It looks like the "user edited flag propagation" node built for the geometry
nodes modifier is unchanged after modifiers are removed. My guess is this
is now a more general problem when modifiers are removed-- I guess the
depsgraph nodes for an object have to be rebuilt when that happens now.
Maybe there is a more general fix, but to fix the crash mentioned in the
report, I tagged the depsgraph for a relations update, which rebuilds the
problematic modifier node.
Pull Request: https://projects.blender.org/blender/blender/pulls/124252
Viewport instances created by the Viewport Render Animation operator
don't get `view_updated` notifications.
Fix update detection by implementing a `last_update` timestamp for
`World`, similar to the ones added in #115196.
Pull Request: https://projects.blender.org/blender/blender/pulls/124114
The issue was that the function `animfilter_action` got a new
code segment at the start which checked `if (action.is_empty())`.
That function didn't check if the `groups` list of the action is empty
though. Regular keyframe animation is usually sorted into
keyframe groups, which means it is not stored under `curves` of the action.
However in the anim filtering code, the function `split_groups_action_temp`
moves the fcurves to their groups under the `groups` listbase.
Pull Request: https://projects.blender.org/blender/blender/pulls/124172
The openvdb API was used incorrectly before. I didn't notice that this is
a static function and requires the type to be passed in.
Note that this does not solve the cache invalidation issue yet. Instead
it handles the case more gracefully when the stored .vdb file changes
when Blender doesn't expect it to change.
Some users were concerned that the tooltip wasn't clear enough on
indicating that Blender doesn't really inforce an offline mode.
The final text was a collaborative effort together with the UI team and
participants from the user-interface-module chat.
The tooltip talks about internet instead of specifically extensions or
repositories to leave room to be used in the future for other things
(e.g., for Blender to check for new release updates).
Caused by 396ad5db83
It is possible that some edge collapsing of non-manifold mesh will
eventually result in extra wire edges, and loose vertices. This was
not properly handled in the boundary checks, assuming that all
modifications preserve mesh manifold.
This fix avoids the crash by adding nullptr check in the boundary
check.
While this is not fully ideal from the result perspective, it is
a safe change for 4.2. Ideally the wire edges and loose vertices
will be removed, but this is a bigger and more risky change. Also,
in Blender 4.0 it was possible to generate loose geometry in
dyntopo as well, so it just a general improvement to happen (and not
a regression).
Pull Request: https://projects.blender.org/blender/blender/pulls/124236
EEVEE Raytracing on Intel Arc wasn't working as there were no rays
generated. The reason was that the raytract tile compact shader
didn't set count the correct tiles it needed due to atomic operations
that were initialized without atomic.
This PR solves the issue by using atomic operations to initialize
the counts. We also tried memory barriers but that didn't fix the
problem.
Other shaders were also tested if they have this issue, but they were
setup in a specific shader or where done using a race condition (SSS) which
is working.
Pull Request: https://projects.blender.org/blender/blender/pulls/124213
Instead of making the dark outline be bright for active but unselected
strips, make it have dark outline just like all other strips,
and a 1px inner bright line inside of that.
Pull Request: https://projects.blender.org/blender/blender/pulls/124210
Because a thumbnail image can get cropped, we can not calculate
horizontal zoom factor once for all thumbnails. Cropping happens
at integer coordinates, so recalculate zoom_x for each thumbnail
based on final image size.
Pull Request: https://projects.blender.org/blender/blender/pulls/124178
Old python STL exporter, as well as other exporters like OBJ,
reverse the face order when object being exported has odd number
of negative scales in the matrix. The C++ STL exporter was lacking
that, resulting in the exported object looking "inside out".
The extra branch inside triangle export inner loop has no measurable
performance impact, probably because it is entirely predictable.
Pull Request: https://projects.blender.org/blender/blender/pulls/124219
When linking object data between two armatures the selection would not work
on the armature which got the new armature data assigned.
That is until a bone is added or something else is done to the armature that triggers a rebuild.
The fix is to trigger the rebuild in the link operator.
