GitLab

ParaView/Servers/ServerManager/Testing/Python/CMakeLists.txt is modified
to handle specially for PythonFilters.py which tests the Python
programmable filter functionality. The modification adds PythonFilters.py
to the test list, it also makes the ctest system parses the outputs of
PythonFilters.py for regular expression that contains "Error". The reason
for this special behavior is that the programmable filter back end
constructs Python scripts on the fly after it receives input expressions.
If there is any error in the input expressisons, the errors or exceptions
that are caused when the constructed script is run, will not be catched
by the constructing environment. So even the test fail, there is no way
to inform ctest about it since the exception is not catchable. The only
possible way is to let ctest parse the outputs and when it sees anything
containing "Error", assuming that the test failed. Of course, this is not
comprehensive, for example, if "Warning" is thrown, ctest will not detect
it. However, given the constraints, this is the best can be done.

Servers/ServerManager/Testing/Python/PythonFilters.py has been
drastically modified to do a test for all programmable filter
currently supported. The tests also covers all possible combinations.
For example, min works on scalars, vectors and tensors that are
assocaited with points or cells. So there are six combinations for
min in total. All cases are tested.

Too many modifications have been done to algorithms.py to resolve the
conflicts manually. Just simply overwrite the entire file because the
new file is correct for sure.

Use buffer() to wrap C++ pointer so that memory address is correctly
tranfered between Python objects, and between Python and C++ objects.
Also add array association attributes to Python objects so that we can
track whether the input data array is associated with cell or will
point. This is very important because otherwise we will be unable to
get the right input data, or put the output to wrong association.

The coprocessing library tests need to run in sequence
and I'm setting them to only run in serial to make sure that
they work as expected for parallel testing.

Had a problem where variable files with a property name of "P" where being
incorrectly detected as little endian fortran binary files.

This should enable one to record plugin tests without the maintenance of constantly updating test indeces whenever a plugin is added or removed.

Adding more toolbars causes the render window to resize which inturn causes chart test failures because different axis are produced.

FlashReader was reading the version number information incorrectly on BE
machines. Fixed that.

Add more breaks to the toolbars so that all toolbars should have all buttons visible. The goal being to prevent test failures.

In Edit | Settings | RenderView pane, added options to change the topology
resolution parameters.

Otherwise as we make selections, selection events will accumulate and get emited for each connection we created but did not disconnect.

We had a bug where the properties from each node where not being sent to the root process for collection and than propagted back to each node.

* Handled animating slice-type and similar properties.

* GetAnimationTrack() works for Visibility and Opacity.

  Added logic to simple.py to locate/create RepresentationAnimationHelper for a
  source proxy when needed by GetAnimationTrack().

  Added code to pqProxy to discover helper proxies registered from Python. This
  was essential since GetAnimationTrack() could result in creation of an
  RepresentationAnimationHelper which the GUI wasn't aware of resulting in the
  AnimationView showing up incorrectly for such tracks.

vtkPVContourFilter was using vtkAMRDualClip even for point-data arrays. Fixed
that. Also, the GUI never changed the attribute mode in the array-selection
property. This was due to missing FieldDataDomain. Fixed that.

It was using weird case for filename property as well it didn't have
default number of elements set which was causing problems in
pvpython

Updated the grid connectivity tests to also run in client server mode, since
the filter is now fixed.

Install a small .git/hooks/pre-commit script in the source tree at CMake
configuration time, instead of telling the user to do it. When the user
first attempts to commit this bootstrapping hook will trigger and print
instructions to install the real hooks. This approach allows us to
require local hooks for commit, but not affect builds.