Fixes a few render synchronization soundness issues I discovered when investigating
why resetting the pools after the submission fence in #137305 ended up causing #137395.
1. Ensure the number of submission fences in GHOST_ContextVK is the same
as the number of resource pools in VKThreadData. Otherwise, the
fences might get misaligned, making it difficult to predict which
submission a fence's signal corresponds to.
2. In swapBuffers, pass the current m_frame_data's fence and semaphores
to the Vulkan backend callbacks, rather than the following
m_frame_data. This fixes an observed soundness issue where the fences
were off-by-one. Now, the backend callbacks can be sure that both the
next frame is ready for construction (and it's resources can be
cleaned), and will use the correctly aligned fences and semaphores
during command buffer submission.
3. Do not recreate the m_frame_data's fences during recreateSwapchain.
This would lead to unsoundness immediately following
recreateSwapchain where all the fences are signaled but those frames
might still be in flight.
Pull Request: https://projects.blender.org/blender/blender/pulls/137580
- Update to latest ufbx version that adds support for FullWeights
- Handle that in the same way as the Python importer did
- Add test files from ufbx test suite
Pull Request: https://projects.blender.org/blender/blender/pulls/138811
Similar to 93be6baa9c.
It doesn't make sense to store the type as part of the request
since we upload any generic attribute, and the vertex buffer
type is just chosen depending on the attribute's type in the
geometry anyway.
The code in `mesh_cd_calc_used_gpu_layers` is unfortunately
still way too complicated to remove the custom data layer lookup,
but this gets us one step closer.
Pull Request: https://projects.blender.org/blender/blender/pulls/138791
When using motion blur GPU materials and its resources can be freed when
still in use. This fix adds a workaround to store these resources
temporarily in a render discard pile. When rendering is finished (or
between frames) the resources are moved to the regular discard pile.
Pull Request: https://projects.blender.org/blender/blender/pulls/138809
There are many RNA properties on the Mesh ID that were animatable, but
likely shouldn't be. For example:
- The vertex indices of edges and faces.
- The X/Y/Z mirroring toggles (effectively tool settings).
This PR marks these properties as non-animatable. The downside of doing
this is that it reduces opportunities for pranks, such as preventing
your coworkers from being able to toggle X-mirror. But that seems like
an acceptable price to pay for keeping this from happening accidentally,
and in some cases perhaps indirectly causing more serious issues.
Note that there are many, many more erroneously animatable properties
throughout Blender. This PR narrowly addresses just those that are part
of the Mesh ID type.
Pull Request: https://projects.blender.org/blender/blender/pulls/138311
Due to an oversight in the pipeline pool/shader, pipelines that were
invalid (as subresources where destroyed) could still be selected for
reused. This happened more often when using the compositor as it
recreates shaders a lot. It also depended on how handles are reuses by
driver implementation.
This is fixed by discarded any associated pipeline when the pipeline
layout is discarded.
Pull Request: https://projects.blender.org/blender/blender/pulls/138800
Instead of calling all of these functions from `RNA_def_nodetree`, call them
from their respective node callbacks. This didn't work a few months ago, because
these callbacks did not have the `brna` argument back then.
Pull Request: https://projects.blender.org/blender/blender/pulls/138798
This patch adds support for creating Group Input nodes by dragging input sockets
and panels from the group interface tree-view in the side bar. This
significantly simplifies creating a Group Input node for a specific input when
there are lots of them.
These cases are supported:
* Dragging an input socket creates a Group Input node just for that socket.
* Dragging a panel without panel-toggle creates a Group Input node containing
all sockets in the panel.
* Dragging a panel with panel-toggle creates a Group Input for the panel-toggle.
* Dragging a panel with panel-toggle while holding Ctrl creates a Group Input
with the panel-toggle and all sockets in the panel.
This is supported in all node tree types.
Pull Request: https://projects.blender.org/blender/blender/pulls/137739
Previously, whenever a node had a non-trivial storage struct, there would have
to be code for it in `node.cc`. Now there is a general callback that new node
types can use to implement their blend read/write behavior.
Some existing nodes were converted to use this decentralized method. However,
some older nodes can't use it in the same way, because the node types were
introduced before there were node idnames. It's also somewhat hard to reason
about special cases that versioning code might have for these nodes, so they
remain unchanged.
The node callback only writes the non-trivial data, while the main node storage
struct is written automatically by relying on `bNodeType::storagename`. This
simplifies the callback in many cases or makes it unnecessary for trivial types.
Some nodes have specific handling for forward-compatibility. This
forward-compatibility code is kept in `node.cc` for now, because it also affects
the main storage struct and therefore has to be changed before that struct is
written.
Pull Request: https://projects.blender.org/blender/blender/pulls/138722
Removes an inconsistency where sometimes `bke::node_label` is used and sometimes
`node.label` to get the frame label.
Consolidates several duplicated formulas for computing the frame layout into a
single function.
Pull Request: https://projects.blender.org/blender/blender/pulls/138035
For local struct types "Mesh" was used for both edit-mesh and
object-modes.
- Use "MeshObject" non edit-mesh selection (vertex paint selection).
- Use "Mesh" for edit-mesh types.
This follows existing naming convention: `do_lasso_select_meshobject`
for e.g.
There was one test in the volume render test folder that made use of
the ocean sim modifier to add some detail. This lead to test failures
when building Blender without ocean sim.
To fix this issue, this commit applies the ocean sim modifier. This
means it's no longer used, but the render results are the same,
meaning there is no need to update the reference images.
