Adds a SPIR-V cache that skips frontend compilation for shaders
that are already compiled in a previous run of Blender.
Initially this was postponed to 4.4 but it was observed that
the vulkan backend didn't perform well on Windows in debug
builds. The reason is that the compiler would also be a debug
build which makes compiling a shader really slow. Starting
Blender on a debug build could take minutes.
So the decision was made to give this task a higher priority so
the vulkan backend would become more usable to developers
as well.
The cache is stored in the application cache dir. The SPIR-V
binaries can be used by different Blender versions so there
is no version specific cache folder.
**Sidecar**: SPIR-V files are a stream of bytes. There is no
header information that allow us to validate the stream. To
add basic validations we could add our custom header or
a sidecar. It was chosen to use a sidecar as having the SPIR-V
files unmodified allows us to load them directly in
debug tools for analyzing.
**Retention**: Shaders that are not used are automatically
removed with a retention period of 30 days.
**Shader builder**: Shader builder cannot use the SPIR-V
cache as it uses stubs that returns invalid cache directories.
This would load/save the cache to the location where you
started the build.
Pull Request: https://projects.blender.org/blender/blender/pulls/128741
Missed in 0161a19669
Unlike many other brushes, the Layer brush uses the original coordinates
of a given vertex only for the distance test when calculating the factor
and does not use it for other components such as textures.
Note that this commit looks larger than the change would imply, as it
required undoing a recent refactor that is no longer applicable with the
distance changes.
Pull Request: https://projects.blender.org/blender/blender/pulls/128723
We no longer keep the active_vert value in a valid state when the cursor
is not over the mesh. The helper method to access this position can
thus result in incorrect behavior when storing and retrieving this
value.
In this case, we simply avoid setting the `UnifiedPaintSettings` values
related to the last stroke, so that when it is used in
`view3d_navigate.cc` it falls back to the object origin.
Pull Request: https://projects.blender.org/blender/blender/pulls/128716
Fixes a crash caused by a null pointer dereference when the Mesh Filter
tool with Erase Displacement option is used on a mesh that has no
Multiresolution modifier.
Pull Request: https://projects.blender.org/blender/blender/pulls/128632
This fixes#127629. It's still a bit slower than it used to be when there are
lots of instances, but that fixes the main bottleneck that was introduced in
#116582. The issue was that we iterated over all attributes of all instances,
but it should only be necessary to iterate over the instances of each unique
geometry only once.
There is still quite some optimization potential in the Realize Instances code
for the case when realizing lots of instances. Especially the code to gather all
geometries that should be realized can still be made more efficient by reducing
redundant work and using multi-threading.
Pull Request: https://projects.blender.org/blender/blender/pulls/128699
Before version 4.1, the vertex destination during edge sliding for two
independent edges was always the midpoint of the vertices closest to
those edges.
In version 4.1, this behavior was improved to calculate the destination
based on the intersection of the edges. However, when the faces were
coplanar, the behavior would revert to the previous midpoint
calculation instead of using the intersection.
This fix ensures that the vertex destination consistently aligns with
the intersection point, even when the faces are coplanar.
In non-manifold geometry, the original loop could fail to identify the
correct slide direction because it checked for loop equality
(`l_other != l_edge`), which might never be met.
The condition has been revised to check for face equality
(`l_other->f != l_edge->f`), ensuring the loop terminates correctly and
preventing infinite iterations in problematic geometry cases.
Blender crashes when using the GPU compositor sometimes. This is because
compositor render data was accessed before it was updated in the
realtime compositor when detecting compositing device. So fix that by
first updating compositor data before calling any context methods.
During stage load we first look for Prototype prims which are used for
instancing. However, if the instancer itself at the root of that
Prototype hierarchy would later be excluded because of either purpose or
visibility checks, the Prototypes would still be processed. These would
correctly be imported but would end up orphaned because nothing would
later add them to the view layer. Code expecting these objects to be
found within the scene would then fail.
In Animal Logic ALab this situation played out where there was a
`/root/instancer` prim with purpose "render" and we proceeded to load
the prims under the `/root/instancer/Prototypes/...` hierarchy. Since we
were attempting to load just the "proxy" purpose, the `/root/instancer`
was later excluded and the orphaning of those prototypes happened.
