Support VTK 9.x about chaste

[fcooper8472]

I have tested #152 against VTK 9.1, and the entire continuous test pack compiles and runs fine.

VTK 9.2 is going to be slightly more involved. One method was deprecated, which is addressed in 6218397.

However, VTK 9.2 causes TestVtkMeshReader to fail:

Entering TestGetNextFaceData

/home/fcooper/GitRepos/Chaste/Chaste/mesh/test/reader/TestVtkMeshReader.hpp:206: Error: Expected (first_face_data.NodeIndices[0] == 11u), found (0 != 11)
/home/fcooper/GitRepos/Chaste/Chaste/mesh/test/reader/TestVtkMeshReader.hpp:207: Error: Expected (first_face_data.NodeIndices[1] == 3u), found (11 != 3)
/home/fcooper/GitRepos/Chaste/Chaste/mesh/test/reader/TestVtkMeshReader.hpp:208: Error: Expected (first_face_data.NodeIndices[2] == 0u), found (3 != 0)
/home/fcooper/GitRepos/Chaste/Chaste/mesh/test/reader/TestVtkMeshReader.hpp:211: Error: Expected (next_face_data.NodeIndices[0] == 3u), found (0 != 3)
/home/fcooper/GitRepos/Chaste/Chaste/mesh/test/reader/TestVtkMeshReader.hpp:212: Error: Expected (next_face_data.NodeIndices[1] == 8u), found (3 != 8)
/home/fcooper/GitRepos/Chaste/Chaste/mesh/test/reader/TestVtkMeshReader.hpp:213: Error: Expected (next_face_data.NodeIndices[2] == 0u), found (8 != 0)
/home/fcooper/GitRepos/Chaste/Chaste/mesh/test/reader/TestVtkMeshReader.hpp:218: Error: Expected (first_face_data.NodeIndices[0] == 11u), found (0 != 11)
/home/fcooper/GitRepos/Chaste/Chaste/mesh/test/reader/TestVtkMeshReader.hpp:219: Error: Expected (first_face_data.NodeIndices[1] == 3u), found (11 != 3)
/home/fcooper/GitRepos/Chaste/Chaste/mesh/test/reader/TestVtkMeshReader.hpp:220: Error: Expected (first_face_data.NodeIndices[2] == 0u), found (3 != 0)
/home/fcooper/GitRepos/Chaste/Chaste/mesh/test/reader/TestVtkMeshReader.hpp:223: Error: Expected (next_face_data.NodeIndices[0] == 3u), found (0 != 3)
/home/fcooper/GitRepos/Chaste/Chaste/mesh/test/reader/TestVtkMeshReader.hpp:224: Error: Expected (next_face_data.NodeIndices[1] == 8u), found (3 != 8)
/home/fcooper/GitRepos/Chaste/Chaste/mesh/test/reader/TestVtkMeshReader.hpp:225: Error: Expected (next_face_data.NodeIndices[2] == 0u), found (8 != 0)
Failed

It looks as though the indices might be being re-ordered? Needs further investigation.

[jmpf]

There are tests above that one for

• node order/position
• element order
  These both work as expected and Chaste meshes use nodes and elements as the primary data structures.

Here the order of nodes within faces have been given an even permutation. This does not affect chirality so it's something that changes nothing. It's possible that new (or old) VTK gives back the other chirality and we flip it in the constructor.

I suggest putting the indices into std::set and using set equality with a comment to say that 9.2 is giving a different ordering. (Possibly overkill, but we're only doing it a couple of times.)

[fcooper8472]

@jmpf just to make sure I've understood correctly, anything up to a rotation is correct, so {11, 3, 0}, {3, 0, 11} and {0, 11, 3} are all "correct"?

Perhaps there's an elegant way of checking for a rotation...

[jmpf]

@fcooper8472 Yes. These all appear to be correct. The other way to do the test is to #ifdef on the VTK version...

I changed by mind about auto-flipping chirality in the constructor -- we can't do that for a triangular face in 3D.

[fcooper8472]

How about something like this?

c4d048d

[jmpf]

Probably overkill, but the rationale of the test is now crystal-clear.

[fcooper8472]

VTK 9.0, 9.1 and 9.2 appear to all be working.

This issue can be closed as soon as the portability workflow is updated to include the new portability-06 image.