David Gobbi authored 9 months ago and Kitware Robot committed 9 months ago

5ba1dcf0 Use double-precision for polygon normal again
2bd8a82d Make templated cross safe for memory overlap

Acked-by: Kitware Robot <kwrobot@kitware.com>
Tested-by: buildbot <buildbot@kitware.com>
Reviewed-by: Yohann Bearzi (Kitware) <yohann.bearzi@kitware.com>
Merge-request: !11237

VTK has always used a double-precision cross product to compute
normals, ever since VTK 4.4, but VTK 9.3.0 changed one of the
code paths to use single-precision cross product, which can lead
to overflow for large polygons and underflow for small polygons.
This commit restores the cross product precision back to double.

This allows the input and output to use the same memory space,
which is consistent with the other vtkMath::Cross implementations.

David Gobbi authored 9 months ago and Kitware Robot committed 9 months ago

136e4675 Numpy v2 takes additional __array_wrap__ parameter

Acked-by: Kitware Robot <kwrobot@kitware.com>
Tested-by: buildbot <buildbot@kitware.com>
Reviewed-by: Ben Boeckel <ben.boeckel@kitware.com>
Acked-by: Nicolas Vuaille <nicolas.vuaille@kitware.com>
Merge-request: !11238

Without the return_scalar parameter, numpy 2.x reports a deprecation
warning when numpy_interface is used with ufuncs.

Lucas Givord authored 9 months ago and Kitware Robot committed 9 months ago

53f68252 use vtkIdType instead of long
5e30daac must avoid overflow for large particle counts. Use long instead of int.
3ffdfcff Fix crash in vtkImageReslice
31c41ab5 Fix crash in vtkPointLocator when number of points is changed to zero
41b19147 Update obj_with_material.obj file to test optional values and comments.
ede5302c Handle optional last value on `v` and `vt` command
120252d4 Discard leading blankspaces instead of whitespaces
49784eb8 vtkThreadedCallbackQueue, fixing race condition
...

Acked-by: Kitware Robot <kwrobot@kitware.com>
Tested-by: buildbot <buildbot@kitware.com>
Reviewed-by: Nicolas Vuaille <nicolas.vuaille@kitware.com>
Reviewed-by: Mathieu Westphal (Kitware) <mathieu.westphal@kitware.com>
Merge-request: !11197

Jaswant Panchumarti (Kitware) authored 9 months ago and Kitware Robot committed 9 months ago

0898fbbc Fix leading newlines in shader code
3bf1f197 Fixed various clang -Wmissing-variable-declarations warnings
096e7811 Remove globals that use `vtkNew<>`.
5c77cca3 OpenGL2: Use #ifdef instead of #if to fix compiler error
53d1e831 Disable linear filtering for depth textures in GLES 3.0 emscripten
7a0aa2f0 Fix point size bug in charts with GLES 3.0
a4e597d6 Make PrintSelf output useful member variables in Context2D classes
0a3b6606 Add null check in vtkContextScene::SetTransform
...

Acked-by: Kitware Robot <kwrobot@kitware.com>
Acked-by: buildbot <buildbot@kitware.com>
Merge-request: !11196

- The first line of GLES3.0 shaders must be #version 300es

(cherry picked from commit 077fa791)

Sean McBride authored 1 year ago and Jaswant Panchumarti (Kitware) committed 9 months ago

(cherry picked from commit b6860a71)

David Thompson authored 1 year ago and Jaswant Panchumarti (Kitware) committed 9 months ago

… and make SMP test output more verbose.
This prevents crashes in static builds since the order
of initialization of these objects is not guaranteed.

(cherry picked from commit 2c288f41)

(cherry picked from commit 8d73768a)

- closes vtk/vtk#19345
- makes vtkEDLShading work in emscripten

(cherry picked from commit 52666ffc)

(cherry picked from commit b8388a66)

(cherry picked from commit 095135e1)

(cherry picked from commit 1e507b9b)

- closes vtk/vtk#19347

(cherry picked from commit 4ab29c9b)

(cherry picked from commit 98657d84)

(cherry picked from commit 1e76890c)

(cherry picked from commit 6dc5382e)

- closes vtk/vtk#19078
- It was cast to float because `mod` only worked with floats.
- However, the % operator accepts integers and it is
 available on webgl2 when `#version 300 es` is used.