The change here moves `collect_point_instancer_proto_paths` to a method
of `USDStageReader` so it can now access `include_by_purpose` and
`include_by_visibility`. And these are now used to prevent unnecessary
Prototype loading.
Pull Request: https://projects.blender.org/blender/blender/pulls/128564
Blender crashes when opening a file that has an interactive compositor
active, while also having a script that invokes the compositor upon file
load.
This is caused by the same system GPU context being active in two
threads at the same time, which happens when the GPU context for the
compositor is created in the main thread, it is made current during
creation, but it is not reset to the main GPU context of the drawable
because it is null. So when the GPU compositor actually executes, it
makes the GPU context current again but in its own thread, causing a
BadAccess error in X11 and potentially other window systems.
The reason why the drawable is nullptr is because it is reset in the
existing window manager when opening a new file while Blender is open,
but it is never initialized for the new window manager. The drawable
info should be moved from the old window manager to the new window
manager in wm_file_read_setup_wm_use_new, but it is preemptively reset
by the wm_window_clear_drawable call before it it is moved. This is done
in wm_file_read_setup_wm_substitute_old_window.
So to fix this, we move wm_window_clear_drawable after the code block
where wm_file_read_setup_wm_substitute_old_window gets called.
Between 0bfd5e3536
and b1cbd9c889
the main branch is incorrectly processing the file
`draw_debug_info.hh` as GLSL and does some string
preprocessing on it. But the output filename matches
the name of the header source file used for compiling
the gpu module. This file not having been updated
since a long time doesn't get copied from the source
folder when switching to other branch and make compilation
fail.
In order to avoid breaking the buildbot longer, we
rename the incriminating file to force recreate it
when building the release branch.
When operator property is tweaked in preferences > keymap > "any operator",
preferences are not tagged dirty, which prevents writing the userpref
file at `WM_exit_ex` when auto-save is enabled.
Pull Request: https://projects.blender.org/blender/blender/pulls/128516
This catches the exception and shows the error message to the user.
We don't have much control over the error message here currently. Better just
report it for now. In the future we could try to detect what error it is exactly
and produce a more user friendly error.
This was already done in GHOST, but not BKE_appdir_folder_home.
Also null check the return value from getpwuid() as it's not
guaranteed to be non-null.
The paths `C:` (on WIN32) and `/` on other systems was detecting as
relative. This meant `BLI_path_abs_from_cwd` would make both paths
CWD relative. Also remove duplicate call to BLI_path_is_unc.
Regression in [0] which didn't account for entering paths in the
file selector which would create paths without confirming,
warning that the "confirm" property was missing.
Entering `*.*` would create `_._` for e.g in the users CWD for example.
Ref !128568
[0]: 6dd0f6627e.
This is due to wrong `is_pad` is passed as argument. It should be false
for gizmo tooltips. For menus when fields size is zero i.e. no
element is appended yet, the python message will be at the top of
tooltip region.
Pull Request: https://projects.blender.org/blender/blender/pulls/128243
This renames the mode identifiers to be consistent with e.g. the context mode identifiers and other names used for the new Grease Pencil.
For `object.mode`:
* `PAINT_GPENCIL` -> `PAINT_GREASE_PENCIL`
* `SCULPT_GPENCIL` -> `SCULPT_GREASE_PENCIL`
* `VERTEX_GPENCIL` -> `VERTEX_GREASE_PENCIL`
* `WEIGHT_GPENCIL` -> `WEIGHT_GREASE_PENCIL`
For the internal `ob->mode` flag:
* `OB_MODE_PAINT_GPENCIL_LEGACY` -> `OB_MODE_PAINT_GREASE_PENCIL`
* `OB_MODE_SCULPT_GPENCIL_LEGACY` -> `OB_MODE_SCULPT_GREASE_PENCIL`
* `OB_MODE_VERTEX_GPENCIL_LEGACY` -> `OB_MODE_VERTEX_GREASE_PENCIL`
* `OB_MODE_WEIGHT_GPENCIL_LEGACY` -> `OB_MODE_WEIGHT_GREASE_PENCIL`
Resolves#127374.
Pull Request: https://projects.blender.org/blender/blender/pulls/128604
The `paint.brush_colors_flip` option was not exposed in the vertex color
panel.
This adds the toggle in the panel and also adds the keymap `X` to flip
the colors.
The tint color and panel was not used consistently.