This only fixes this particular issue, the issue of bone selection sync between armature instances
is not fixed, but tracked in this report #117892
Pull Request: https://projects.blender.org/blender/blender/pulls/123743
Retiming versioning used incorrect source for strip length, so produced
inverted strips. Also it tried to scale sound strips by `speed_factor`,
but sound strips did not change length when retimed.
Finally, the original versioning code was in versioning_300.cc file,
even though it was released with 4.0 version.
Pull Request: https://projects.blender.org/blender/blender/pulls/123144
With the latest fixes of #124061, the `pipelines.volume.is_enabled`
predicate is now obsolete.
Replace it by `!current_objects_.is_empty()`.
Fixes#124159
For text objects, current code will use another font if you ask for
a character that is not found in the selected font. But what if the
selected font is invalid? This can happen with a saved Blend that uses
a non-packed font that is since deleted. Current behavior will show
nothing. This PR restores earlier behavior where we use the built-in
font in this case. This does not make any changes to error reporting.
Pull Request: https://projects.blender.org/blender/blender/pulls/124184
This commit does multiple things:
- It checks `is_behind_occluder` in all non-extrapolated
cases. This avoid false positive hit caused by tracing
basis X axis being almost parallel to the light.
- It checks time only on the last sample. This avoid
light leaking when the ray poke through some solid
objects because of raymarching steps.
- Add bias to `is_behind_occluder` to avoid
precision issues when ray is parallel to the light
direction.
This reduces noisy area lights but there are still issue
very close to them.
Fixes#123588
Regression from fb8bf8b0c8: Now that the 'automatic override'
relationships between IDs are much more strict, re-creating
sub-hierarchies of overrides as part of the partial resync may not be
able to 'find' all previously used overrides anymore. An additional
remapping from linked (reference) data to existing, unchanged 'old'
overrides is needed.
In the report case, the root of the partial resync is the Curve obdata
itself, which will imply resync of its material. But the Armature object
used as dependency of the material (through the driver target) has
already been checked as not needing resync, so it is not included as
part of this partial resync sub-hierarchy anymore.
However, it means that the newly resync material override will have a
pointer to the linked armature object, which then needs to be remapped
to the matching existing (unchanged) override.
NOTE: this likely means that too many override IDs are still being
resynced currently. In the report-s case e.g., the Material ID likely
does not need to be resynced either.
This is a follow up to #124054, which made all selected bones get their
location keyed after a targetless IK transform. That solved an issue
where auto-keying was failing to insert location keys on bones that
needed them, but in the process that fix also inserts location keys on
many bones that *don't* need it.
This commit addresses one of the simpler cases of unneccesary location
keys: bones that are connected to their parents. If a bone is connected,
its location will no longer get keyed.
Pull Request: https://projects.blender.org/blender/blender/pulls/124145
- Do not translate a label containing the name of the active NLA
action.
- Translate default action name when created by inserting a keyframe.
- Translate "<NoAction>" and other default NLA strip names.
- Translate "<NoAction>" displayed in the UI when no action exists in
the NLA.
- Translate the temporary meta-strip created when moving an NLA strip
around. This uses DATA_() for consistency, even though it is not
really user data.
Issues reported by Gabriel Gazzán.
Pull Request: https://projects.blender.org/blender/blender/pulls/124113
- Operator descriptions use tip_() since they will be displayed in
tooltips.
- Extension messages:
- Split "(Add-on|Theme) \"{:s}\" already installed!" into two
messages.
- Use rpt_() to translate error messages.
- Restore core add-on name and description translation.
- Use DATA_ to translate paint material slot name, so that translation
happens only if the user enabled it for user-created data.
- Node Wrangler contains functions used to build operators' poll
methods. This change allows them to be properly translated by using
str.format() instead of f-strings, and explicit extraction with
tip_().
Pull Request: https://projects.blender.org/blender/blender/pulls/123795
Code that implemented "show missing media in magenta" in VSE preview
only implemented that for images and movies, and forgot that scene
strips can also be missing.
