- This commit is a first step to getting the widget to deserialize
correctly because during deserialization a different renderer
is assigned as the default renderer.
- When another renderer is provided, this commit carefully removes
the previous reorientation observer. If the widget was previously
enabled, it replaces the older renderer in the render window with the
new renderer. Otherwise, upon deserialization, the render window had
an extra renderer which originated from the constructor of
`vtkCameraOrientationWidget``.
- Updated baseline images which used the camera orientation widget
  because the widget now resizes itself to correct proportions

It was not producing the correct point IDs for meshes with polyhedrons only.
Also fast path was added when no polyhedrons are present. Additionally,
documentation was added about why copying happens, and the code was simplified.

Mickael PHILIT authored 5 months ago and Mathieu Westphal (Kitware) committed 4 months ago

(cherry picked from commit c0b29bb1)

Mickael PHILIT authored 5 months ago and Mathieu Westphal (Kitware) committed 4 months ago

(cherry picked from commit a7bdf2ee)

Jean Fechter authored 6 months ago and Mathieu Westphal (Kitware) committed 4 months ago

(cherry picked from commit d9ae0820)

Jean Fechter authored 6 months ago and Mathieu Westphal (Kitware) committed 4 months ago

The wrong copy constructor of Node was called, taking the ownership of data it should not.
Therefore, the data was freed too early.

(cherry picked from commit 7bd1f0b3)

Mickael PHILIT authored 5 months ago and Mathieu Westphal (Kitware) committed 4 months ago

(cherry picked from commit 5c4687d6)

Ben Boeckel authored 5 months ago and Mathieu Westphal (Kitware) committed 4 months ago

The header has a method which returns a deprecated type. MSVC warns
about such usage, so suppress it.

(cherry picked from commit 700f5d1e)

Ben Boeckel authored 5 months ago and Mathieu Westphal (Kitware) committed 4 months ago

(cherry picked from commit 35c93c98)

Ben Boeckel authored 5 months ago and Mathieu Westphal (Kitware) committed 4 months ago

(cherry picked from commit 230a3254)

Ben Boeckel authored 5 months ago and Mathieu Westphal (Kitware) committed 4 months ago

The macro goes after the `struct` keyword.

(cherry picked from commit cf3e8667)

Ben Boeckel authored 5 months ago and Mathieu Westphal (Kitware) committed 4 months ago

(cherry picked from commit 6fea24e9)

Previously, only float values were accepted as time values, preventing
"time" or "TimeValue" nodes with dtype of double or int from being
used to define time. Fixed by checking that an that the dtype is a
number instead.

Léon Victor authored 5 months ago and Jean Fechter committed 4 months ago

The tetrahedron reading logic used a non-tetrahedron-specifc `vtkUnstructredGrid::InsertNextCell` call which performed poorly due to a large amount of time spent managing a `std::set`, and struggling with memory allocation previsions.

With this change we instead handle the cell ID uniqueness ourselves and use the tetrahedron-specific API, resulting in a 2~3x speedup on a large dataset.

The logic is also refactored a bit for readability

Fix an int overflow that could occur when seeking a very large offset (it did happen in the geometry file)

Overall performance improvements

Reader regarding overlappingAMR can have an overflow issue due to his
offset. Few places in the writer should rely on vtkIdType instead of int
too.

INFO log messages will be logged by default to stdout, but this
logging level was used in a number of places throughout VTK, leading
to unexpected output messages. VTK should not print messages to stdout
by default except when reporting a warning or error condition. This
commit changes messages that do not report warnings/errors to TRACE so
messages will be logged only when the logging level is specifically
requested to be more verbose.

Since vtkIOEnSightCxxTests-MPI is optional, it is not necessary
to explicitly list ParallelMPI in TEST_DEPENDS.

Make sure to test temporal overlappingAMR for the version 2.2 and 2.3

The VTKHDF File Format contains a bug and has been bump to 2.3. The
vtkHDFReader has been updated to fix this issue too.

This verifies number of cells in patches that have been written out as
Element_t nodes.

Previously only surface elements in `BC_t` nodes would be read.  The new option
`LoadSurfacePatch` allows to read 2D elements that are not `BC_t` nodes(handled
by the LoadBndPatch option)  but rather `Element_t` nodes.  This allows to read
boundary elements no matter if boundary conditions are assigned to them.

The cctype header is needed for std::isspace().

Deprecation usage in this class was wrong, we deprecate the variable
directly instead of the method.

Remove also the deprecation for SetHasTemporalData as it shouldn't be
deprecate.

The header has a method which returns a deprecated type. MSVC warns
about such usage, so suppress it.

The macro goes after the `struct` keyword.

With [recent changes to Fides], it is no longer necessary to supply
paths when loading a JSON file (as long as the JSON gives a file path
relative to the JSON path). Because this case can happen, the
`RequestInformation` of the Fides reader should continue to process the
meta data even if the paths object is empty.

This is particularly important in ParaView to populate the field
selection before they `Apply` button is hit. Otherwise, all the fields
must be loaded before they can be deselected.

[recent changes to Fides]: vtk/fides!172

Remove the threshold that existed before because
the new image testing framework is now used, whose
threshold ranges from 0 to 1, and 0.05 is a good default.