While testing the RX 5700XT with HIP on Linux, I noticed that under
specific conditions, shadows would not render properly. The specific
conditions being a combination of:
- Curve rendering
- Subsurface Scattering
- Custom mesh light shaders
- Specific GPU and HIP compiler combinations
This specific combination was not present in the render test suite.
This commit adds a heavily simplified verion of a file that is affected
(Cosmos Laundromat demo) to the test suite to test for this specific
issue.
Ref: blender/blender-test-data!41
The Cycles render engine switches between two triangle light sampling
methods depending on the distance to the triangle, and the size of the
triangle.
In a earlier version of Blender, one of the sampling methods was broken
during a refactor. This commit adds a test scene that clearly shows off
both sampling systems so if it breaks again, it will easily
and clearly be shown in the test results.
References:
Original bug report: blender/blender#117771
Fix commit: blender/blender@583af2a6e5
Test repository PR: blender/blender-test-data!42
There are a few simple shader nodes that were missing test files.
This commit adds tests for those nodes:
- Camera data
- Fresnel
- Layer weight
- Hue/Saturation/Value
- Invert
- Normal
- Wavelength
Ref: blender/blender-test-data!39
Unfortunately, the tests as they are right now will likely not work with
4.2 and previous versions, as the brush asset changes will need to be
accounted for.
Pull Request: https://projects.blender.org/blender/blender/pulls/132827
Simply load factory startup file, and recursively iterate over
all (valid) `POINTER` and `COLLECTION` properties of all IDs.
NOTE: A few cases are currently returning invalid rna paths (at least
from quick look), these are skipped for now.
Pull Request: https://projects.blender.org/blender/blender/pulls/132809
This scene consists of two spheres (shade flat and shade smooth)
having their normals baked onto a plane using the "Bake from selected"
option.
Ref: blender/blender#132716
Ref: blender/blender-test-data!37
This commit modifies the `performance/api/sculpt.py` test to add a new
set of tests that run with the multires modifier present. New tests
have `multires_` prepended to the brush name to group them separately
from the existing mesh tests.
Part of #130772
Pull Request: https://projects.blender.org/blender/blender/pulls/132822
The hydra render engine can not render a colour connected directly
to a material output. This was used a lot in the texture coordinate
tests, and as such these tests produced blank results.
This commit adds a emission shader between the colour and the output
so hydra can produce more useful results.
There is no change in the output for Cycles and EEVEE, so those
reference images are not updated.
The rest of blender does handle multi-layer EXR images, using the
"combined" or RGBA/RGB layers when the visual result is needed. Make
VSE do the same.
While fixing this, I found several issues in other not well tested code
and had to fix them:
- IMB_buffer_float_from_float_threaded was wrongly using source channels
as destination channels, producing garbage result.
- IMB_scale_into_new was not assigning channels to destination image.
Pull Request: https://projects.blender.org/blender/blender/pulls/132790
When creating Action Slots for legacy Actions, always name those slots
"Legacy Slot". Before this commit, the slot was named after the ID that
is animated by the Action; this matches what Blender will do when
animating that ID from scratch.
The old versioning behaviour caused issues when dealing with legacy
Actions: if there are multiple Actions in a file (for example multiple
run cycle animations) and only one of those was assigned to the
character (the rest has a fake user), Blender would only name that one
action slot after the character. The "fake user" Actions would just be
get a slot named "Slot".
This causes issues when toggling between the different Actions, as
Blender will not automatically assign a slot when switching from the
character- named one to the "Slot" one.
Ref: #129563, #130261
Pull Request: https://projects.blender.org/blender/blender/pulls/131425
The max length of the RNA property `ActionSlot.identifier` was set
incorrectly. The setter code did manage the length properly, but the
getter was checking agains that incorrect max length, and rightfully
complained.
Pull Request: https://projects.blender.org/blender/blender/pulls/132691
This patch adds support for using integer sockets in compositor nodes.
This involves updating the Result class, node tree compiler, implicit
conversion operation, multi-function procedure operation, shader
operation, and some operations that supports multiple types.
Shader operation internally treats integers as floats, doing conversion
to and from int when reading and writing. That's because the GPUMaterial
compiler doesn't support integers. This is also the same workaround used
by the shader system. Though the GPU module are eyeing adding support
for integers, so we will update the code once they do that.
Domain realization is not yet supported for integer types, but this is
an internal limitation so far, as we do not plan to add nodes that
outputs integers soon. We are not yet sure how realization should happen
with regards to interpolation and we do not have base functions to
sample integer images, that's why I decided to delay its implementation
when it is actually needed.