In Vertex Paint mode we're using the unified paint settings,
but in Draw mode, we only use the brush color for tinting.
This fixes the issue by using the unified paint settings
for all the uses of the vertex color.
Enabling render cropping while using the GPU compositor can cause the
render to be distorted in certain resolution and border configurations.
This is due to a difference in how the compositor and the render
pipeline compute the effective bordered size of the render. While
mathematically identical, difference in rounding can cause off by one
errors in the computed size. The render pipeline computes the integer
bounds of the border then computes the size from that, while the
compositor multiplies the float size of the border to the full size of
the render.
To fix this, we adjust the BKE_render_resolution function to compute the
border size using integer bounds like the render pipeline, which is the
function used by the compositor. The cropped version of that function is
used in two other places in the image editor for displaying render
results, so we should expect no difference in outputs with minimal but
more accurate difference in display.
Pull Request: https://projects.blender.org/blender/blender/pulls/128576
When attaching a layered image with offset, the size of the attached
layers should be decreased. Otherwise an image view is created that can
access incorrect data.
Pull Request: https://projects.blender.org/blender/blender/pulls/128583
Activating render region while using the CPU compositor produces corrupt
output for areas outside of the region when using the File Output node.
That's because the Full Frame compositor ignored nodes' render_border
flag, so the areas of interest of nodes that didn't consider render
border like the File Output were corrupt, producing uninitialized
outputs.
To fix this, we just consider the render_border flag when determining
output areas for output nodes in the Full Frame compositor.
Pull Request: https://projects.blender.org/blender/blender/pulls/128546
The Legacy Cryptomatte node doesn't work in GPU execution mode if
Precision is set to Auto. That's because the colors picked from the Pick
layer might be in half precision and thus will not match the colors in
the Cryptomatte layers. This is due to the compositor using the
context's precision for Viewer outputs as opposed to the precision of
the image that actually needs to be viewed in the Viewer node.
To fix this, we set the Viewer node precision to be the precision of its
input, that way, the Cryptomatte pick layer will be output in full
precision as intended.
Pull Request: https://projects.blender.org/blender/blender/pulls/128495
When opeing the 2D Animation template, there was a UI error:
`AttributeError: 'Context' object has not attribute 'grease_pencil'`
This was because the context function `grease_pencil` was missing.
The fix adds this function.
Pull Request: https://projects.blender.org/blender/blender/pulls/128580
There was a bug where vertex colors were not shown in solid view in
draw or vertex paint mode. The check to use `V3D_SHADING_VERTEX_COLOR`
was outdated and checked the wrong mode flags.
The fix makes sure that we use `V3D_SHADING_VERTEX_COLOR` in
draw and vertex paint mode.
Pull Request: https://projects.blender.org/blender/blender/pulls/128581
Removes many of the operators, panels, and menus used exclusively by Grease Pencil v2 that are no longer needed in v3.
No functional changes are expected.
Some operators are still used by the annotations system and have to be kept around. These may be renamed in future.
Pull Request: https://projects.blender.org/blender/blender/pulls/128521
The armature shape drawing was broken on NVIDIA. The reason is that not
all the elements in of the geometry shader out stages are written to. On
NVIDIA platforms this leads to not storing all the output data.
This could be that geometry shader on those plaforms write directly to
the final location in GPU memory. Other platforms might keep the data in
a local registry and copy it to the final location when emitting the
vertex.
Blender 4.4 might not have this issue as overlay next will remove the
need of the geometry shader.
From Khronos GLSL spec:
```
GS code writes all of the output values for a vertex, then calls EmitVertex().
This tells the system to write those output values to where ever it is that
output vertices get written. After calling this function, all output variables
contain undefined values. So you will need to write to them all again before
emitting the next vertex (if there is a next vertex).
```
Pull Request: https://projects.blender.org/blender/blender/pulls/128578
After feedback from the brush assets, we want to not only set the
default visibility, but also remember the visibility per mode. This way
we can keep the shelf hidden in modes where there's only one bundled
brush per available tool. While having these tools support brushes can
be useful, we don't expect to bundle more brushes soon. So to avoid a
mostly empy asset shelf for a single brush, allow keeping the shelf
hidden in these modes unless the user expands it. The following commit
will set the default visibility for these modes.
More strictly speaking this is entirely context sensitive, not just per
mode, but with current shelf types is practically per mode.
