Seems to be a copy-paste error: the first 3 columns
were only compared as float3, not as float4.
Only affected validness of regression tests which might have
missed an actual mismatch between matrices. Likely, all tests
were valid, and this change did not discover failures.
The -Wundef was already added to the GCC compiler, but not
to the Clang compiler.
This allows catching cases when code accesses define variable
which has not been defined yet, for example `#if SOME_VAR` without
having `#define SOME_WAV <value>`.
The exact difference between undef and undef-prefix is not fully
clear, this is just something that seems empirically be needed.
This change discovers access to undefined WITH_METAL in the
GHOST_ContextCGL.mm, which needs to be looked into separately.
`Project Individual Elements` was never supported for `Vert Slide` and
`Edge Slide`, however, albeit erroneously, this option still affected
those operations.
In Blender 3.6 this situation of `Project Individual Elements`
affecting the result, has been "fixed". But users still preferred the
old behavior.
Therefore, instead of falling back to `Snap To Face`, support
`Project Individual Elements` for `Vert Slide` and `Edge Slide`.
This is more like how it worked previously.
Needed for the asset shelf context menu to work, see #104831.
Ensures the view item button's context is passed on to the context menu.
Otherwise it cannot display operators relying on this context.
This PR adds support to override the world with the HDRI of the
viewport.
Blurring of the background and camera based rotation isn't supported.
Switching the HDRI would still use the world diffuse light as that is
baked into the irradiance cache.
These issues have been added to #109646
Pull Request: https://projects.blender.org/blender/blender/pulls/109548
Renames `OB_MODE_EDIT_GPENCIL`, `OB_MODE_PAINT_GPENCIL`, `OB_MODE_SCULPT_GPENCIL`, `OB_MODE_WEIGHT_GPENCIL`, `OB_MODE_VERTEX_GPENCIL, and the context modes` to `*_LEGACY`.
Pull Request: https://projects.blender.org/blender/blender/pulls/109648
No user visible changes expected. The feature is not exposed yet.
For the asset shelf (#102879), design is to use a paginated scrolling
style. That means, that any scrolling will always snap to a multiple of
the page size (the size used when pressing the Page Up/Down keys).
Together with strict region size snapping (implemented in the asset
shelf patch, #104831), this gives a clean scrolling experience where
partially visible rows are avoided (impossible even).
Introduces:
- `View2D.flag` value `V2D_SNAP_TO_PAGESIZE_Y`, which will cause any
scrolling to only use multiples of the page size.
- A custom page size via `View2D.page_size_y` for when the full region
size is not the appropriate page size value.
- API function `UI_view2d_offset_y_snap_to_closest_page()` to enforce
the snapping from outside View2D. The asset shelf uses this to keep
strict scrolling over DPI changes.
Pull Request: https://projects.blender.org/blender/blender/pulls/109154
Use octahedral mapping to store probe textures. Octahedral textures are easier to sample
and will increase performance. When extracting the world probe it will first be rendered into
a cubemap. This cubemap will then be remapped using octahedral texture coordinates.
* `CaptureView` captures the world light into a cubemap.
* `CaptureView` triggers the reflection probe module to update the octahedral texture using
the cubemap.
* When sampling reflection probes it will convert the (cubemap) direction to octahedral
coordinate and read from the octahedral texture.
Pull Request: https://projects.blender.org/blender/blender/pulls/109559
Ensure correct SSBO bindings are present for shadow tests.
Metal validation errors occur if SSBO bindings that are expected are
not bound. In this case, we can bind empty SSBOs, but these should
be of the correct type for the tests.
Also adding missing zero-initializations for required members within
LightData. Without these, unit tests fail with various issues including
prevalence of OOB reads.
Co-authored-by: Michael Parkin-White <mparkinwhite@apple.com>
Pull Request: https://projects.blender.org/blender/blender/pulls/109645
Happens with specific GPUs like NVIDIA A100-SXM4-40GB. They use the
compute capability 8.0, which is not explicitly compiled as a cubin,
and since 7fca0ee76a the PTX is not suitable for it either.
The safest solution is to revert the change to a known good state,
and re-iterate as needed.
Revert "Cycles: Increase the compute model for the PTX kernel"
This reverts commit 7fca0ee76a.
This change would need to be cherry-picked to the 3.6 LTS.
Pull Request: https://projects.blender.org/blender/blender/pulls/109636
Previously, a region that used size snapping could be dragged to a
snapped size (say 3 times a column width), but then would be clamped to
an unsnapped size (say 2.7 times a column width) to make it fit the
available space.
Instead clamp the size before snapping, so that snapping respects the
available width/height (resulting in 2 times a column width for
example). Put differently, the region will not be made taller if there's
not enough space to fit the region up to the next snapping point.
Implemented as part of #104831.
Pull Request: https://projects.blender.org/blender/blender/pulls/109027
The Convert Colorspace node uses "Convert Colorspace" as a name on
registration, but it uses "Color Space" as a `DefNode()` in
NOD_static_types.h.
As a result, the node name is not extracted to the .po files, and thus
not translated.
This commit uses the same name in all definitions of the node.
Pull Request: https://projects.blender.org/blender/blender/pulls/109419
Caused by d9f5ce2546 which moved the NLA solo button to the right, but
didnt make the backdrop behind the right-hand-side widgets bigger. So
the first widget seemed "transparent" over the channel name.
Now make the backdrop bigger if the solo button is drawn.
Pull Request: https://projects.blender.org/blender/blender/pulls/109498
Based on code-comments it seems vsnprintf didn't return the un-clamped
string length on MS-Windows, this is no longer the case.
BLI_dynstr used allocation doubling in a loop (with a 65536 limit)
in an attempt to allocate sufficient space.
This workaround isn't needed anymore. Expose BLI_sprintfN_with_buffer &
BLI_vsprintfN_with_buffer functions that take a fixed buffer to avoid
allocation for smaller strings.
This simplifies BLI_dynstr_appendf considerably.
Format the prefix once, then prepend it using a function shared between
BKE_report_prependf & BKE_report_prepend.
Replace BLI_dynstr with BLI_vsprintfN for string formatting.
On my system `int(PIL_check_seconds_timer() * 1000000.0)` always game
the same result because the double value was out of the int range.
Use PIL_check_seconds_timer_i instead (as is done elsewhere).
When the vertex indices are already ordered like curve points, the
attribute arrays can be shared with the result mesh. This reduces
memory and saves time copying the data. The improvement is
larger when the mesh contains more point domain attributes.
The `vert_dest_map` array, which contains a map indicating the index of
destination vertices, was being copied to another array (`WeldVert`)
unnecessarily.
By directly using the `vert_dest_map` array, we achieved a performance
improvement.
The average execution time of different operations was reduced from
267.4ms to 261.3ms, resulting in a 2.3% improvement in overall
performance.
Fix of assert for debug build. Offsets have one last extra
element, used to contain offset for last real element. When
copying values to offsets, the last element has to be ignored.
Pull Request: https://projects.blender.org/blender/blender/pulls/109476
`BKE_mesh_nomain_to_mesh` already copies the "parameters" to the
result mesh, and the result mesh here wasn't an "evaluated" mesh like
the removed function expected.
Fairly recently `OPTYPE_DEPENDS_ON_CURSOR` was added, which makes
the modal callback used to allow clicking on a face set unnecessary.
There is no more instructional text in the status bar, but the changed
mouse cursor should make that clear anyway. Removing the modal
callback makes the behavior more standard, and avoids hardcoding
the operator's keymap.
Pull Request: https://projects.blender.org/blender/blender/pulls/109608