Pull Request: https://projects.blender.org/blender/blender/pulls/132599
In a recent version of Blender, Cycles and EEVEE gained support for
true sphere lights.
This commit adds tests for these sphere lights inside of volumes.
Ref: blender/blender#123012
Ref: blender/blender-test-data!32
* Use .empty() and .data()
* Use nullptr instead of 0
* No else after return
* Simple class member initialization
* Add override for virtual methods
* Include C++ instead of C headers
* Remove some unused includes
* Use default constructors
* Always use braces
* Consistent names in definition and declaration
* Change typedef to using
Pull Request: https://projects.blender.org/blender/blender/pulls/132361
This commit adds render tests for the outputs of the
texture coordinate node tested on:
- A scaled and rotated mesh
- A instanced scaled and rotated mesh
- A mesh with and without smooth shading
- A mesh with a volumetric shader
- The world background chader
And a few extra objects for specific outputs. Examples including:
- A deformed mesh for the `Generated` output
- View meshes through a reflection for view dependent outputs
- Instanced meshes using the old instancing system to test the
"from instancer" option with the UV and Generated outputs
- Additional objects with extra UV maps to ensure the correct UV map
is loaded
- And more
There are other aspects that we would ideally test
(E.g. Point clouds and hair), but these will be handled by a
separate commit and set of tests.
Ref blender/blender-test-data!30
Pull Request: https://projects.blender.org/blender/blender/pulls/132512
This rescales the whole scene by its root transform to match the same
visual size while not forcing the user to wait for scale to be applied to
each object.
This is requested by studios whose main applications / USD scenes are
in CM, because referencing and payloading scenes from disparate scales
can cause issues at resolution time.
If "Apply Unit Scale Conversion" is unchecked on import, the user now
has the ability to bring the objects in with a scale factor of 1.0, so that the
objects may be edited as if Blender's scene units matches the imported
stage's.
At export time, a "Stage Meters Per Unit" value can be chosen from a list
of common measurements, as well as setting a custom value.
Co-authored-by: kiki <charles@skeletalstudios.com>
Co-authored-by: Michael Kowalski <makowalski@nvidia.com>
Pull Request: https://projects.blender.org/blender/blender/pulls/122804
The issue is that certain RNA paths cannot be generated to come
from an ID and `RNA_path_from_ID_to_property` will return no value.
We are using that function in the keyframing code to allow passing
in a pointer to a bone and a path relative to that bone.
Since there is currently no good way to find the path from the ID to an arbitrary
struct pointer (see #122427), this patch is a workaround that uses
the struct_pointer IF that happens to be an ID pointer.
Of course that still has the core limitation in place but until a
better solution is available on the RNA side this is the best we can do.
Pull Request: https://projects.blender.org/blender/blender/pulls/132552
The reason for this probably was the const nature of the shader data.
However, this is something counter-intuitive, as it potentially leads
to multiple BSDFs re-using the same LCG state.
Pull Request: https://projects.blender.org/blender/blender/pulls/132456
In a previous commit I accidentally added a test that had
these properties:
- High resolution (512x512)
- High sample count (512)
- Had high quality denoising enabled
By reducing the resolution to 128x128 and sample count to 32
and disabling denoising, the test will hopefully be 256x faster.
The goal of this test is to try to import some critical py scripts with the
system python of the building machine.
The main target is to ensure that these py scripts remain usable by all
buildbot machines, as some of them are using fairly outdated python
versions.
Current status:
* Scripts in `build_files` and `docs` are checked.
* Some python scripts in `build_files` were 'reverted' to be compatible
with older required python version currently (3.6).
* A few scripts are excluded from the test, mostly because they use Blender's
`bpy` module, which means they are only intended to be ran with Blender's
python anyway.
* The test is only enabled for Linux buildbots currently, as they use the
oldest Python by far.
Notes:
* Some more scripts are likely to be moved around in the future.
* Whether these tests need to be enabled on windows or macos platforms remains
an open question.
Pull Request: https://projects.blender.org/blender/blender/pulls/130746
The Movie Distortion node output is cropped to the bounds of the
original input. This patch fixes that by extending the bounds depending
on how the distortion extends the image bounds.
Importing USD `quatf` types was erroneously left disabled after recent
work this past summer for better attribute support. It is already
correctly enabled and validated for Export.
The `half` and `double` variants must still be skipped for now though.
Pull Request: https://projects.blender.org/blender/blender/pulls/132252
Import edge crease values and properly configure any added subdivision
modifiers with correct UV and boundary settings. These were already
exported so this now completes our subdivision support.