Pull Request: https://projects.blender.org/blender/blender/pulls/138729
When building Blender WITH_MOD_FLUID=OFF, the OpenVDB render tests
would fail as some of them make use of the WITH_MOD_FLUID features.
This commit fixes this by disabling OpenVDB render tests unless
WITH_MOD_FLUID is active.
Pull Request: https://projects.blender.org/blender/blender/pulls/138728
Previously, it was not possible to switch a menu based on another menu. This
patch adds support for this.
Usually, menu sockets are drawn without the label in nodes currently. Now there
is one exception: the Menu Switch node when it switches another menu. If the
label is not shown, the UI is missing crucial information.
Pull Request: https://projects.blender.org/blender/blender/pulls/138704
Previously, Switch node were using the fallback behavior when they were muted.
That implied that the generated internal link was generally not very useful.
This patch makes the behavior of muted switch nodes explicit. Now the internal
link will always point to the first value input and never to the condition
input. Note, for the Menu Switch node this does not make a difference yet,
because menu sockets are not supported there yet (#138704).
Pull Request: https://projects.blender.org/blender/blender/pulls/138724
Regression in [0] which missed updating the old-buttons data,
causing it to reference freed Main & ID pointers after undo.
Apply the same fix for uiBut::pushed_state_func.
[0]: 3e03576b09
Prior to this commit, the Dyntopo panel was hardcoded to display as
faded if the current brush was the mask brush. This conflicts with the
common set of brushes which we do not support dyntopo for, which
includes many other brushes.
This commit changes the check to better indicate that certain brushes do
not support dyntopo. The scene-level settings can still be changed & the
detail flood fill operator remains functional even if the a different
brush is selected.
Additionally, the panel is no longer greyed out if the "Manual" option is
active.
Pull Request: https://projects.blender.org/blender/blender/pulls/138634
This converts the public `uiItemO` function to an object oriented
API (`uiLayout::op`).
Also this rearranges `idname` paramether, since this the only one
required, and to make format similar to `uiItemFullO`
Note: One of the benefits of moving from a public function to class
method is to reduce API usage difference between C++ and Python. In
Python this method is called `UILayout::operator`, however `operator`
is a reserved keyword in C++.
Part of: #117604
Pull Request: https://projects.blender.org/blender/blender/pulls/138776
Similar to b21cb20eeb.
This time the domain is removed. The idea is that the domain doesn't
change anything about how the attribute is stored on the vertex buffers
so it doesn't make sense as part of the request. If we continue that
logic to also remove the data type, we can avoid searching through the
geometry when creating the requests, instead handling invalid requests
when creating the buffers.
The complexity of the change comes from the fact that the request's
domain was used to determine whether the Curves drawing code needed to
interpolate the attribute to the evaluated points. This is now stored
separately in the curves cache. The change in the sculpt code is also
non-trivial since we delay more of the logic until after we have
looked up the attribute from the geometry.
Pull Request: https://projects.blender.org/blender/blender/pulls/138619
Fix the recently implemented ShaderCompiler::batch_cancel.
Expose it with GPU_shader_batch_cancel and
GPU_shader_specialization_batch_cancel.
Use them in the EEVEE ShaderModule destructor, to prevent blocking on
destruction when there are in-flight compilations.
Pull Request: https://projects.blender.org/blender/blender/pulls/138774
Corrects behavior with NURBS knot values in .obj exporter. Knot values
denoting the curve parameter range and values at the boundary region
in the span ends had hardcoded knot values. It also implemented its own
knot calculation, which is not ideal...
Importer is updated to not try to second guess the knot values.
Not entirely sure what it was trying to do but it used wrong indices
and missed writing the end of the knot vector. Combined the changes
should make it possible to import and export a simple NURBS curve with
custom knots and leaving it intact.
This replaces some of the erronous behavior using functions from [new]
Curves implementation. Mixing new and legacy curve implementation is not
ideal but exporter is exporting POLY curves as NURBS while legacy method
does not support computing the knot vector. To avoid introducing
additional branch cases nor update legacy functions, using the new
functions seems to be the correct choice. These functions should be
functionally equivalent but is not identical (e.g. legacy curve returns
knots in [0, 1] range). It should also make it easier to transition to
exporting new Curves.
Pull Request: https://projects.blender.org/blender/blender/pulls/138732
Pretty bare bones but gets the job done, unlike the gcc
tooling, this will work for release builds, the performance cost
of it is on the high side of things, the full test suite tests take over
an hour for me with code coverage support enabled on a release build.
I have not timed a debug build. Given developers can just run their
tests to get coverage data over what they are working on, I feel this
is still useful tooling to have.
This adds the 3 targets for clang and adds a single gcc target
coverage-reset - this removes the collected code coverage data and
report
coverage-report - This merges the collected data and generates the
report (new for gcc)
coverage-show - This merges the collected data and generates the report
and opens it in the browser
This relies on llvm-cov and llvm-profdata being available if not found
code coverage is disabled.
Note: A full test run requires an obscene amount of disk space, a
complete test run takes about 125GB and takes 12 minutes to merge, so
provision the COMPILER_CODE_COVERAGE_DATA_DIR folder accordingly
Example report in PR
There are very rare use-cases for having multiple Group Output nodes. However,
it's something that's been supported for a long time and removing it may be a
regression for some. In practice, it's usually a mistake when someone has
multiple Group Output nodes.
This patch adds a simple warning on inactive Group Output nodes to help the user
notice this case.
Pull Request: https://projects.blender.org/blender/blender/pulls/138743
