For non-trivial custom data types, the undo system did a shallow copy of the
base array which implicitly transferred ownership of the data to the undo
system. Combined with implicit sharing, this was hacky at best, and quite
wrong at worst, since it freed the implicit sharing info incorrectly.
To fix this, free the mesh custom data with the standard function for that
and add the non-trivial layers to the undo state using implicit sharing to
avoid another copy.
Alternative to #123894 and #123884.
Pull Request: https://projects.blender.org/blender/blender/pulls/123991
Remove the filtering of factors from the brush distance calculation
function to simplify its responsibilities and to allow passing a custom
radius from the outside. Factor filtering is now done in a separate
function. Also rename the distance calculation functions.
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
The issue is caused by an external dependency to
/opt/homebrew/opt/libb2/lib/libb2.1.dylib
This change adds a test to catch the same issue in the future,
and moves the arm64 libraries to a newer hash with fixed Python.
Pull Request: https://projects.blender.org/blender/blender/pulls/124096
Cancelling operator with RMB is hardcoded but some wants RMB to invoke
the operation as well. In such case, don't use RMB to cancel operator
and reset the radius, i.e. when both `rc->init_event` and `event->type` is RMB
Pull Request: https://projects.blender.org/blender/blender/pulls/123053
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
Embree device pointer can end up being nullptr even when Embree on GPU is
expected to be used. Previous implementation overlooked this possibility,
leading to a completely silent fallback to the non-Hardware ray-tracing path,
this commit fixes it. We've noticed this as now Embree relies on a driver
component: https://github.com/intel/level-zero-raytracing-support that can
potentially be missing from a system.
Pull Request: https://projects.blender.org/blender/blender/pulls/124085
SYCL runtime currently relies on an internal driver behavior that will
break the driver version string returned by SYCL if it changes:
https://github.com/oneapi-src/unified-runtime/issues/1777
This will be fixed at SYCL runtime level but until we use a new enough
one, we need to add additional verifications to avoid blocking execution
on a driver that will change this internal behavior.
Pull Request: https://projects.blender.org/blender/blender/pulls/124084
As desribed in #90039, we dont have a context object when pinning data,
this was accessed though leading to python errors.
The object was only used for the active state of certain layouts, do to
resolve the errors (and disappearing UI), check if we actually get an
object from context. That means the correct active state of certain
layouts is lost (it could either be always ON or always OFF, PR is using
the former -- since the default visibility is ON, so that would be more
common).
Pull Request: https://projects.blender.org/blender/blender/pulls/124091
Caused by 9d4d1aea98. Not every operator
that uses lasso gesture has `smooth_stroke/radius` properties defined,
they are sculpt tool specific. When using `node link cut` tool, console
is flooded with the "property not found error".
Pull Request: https://projects.blender.org/blender/blender/pulls/124040
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.
Part of #118145.
Multithread the loop over all grid vertices, replace the PBVH vertex
iterator macro, and avoid subtractingt the limit positions in the first
loop, since we do that later anyway.
The `sample_disk` returns samples inside a disk of radius 1
which was spanning 2 pixels. The blackmann-haris filter has
not much energy at the edge of the filter and since we
don't importance sample the filter, we have very low weight
samples at some time steps. Improve this by biasing the
distribution towards the center.
This is a temporary solution until we have proper importance
sample of the filtering function for this case.
Fix#123630
The output of the Texture node in the compositor does not update when
the texture settings or the texture type is adjusted. That's because
texture evaluations are cached, and the cache is not invalidated when
needed.
This happens because we rely on the depsgraph to tag changes to textures
so that the cache is invalidated. The tagging doesn't happen because the
texture ID is actually not part of the depsgraph in some cases. This is
so because the Texture ID property in the node does not update depsgraph
relations when it changed, so the depsgraph is out of sync with the node
tree.
This patch fixes that by invoking a depsgraph relations updates when
ever an ID property used by the compositor is changed.
Pull Request: https://projects.blender.org/blender/blender/pulls/124050
This patch cleanup and refactors the render pipeline compositor render
code to deduplicate code and clarify usage.
The unused this_scene arguemenet was removed, per-node functions were
introduced to simplify loops, C++ Set was used instead of GSet, and
scene change is now detected by any rendered scene in the set.
Pull Request: https://projects.blender.org/blender/blender/pulls/124028
Don't store pointers to `blender::Map` items since, unlike `std::map`,
they're not guaranteed to stay valid.
This resulted in shader binaries being loaded into the wrong
specialization.
Pull Request: https://projects.blender.org/blender/blender/pulls/124051
This is a workaround to allow user to keep working without loss of data
when an issue like #124049 happens.
This commit also expose again the `use_all_linked_data_direct` debug
option, no idea why that one was removed.
