Blender

Recent Commits

• Bart van der Braak (16 Jun 26)

  Build: Remove `libdecor` and warn when Python unsupported - `libdecor` is not used anymore - warn the user when python version is lower than 3.10 Pull Request: https://projects.blender.org/blender/blender/pulls/160227

• Guillermo Venegas (16 Jun 26)

  Cleanup: UI: Sort button flag enum values `BUT_DRAG_LOCK_X` was introduced in https://projects.blender.org/blender/blender/commit/2bbd90ffa0e6feec449c2f94b14ece105cb64731, `21` was the only bit flag available Pull Request: https://projects.blender.org/blender/blender/pulls/160243

• Hans Goudey (16 Jun 26)

  Mesh: Improve and simplify modifier evaluation logic Currently processing meshes and other geometry components through the mesh modifier stack is quite convoluted. For example, we keep removing the mesh from the geometry set and adding it back again. This makes dealing with the more flexible ownership provided by implicit sharing too complex, so we miss some possible performance benefits. This PR changes to passing a geometry set between modifiers. The deform and cage meshes are now stored as "edit hints," similar to curve deform positions which were added more recently. `ID_TAG_COPIED_ON_EVAL_FINAL_RESULT` is removed because the same mesh might be used for the initial copy-on-eval copy and the final result, we don't always copy for the latter. Overall, this allows us to avoid mesh copies more often. This would have been a deeper fix to #119938, and is part of fixing #153195. Pull Request: https://projects.blender.org/blender/blender/pulls/119968

• Miguel Pozo (16 Jun 26)

  Merge branch 'blender-v5.2-release'

• Miguel Pozo (16 Jun 26)

  Fix: Overlay: FlatObjectRef::flat_axis_index_get for handle ranges Don't call `BKE_object_dimensions_get` on duplis. This updates the code to only take object data dimensions into account. We could check individual instances, but it seems too overkill, especially since it's not possible to select individual dupli instances anyway. Related: #160077 Pull Request: https://projects.blender.org/blender/blender/pulls/160167

• Brecht Van Lommel (16 Jun 26)

  Merge branch 'blender-v5.2-release'

• Brecht Van Lommel (16 Jun 26)

  Fix #159577: Cycles can't read velocity volume grid with custom name The code was incorrectly making the grid only available by name or standard, and not always setting the number of channels correctly. Pull Request: https://projects.blender.org/blender/blender/pulls/160185

• Brecht Van Lommel (05 Jun 26)

  Cleanup: Cycles: Clarify path visibility static_assert There was an assert to check it fits in 16 bits which is correct for the related code, but better also check it fits in 8 bits which is how we store it in the integrator state. Pull Request: https://projects.blender.org/blender/blender/pulls/159666

• Brecht Van Lommel (05 Jun 26)

  Fix: Cycles: Incorrect generated coordinates for subdivision This should get the position attribute from the subdivision attributes, not get_position(). This code does not run in the Blender integration. This issue was introduced in 24b981b1cd0. Pull Request: https://projects.blender.org/blender/blender/pulls/159666

• Brecht Van Lommel (05 Jun 26)

  Fix: Cycles: Uninitialized rtc_device_is_sycl and missing space in log The missing space for the log was introduced in 3be313be555. Pull Request: https://projects.blender.org/blender/blender/pulls/159666

• Brecht Van Lommel (05 Jun 26)

  Fix: Cycles: OptiX optional modules issue with runtime compilation This case was not properly handled in the recent refactor. This issue was introduced in c51fcf73a7c. Pull Request: https://projects.blender.org/blender/blender/pulls/159666

• Brecht Van Lommel (05 Jun 26)

  Cleanup: Cycles: Add missing asserts for MNEE state Pull Request: https://projects.blender.org/blender/blender/pulls/159666

• Brecht Van Lommel (05 Jun 26)

  Fix: Cycles: MNEE shade surface kernel incorrectly runs for casters Only receivers need to initiate an MNEE walk. Pull Request: https://projects.blender.org/blender/blender/pulls/159666

• Brecht Van Lommel (05 Jun 26)

  Fix: Cycles: Denoising motion pass does not check view layer filter It was inconsistent with the C++ code. This issue was introduced in 6fef00cace6. Pull Request: https://projects.blender.org/blender/blender/pulls/159666

• Brecht Van Lommel (05 Jun 26)

  Fix: Cycles Hydra issue rendering subdivision meshes Subdivision mesh synchronization did not use the new ATTR_STD_POSITION properly. This issue was introduced in edafe4f5eee. Pull Request: https://projects.blender.org/blender/blender/pulls/159666

• Brecht Van Lommel (05 Jun 26)

  Fix: Cycles: Hydra instancer rotation not cleared across syncs This issue was introduced in 51977de03a3. Pull Request: https://projects.blender.org/blender/blender/pulls/159666

