Tests using pygmsh fail due to upstream changes

[kinnala]

This is due to the commit: nschloe/pygmsh@ed581be#diff-bc0dadc6113a14c1c7891320765ee767

[kinnala]

Seems to be a simple enough fix: generate_mesh was turned into a method of Geometry in 7.0.0.

[kinnala]

Wow, actually this new pygmsh release using gmsh package fails pretty fatally on my operating system:

        if handle is None:
>           self._handle = _dlopen(self._name, mode)
E           OSError: libXcursor.so.1: cannot open shared object file: No such file or directory

Not sure why it's linking to libXcursor but anyhow, need to do some adjustments.

[kinnala]

nschloe/pygmsh#365 causes the tests to fail but nschloe/pygmsh#360 is the PR which causes this to fail on my laptop.

[kinnala]

I need to look into this later, I'm now a bit busy with the grant application.

[kinnala]

Ok, the grant deadline was extended so I can now look into this.

[kinnala]

(vedoenv) conda-shell-chrootenv:tom@tunkki:~/src/scikit-fem$ python -c "import gmsh"
Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "/home/tom/.conda/envs/vedoenv/lib/python3.8/site-packages/gmsh-4.6.0-Linux64-sdk/lib/gmsh.py", line 39, in <module>
    lib = CDLL(libpath)
  File "/home/tom/.conda/envs/vedoenv/lib/python3.8/ctypes/__init__.py", line 373, in __init__
    self._handle = _dlopen(self._name, mode)
OSError: libXcursor.so.1: cannot open shared object file: No such file or directory

[kinnala]

I'm getting some of the required libraries from nixpkgs but it seems that the binaries distributed in gmsh package are compiled against some newer versions of xorg-related libraries that I can easily to install.

Workaround is to restrict to an older version of pygmsh which I'm going to do now. Maybe this is finally the reason to do #333 .

[kinnala]

Pinned pygmsh to 6.1.1. Making another issue to support 7.0.0.