Pull Request: https://projects.blender.org/blender/blender/pulls/124124
This is a follow-up to #123998 which fixed autokeying on objects when
the "Only Insert Needed" user preference was enabled. This fixes an
essentially identical bug for auto-keying bones, but *also* fixes a
different-but-related bug unique to bone keying.
In the first bug the location channel would erroneously not get keyed
sometimes when the "Only Insert Needed" user preference was enabled.
This fixes that by passing in whether more than one bone is being keyed
or not, and using that to accurately determine if location actually
needs to be keyed.
In the second bug, the location channel would erroneously not get keyed
when "Auto IK" was used to transform bones. This fixes that by just
being over-conservative. This unforunately leads to more bones getting
their location channel keyed than needed when Auto IK is used, but this
is better than *failing* to key needed channels, which was the previous
behavior. I left a comment explaining the situation, along with a TODO
to address this in a more satisfying way in the future.
Pull Request: https://projects.blender.org/blender/blender/pulls/124054
The issue was that the wire width was not passed to the shader
when the object was of type curve.
This resulted in the bone wire thicker than expected and
not responding to the parameter.
Pull Request: https://projects.blender.org/blender/blender/pulls/124053
For text objects, current code will use another font if you ask for
a character that is not found in the selected font. But what if the
selected font is invalid? This can happen with a saved Blend that uses
a non-packed font that is since deleted. Current behavior will show
nothing. This PR restores earlier behavior where we use the built-in
font in this case.
Pull Request: https://projects.blender.org/blender/blender/pulls/122851
Use the correct property names in the up and forward axis update
callbacks. Otherwise they don't work as intended and the following will
be traced to the console:
```
RNA_enum_get: WM_OT_usd_export.forward_axis not found.
RNA_enum_get: WM_OT_usd_export.up_axis not found.
```
Pull Request: https://projects.blender.org/blender/blender/pulls/124112
Added a new custom_properties_namespace USD export option, to
allow replacing or omitting the current default "userProperties"
namespace prefix.
Note that this option does not apply to names that already have a prefix
(e.g., it would apply to name "bar" but not "foo:bar"). It also does not apply
to the internal Blender "object_name" and "data_name" properties which
always have the prefix "userProperties:blender".
Also added logic to handle ":" namespace delimiters in property names.
Pull Request: https://projects.blender.org/blender/blender/pulls/124067
This was caused by the effective radius of some
components were below the first sample radius.
This meant that the accumulation resulted in zero
weight for these components.
This patch introduce a minimum radius that can
be represented depending on the sample closest
to the center pixel. This makes sure that we
have at least one sample that will have
non-zero weight.
Fix#124031
Code was using seq->start + seq->len to figure out where the "right
side of content" is, seemingly since forever (i.e. since 997b5fe4
added strip thumbnails). But that is incorrect starting with 3.3,
where seq->len is "number of frames / samples in source media",
but that does not necessarily match number of frames in the timeline.
For movies this happens when media framerate does not match
sequencer framerate.
Use SEQ_time_content_end_frame_get instead to calculate where the
content ends.
Pull Request: https://projects.blender.org/blender/blender/pulls/124081
Only apply EEVEE-Next material shadow versioning to materials with
surfaces. Otherwise this will break the rendering of some materials
(E.g. Volumetrics) by making the surface opaque to the camera.
Pull Request: https://projects.blender.org/blender/blender/pulls/124076
This patch allows the size of the Glare node to be below 6. This is not
really a fix, but it is targeting the release branch because of concerns
about this limitation which din't exist for old EEVEE bloom.
Pull Request: https://projects.blender.org/blender/blender/pulls/124092
The Bloom mode of the Glare node crashes if the input image is too
small. This is because bloom is computed using a down-sampling followed
by an up-sampling chain, and if the user supplied size is not maximum,
the computed chain length might be zero or negative, which is not
handled gracefully. To fix this, we just sanitize the chain length.
Pull Request: https://projects.blender.org/blender/blender/pulls/124089
- Report when a repository is added since it's not so clear from the
popup that a new repository has been added.
- Fix for missing redraw after selecting all/none tags.
- Also access the tags property once instead of getting it for each
button.