• Brecht Van Lommel (05 Jun 26)

  Fix: Cycles: Build error without OpenColorIO Not a required dependency for Cycles standalone yet, even if it is for Blender. This issue was introduced in 5d2b8f8a529. Pull Request: https://projects.blender.org/blender/blender/pulls/159666

• Brecht Van Lommel (05 Jun 26)

  Cleanup: Cycles: Avoid potentially dangling XYZ-to-scene-linear reference It's not really possible in practice for OpenColorIO configs to switch mid render, but better not to return a reference to something that is mutex protected. This issue was introduced in 5d2b8f8a529. Pull Request: https://projects.blender.org/blender/blender/pulls/159666

• Brecht Van Lommel (05 Jun 26)

  Cleanup: Cycles: Glass BSDF thin film IOR default mismatch This was missed in b02367970113. Pull Request: https://projects.blender.org/blender/blender/pulls/159666

• Sybren A. Stüvel (16 Jun 26)

  Fix: Animation playback loop mode 'Stop at End' only works once The animation playback loop mode 'Stop at End' stops playback at the scene end frame (or the preview range end frame, if it's active). This means that restarting playback from where it stopped (so on the end frame) would immediately stop again. This is now resolved, by detecting this special case and actually restarting playback at the start frame. This takes into account reverse playback, where the meaning of 'start' and 'end' are flipped. Pull Request: https://projects.blender.org/blender/blender/pulls/160236

• Guillermo Venegas (16 Jun 26)

  Refactor: UI: Replace some retval based callbacks with std::function callbacks Continuation from https://projects.blender.org/blender/blender/commit/b16295050befb1a6c3c240e87472bb8c797ae371 Pull Request: https://projects.blender.org/blender/blender/pulls/159646

• Philipp Oeser (16 Jun 26)

  Merge branch 'blender-v5.2-release'

• Philipp Oeser (16 Jun 26)

  Fix #156178: Render Sequence Animation stopped showing total progress Probably caused by the sequence of these: ``` a41e73770008 Fix: Cycles renders not showing progress in compositor and sequencer 4bf5fcafd1a3 Refactor: Sequencer: Pass Render struct in render context 48c507316513 Refactor: Move GPU contexts and display callbacks into new RenderDisplay ``` The progress reporting (how far along in the whole sequencer render job we are) from `do_render_sequencer` was actually intact, it was just constantly overwritten by the frame progress via `RE_engine_update_progress`. (also the reason why it still worked with EEVEE -- EEVEE does not report frame progress...) While debatable which progress we are most interested in, I would think the progress in the sequence is more important, thus avoid calling the frame `RenderDisplay` `progress()` from the render engine when it is shared (aka recursive compositor or sequencer renders). This only affects the progressbar afaict, the Image Editor e.g. is still updating nicely. Pull Request: https://projects.blender.org/blender/blender/pulls/160223

• Philipp Oeser (16 Jun 26)

  Merge branch 'blender-v5.2-release'

• Omar Emara (16 Jun 26)

  Nodes: Track animated properties in node dependencies Node dependencies do not consider animated properties in determining if the node tree depends on time, so this patch does track such animations. Pull Request: https://projects.blender.org/blender/blender/pulls/160207

• Christoph Lendenfeld (16 Jun 26)

  Fix: pose sliding with identically named bones in multiple armatures Oversight of 7017225b2083815626381164274bccc6781bf885 When using pose slide on multiple armatures, the names of pose bones may be identical. The `Set` used bone names to avoid adding double entries but did so by name. The fix is to use the `bPoseChannel *` directly. Pull Request: https://projects.blender.org/blender/blender/pulls/160211

• Brecht Van Lommel (16 Jun 26)

  Merge branch 'blender-v5.2-release'

• Brecht Van Lommel (16 Jun 26)

  Fix #158706: Cycles Metallic BSDF F82 black in glossy pass Always do the table lookup for the albedo in case it's needed for the glossy or transmissions pass even if not needed for the combined pass. This also improves sampling quality for some rough glass, due to better sampling weights. Pull Request: https://projects.blender.org/blender/blender/pulls/158819

• Brecht Van Lommel (15 Jun 26)

  Fix #95995: Freestyle lines missing with many objects Each stroke is placed at its own Z-depth, and there was ineffective compensation for precision issues. This may have made more sense for the Blender Internal z-buffer, but for EEVEE and Cycles this was causing more problems that it solved. The code to increase the z-delta could have just been removed to fix the reported bug, but we improve the precision a bit further by computing the delta by dividing the range by the number of vertices. Pull Request: https://projects.blender.org/blender/blender/pulls/160187

• Brecht Van Lommel (15 Jun 26)

  Fix #85451: Freestyle collection filter ignores nested collections Use object iterator that takes into accound child collections.