Pull Request: https://projects.blender.org/blender/blender/pulls/131569
This patch disables compositor tests that rely on anisotropic filtering
for GPU testing. This is done until we make sure they pass universally
by not relying on hardware filtering.
Test files that rely on anisotropic filtering were moved to their own
tests to be able to disable them for GPU only.
This commit implements most features needed for simple text editing.
Active text strip can be edited in preview by pressing tab key, which
enabled text editing mode. With this mode active, outline matches text
boundary box and cursor is drawn.
Cursor can be moved with usual keys. Pressing shift starts selection.
Selection and navigation works when text is scaled or rotated. Mirrored
text is not supported in this PR. it can be done, but the text is
unreadable that way, so I kept it simple.
Multi line text is supported. Pressing return key starts new line.
Copy/paste operator uses OS copy paste buffer, so text from other apps
can be pasted.
Text is still limited to 512 characters. Text string property still
exists in side panel and is limited to single line. Individual
characters can not be styled in different way like in 3D viewport, but
the code is mostly ready for such feature.
Ref: #126547
Pull Request: https://projects.blender.org/blender/blender/pulls/127239
This patch removes the references for the GPU tests and uses the CPU
references for both CPU and GPU, since they are expected to match. This
also unifies the tests scripts into a single script with an argument
for execution device.
This patch changes how transformations are realized by adjusting the
computed size of the new domain after transformation. Previously, this
was computed with the lower left corner of the domain as the origin of
transformation, while now, the center of the domain is used as the
origin. Consequently, domains shrinks/grows around their center, which
results in a more stable output as transforms are animated.
A consequence of this change is that we can no longer scale odd sized
domains to even sized domains or vice versa, since it grows/shrinks by
the same amount on both sides. Supporting this case requires further
investigation and will probably require passing down information to the
realization functions themselves.
This patch implements Font to Curves, Legacy curves to Grease Pencil and
Font to Grease Pencil conversions.
Note that Font to Grease Pencil is done by converting Font to Curves
then converting Curves to Grease Pencil. Currently we do not have
direct conversion APIs.
Part of #131595 and #130518
Pull Request: https://projects.blender.org/blender/blender/pulls/131612
Contains the following changes:
- Uses new `render_layer` images in test repo
- Excludes known broken tests with Hydra Storm from USD 24.05
- packed float and packed half images show corruption
- USD export doesn't support the current light-tree .blend
- Bump the `image_colorspace` and `image_mapping` thresholds to
account for image filtering differences from OS/drivers
Pull Request: https://projects.blender.org/blender/blender/pulls/132044
Both the draw manager and gpu backend used the same compilation
directive for enablement. This PR seperates them into
`WITH_GPU_DRAW_TESTS` for draw manager related tests and
`WITH_GPU_BACKEND_TESTS` for gpu backend related tests.
Pull Request: https://projects.blender.org/blender/blender/pulls/132018
With the HIP-RT BVH on AMD GPUs, instances that have undergone
two sets of transformations will not render properly.
This manifests as:
- Incorrect mesh normals
- Improperly positioned, scaled, or rotated meshes
- Missing intersections
This commit adds a test for this issue to make it easier to test,
and so we can hopefully catch similar issues if we ever add more
BVH options in the future.
Original report: blender/blender#117567
Ref: blender/blender-test-data!29
Pull Request: https://projects.blender.org/blender/blender/pulls/131352
OptiX OSL tests were previously disabled due to a GPU driver bug
resulting in many tests failing unexpectedly.
The new driver version is now out with the fix so we can now enable
OptiX OSL testing.
This commit also updates the OptiX OSL block list with better comments,
and more tests that are known to fail and need investigating.
Ref: #123012
Pull Request: https://projects.blender.org/blender/blender/pulls/129280
Until the newer Hair Curves system can fully replace particle hair, add
a small test to ensure this continues to work.
Since the hair is exported as cubic bspline curves, we can also use this
same file to test bspline import now too.
Pull Request: https://projects.blender.org/blender/blender/pulls/131997
While adding tests I found that metaball export has been broken since
Blender 3.4. It would export each metaball geometry twice.
This looks to have been a side effect of a change to `object_dupli.cc`
which no longer sets the `no_draw` flag for metaballs[1]. With the flag
unset we would end up visiting this particular object twice.
Use a direct check for Metaballs now and add test coverage for the
scenario in general.
[1] eaa87101cd
Pull Request: https://projects.blender.org/blender/blender/pulls/131984
