Include all the modules that the official VTK wheel supports?

[whophil]

Comment:

In #252 , I manually enabled a module using a compiler flag. The module in question is included by default in the VTK wheels available on PyPi, but was not available in the conda-forge build.

This seems like a band-aid solution - to one-by-one enable modules which I find to available in the wheel but not in the feedstock.

I wonder if there is a better way to ensure that all the modules available in the wheel are available in the conda-forge build as well.

It appears that all the flags required to build the PyPi wheels can be found here: https://gitlab.kitware.com/vtk/vtk/-/tree/master/.gitlab/ci

The wheel build and conda-forge builds are quite different, but it seems that in general the conda-forge build enables features that cannot be easily packaged with the wheel. Is it reasonable that the conda-forge build should contain all modules/features available in the wheel, and then some?

I recognize that this is a complicated feedstock, but would love to hear from the feedstock maintainers on this!

CC @banesullivan

[mathstuf]

I wonder if there is a better way to ensure that all the modules available in the wheel are available in the conda-forge build as well.

Use VTK_BUILD_ALL_MODULES and then disable what doesn't work.

[mathstuf]

Note that if conda supports things like "with MPI", you should use those to toggle flags like VTK_USE_MPI and the like. As a seed for what things might make sense, you can look at what we use in CI to select these features.