GitLab

First issue is that a tolerance squared was being passed
into a function needing distance. Needed a sqrt.
The second issue is that the default computed tolerance for
the proble filter was absurd and wasn't computed at all
just hardcoded to max double which causes numerous overflows
and incorrect behavior. Changed the logic to look at some
actual cells and compute a tolerance from them.

Point arrays were being used when cell attributes were generated.
Fixed that and updated some tests.

Memory allocation error when a contour value resulted in no output
primitives, and when the contour was just one of many contours.

vtkTubeFilter removed non-unique points from cells of its input polydata. This data corruption caused issues for example when the same polydata was used as input by various filters.

Fixed by making a copy of point indices before removing non-unique points.

This filter wasn't streaming because the input's whole extent was
updated before vtkImageMarchingCubes::RequestData() was called.
In order for streaming to work properly, the input data must not
be updated before RequestData() updates it.  To state this in
another way, setting a memory limit for the input has no effect
if the whole input is filled before vtkImageMarchingCubes even
executes.

To properly test this filter, the test was modified to use a
streaming reader (vtkImageReader2 instead of vtkVolume16Reader),
and the InputMemoryLimit was reduced.

minro fix

This brings in VTK commit 67ec1f63 and subsequent changes
to the appendfilter over to the VTK branch that ParaView's
release branch uses in order to fix the merge blocks
filter in ParaView 5.6.x.

See paraview/paraview#18991

These changes address issue #17597.

This implements a SMP version of the LagrangianParticleTracker in order to increase performance.
It make the whole algorithm thread safe and uses multiple vtkSMP::For and mutexes instead of reduction

This adds a new UserData parameters to FunctionValues
so any type of data can be passed down.

Runtime warnings.

There were replaced by:

vtkParticleTracerBase, vtkParticleTracer, vtkParticlePathFilter,
and vtkStreaklineFilter
and their parallel versions.

See
https://blog.kitware.com/improvements-in-path-tracing-in-vtk/

This is a mess because it seems like the pipeline created
has no indication of what nodes are what level in the
tree. So we end up having to iterate over the last
filters input to count non lowest level nodes and
then use that count when iterating over the mappers
input.

minor fix

They were drawing thee layers of overlapping data
Depending on pointer values the test could fail
due to which level was drawn last.

Similar to PolyLineSource which was made a subclass of this.
Useful to have a gui driven point source with explcit points.

Add a check whether a data object can be processed by
vtkContour3DLinearGrid when contouring by the named array. Conditions
are:

* all cells are linear
* the contour array is a supported type

Added some tests of this function.

These warnings are for input state that does not affect the
correctness of the filter output, so convert them to vtkLog
informational messages.

Fixing errors within vtkCellValidator that caused a wrong state mask
evaluation, returning wrong error messages in case of errors.
Changing the IsConvex method of vtkPolyhedron to use local instead
of global face ids to access points. Global face ids resulted in accessing
the wrong points causing erroneous face centroids and normals.