(cherry picked from commit e919cb88)

Sankhesh Jhaveri authored 1 year ago and Jaswant Panchumarti (Kitware) committed 9 months ago

(cherry picked from commit 37cad402)

Sankhesh Jhaveri authored 1 year ago and Jaswant Panchumarti (Kitware) committed 9 months ago

Render surface with edges uses a geometry shader to draw a triangle
strip at the edges of polygons. Apart from calculating the edge
equation, it was offsetting the vertex positions to accommodate wide
lines. However, this would lead to overlapping inner edges. This change
removes this offset.

(cherry picked from commit 2fe546cc)

(cherry picked from commit 3bd2e838)

(cherry picked from commit 902919df)

(cherry picked from commit 02d9ff6f)

- this ensures the render passes are available on the actor's property keys and execute when rendering

(cherry picked from commit 47fcfae7)

Ben Boeckel authored 9 months ago and Kitware Robot committed 9 months ago

87bf5f9c gitlab-ci: wait 4 hours to start merged pipeline jobs

Acked-by: Kitware Robot <kwrobot@kitware.com>
Rejected-by: buildbot <buildbot@kitware.com>
Reviewed-by: Mathieu Westphal (Kitware) <mathieu.westphal@kitware.com>
Reviewed-by: Cory Quammen <cory.quammen@kitware.com>
Merge-request: !11205

This should help alleviate pressure on the CI infrastructure.

David Gobbi authored 9 months ago and Kitware Robot committed 9 months ago

75af395b Fix repeated render in vtkCocoaRenderWindow

Acked-by: Kitware Robot <kwrobot@kitware.com>
Tested-by: buildbot <buildbot@kitware.com>
Reviewed-by: Sean McBride <sean@rogue-research.com>
Reviewed-by: Mathieu Westphal (Kitware) <mathieu.westphal@kitware.com>
Merge-request: !11194

Jean M. Favre authored 10 months ago and Lucas Givord committed 10 months ago

(cherry picked from commit 8e6bb4eb)

Jean M. Favre authored 10 months ago and Lucas Givord committed 10 months ago

(cherry picked from commit 0cd5314e)

Andras Lasso authored 10 months ago and Lucas Givord committed 10 months ago

When an input vtkImageData did not contain point scalars (e.g., because it contains tensors, or no point data at all) then vtkImageReslice crashed.
It is not a problem if the filter only processes scalar data, but it should never crash, for any kind of input data.

(cherry picked from commit 064d36d5)

Sunderlandkyl authored 10 months ago and Lucas Givord committed 10 months ago

When the number of points in the input dataset was set to zero, BuildLocator would be called but the HashTable would not be invalidated.
The FindClosestPoint methods would continue on with the invalid HashTable, resulting in a crash when invalid point IDs were accessed.

Fixed by calling FreeSearchStructure before returning from BuildLocatorInternal so that the HashTable is invalidated.
FreeSearchStructure now also re-initializes variables that are calculated during BuildLocatorInternal so that Bounds/H/Divisions will be consistent for an empty dataset.
Removed vtkErrorMacro when the dataset is empty or has no points since it is a valid input.

(cherry picked from commit c14d87f8)

Alexy Pellegrini authored 10 months ago and Lucas Givord committed 10 months ago

(cherry picked from commit 26ea1052)

Alexy Pellegrini authored 10 months ago and Lucas Givord committed 10 months ago

They are now checked but ignored.

(cherry picked from commit d2110e04)

Alexy Pellegrini authored 10 months ago and Lucas Givord committed 10 months ago

This fixes an issue when the first comment ended with an empty comment:

```
# like this
#
...
```

(cherry picked from commit 6979a976)

Yohann Bearzi (Kitware) authored 10 months ago and Lucas Givord committed 10 months ago

The boolean member `Destroying` needs to be atomic because there is one
location in the code that uses it without mutex
protectionThreadedCallbackQueue, fixing race condition

The boolean member `Destroying` needs to be atomic because there is one
location in the code that uses it without mutex protection

(cherry picked from commit 0cc57f26)

Sankhesh Jhaveri authored 10 months ago and Lucas Givord committed 10 months ago

(cherry picked from commit 2a01ca8d)