/118958
Regression from 427eed292d, in IDProperty case the 'ensured' PropertyRNA
pointer is some sort of empty template, and cannot be used for many
things.
Instead of calling `rna_ensure_property` twice for each call to
`RNA_property_animateable`, rather add an optional parameter to static
helper `rna_property_editable_do`, and call it directly from
`RNA_property_animateable`.
