CMake: allow use of system-provided meshlab & pybind11

[nim65s]

follows cnr-isti-vclab/vcglib#248

NB: this PR is way easier to review by hidding whitespaces

[alemuntoni]

@nim65s did you tested the pymeshlab package built and installed in an environment where meshlab was already installed?
Right now I don't have the time to do it myself... But I would be really surprised if it just works out of the box :D

[nim65s]

Yes, I did something like this:

>>> from pymeshlab import MeshSet
>>> ms = MeshSet()
>>> ms.load_new_mesh("/…/base_link.STL")
>>> ms.meshing_decimation_quadric_edge_collapse()
>>> ms.save_current_mesh("/…/base_link2.STL")

And base_link2.STL is created and lighter than the initial one.

I did that with nixpkgs though, which add some additional magic to make certain things autowork and break others.

I had to copy libio_* & libfliter_* plugins from meshlab to pymeshlab for example. I guess the loading mechanism can be improved for this, but I did not want to put that in the same PR set

[nim65s]

I tested that on linux and macos, with the following recipe: https://github.com/nim65s/nixpkgs/blob/meshlab/pkgs/applications/graphics/pymeshlab/default.nix

[alemuntoni]

Ok, that makes sense.

I would just put a warning somewhere to say that the package would use meshlab built with single float precision, which could lead to different behaviors w.r.t. the pypi pymeshlab package (which uses double precision).

Anyway, that's really great! Thanks for your work!!

[nim65s]

out of curiosity, what is the rationnale for the default single on meshlab but default double on pymeshlab ?

[alemuntoni]

On pymeshlab it is default double because it is default also in python.
In my opinion it should be default double also meshlab, but since it was born with floats (which made sense 20 years ago) we keep it float mainly for backward compatibility.

[nim65s]

I've added a warning :)

[alemuntoni]

Thank you @nim65s!! :)