[BUG] Tests fail with NumPy 2.2.0

[penguinpee]

Fedora has recently upgraded NumPy to 2.2.0 in the development branch. Building meshio with that version results in a number of tests failing:

=========================== short test summary info ============================
FAILED tests/test_dolfin.py::test_dolfin[mesh3] - ValueError: could not conve...
FAILED tests/test_gmsh.py::test_gmsh22[False-mesh9] - ValueError: cannot resh...
FAILED tests/test_gmsh.py::test_gmsh22[False-mesh10] - ValueError: cannot res...
FAILED tests/test_gmsh.py::test_gmsh22[False-mesh11] - ValueError: cannot res...
FAILED tests/test_gmsh.py::test_gmsh22[False-mesh12] - ValueError: cannot res...
FAILED tests/test_gmsh.py::test_gmsh22[False-mesh13] - ValueError: cannot res...
FAILED tests/test_gmsh.py::test_gmsh22[False-mesh14] - ValueError: cannot res...
FAILED tests/test_gmsh.py::test_gmsh40[False-mesh8] - ValueError: cannot resh...
FAILED tests/test_gmsh.py::test_gmsh40[False-mesh9] - ValueError: cannot resh...
FAILED tests/test_gmsh.py::test_gmsh40[False-mesh10] - ValueError: cannot res...
FAILED tests/test_gmsh.py::test_gmsh40[False-mesh11] - ValueError: cannot res...
FAILED tests/test_gmsh.py::test_gmsh40[False-mesh12] - ValueError: cannot res...
FAILED tests/test_gmsh.py::test_gmsh40[False-mesh13] - ValueError: cannot res...
FAILED tests/test_gmsh.py::test_gmsh41[False-mesh8] - ValueError: cannot resh...
FAILED tests/test_gmsh.py::test_gmsh41[False-mesh9] - ValueError: cannot resh...
FAILED tests/test_gmsh.py::test_gmsh41[False-mesh10] - ValueError: cannot res...
FAILED tests/test_gmsh.py::test_gmsh41[False-mesh11] - ValueError: cannot res...
FAILED tests/test_gmsh.py::test_gmsh41[False-mesh12] - ValueError: cannot res...
FAILED tests/test_gmsh.py::test_gmsh41[False-mesh13] - ValueError: cannot res...
FAILED tests/test_ugrid.py::test_io[1e-07-.ugrid-mesh0] - meshio._exceptions....
FAILED tests/test_ugrid.py::test_io[1e-07-.ugrid-mesh1] - meshio._exceptions....
FAILED tests/test_ugrid.py::test_io[1e-07-.ugrid-mesh2] - meshio._exceptions....
FAILED tests/test_ugrid.py::test_io[1e-07-.ugrid-mesh3] - meshio._exceptions....
FAILED tests/test_ugrid.py::test_io[1e-07-.ugrid-mesh4] - meshio._exceptions....
FAILED tests/test_ugrid.py::test_io[1e-07-.ugrid-mesh5] - meshio._exceptions....
============ 25 failed, 714 passed, 6 skipped, 25 warnings in 4.68s ============

The set of errors is:

ValueError: could not convert string to float: 'np.float64(63.69616873214543)'
ValueError: cannot reshape array of size 1 into shape (4,2)
meshio._exceptions.ReadError: Header of ugrid file is ill-formed

The full log of all test failures is too big to post here. But this should be rather easy to reproduce in a venv. On my Fedora box I'd do:

python3.13 -m venv .venv
. .venv/bin/activate
pip installe -e .
pip install pytest netCDF4 h5py
pytest -v

For a quick glance, the log from the latest build attempt is accessible online:

https://kojipkgs.fedoraproject.org/work/tasks/8425/126988425/build.log

It will be cleaned up over time, however.

[musicinmybrain]

FAILED tests/test_dolfin.py::test_dolfin[mesh3] - ValueError: could not conve...

Fixed by #1506.

FAILED tests/test_gmsh.py::test_gmsh22[False-mesh9] - ValueError: cannot resh...
FAILED tests/test_gmsh.py::test_gmsh22[False-mesh10] - ValueError: cannot res...
FAILED tests/test_gmsh.py::test_gmsh22[False-mesh11] - ValueError: cannot res...
FAILED tests/test_gmsh.py::test_gmsh22[False-mesh12] - ValueError: cannot res...
FAILED tests/test_gmsh.py::test_gmsh22[False-mesh13] - ValueError: cannot res...
FAILED tests/test_gmsh.py::test_gmsh22[False-mesh14] - ValueError: cannot res...
FAILED tests/test_gmsh.py::test_gmsh40[False-mesh8] - ValueError: cannot resh...
FAILED tests/test_gmsh.py::test_gmsh40[False-mesh9] - ValueError: cannot resh...
FAILED tests/test_gmsh.py::test_gmsh40[False-mesh10] - ValueError: cannot res...
FAILED tests/test_gmsh.py::test_gmsh40[False-mesh11] - ValueError: cannot res...
FAILED tests/test_gmsh.py::test_gmsh40[False-mesh12] - ValueError: cannot res...
FAILED tests/test_gmsh.py::test_gmsh40[False-mesh13] - ValueError: cannot res...
FAILED tests/test_gmsh.py::test_gmsh41[False-mesh8] - ValueError: cannot resh...
FAILED tests/test_gmsh.py::test_gmsh41[False-mesh9] - ValueError: cannot resh...
FAILED tests/test_gmsh.py::test_gmsh41[False-mesh10] - ValueError: cannot res...
FAILED tests/test_gmsh.py::test_gmsh41[False-mesh11] - ValueError: cannot res...
FAILED tests/test_gmsh.py::test_gmsh41[False-mesh12] - ValueError: cannot res...
FAILED tests/test_gmsh.py::test_gmsh41[False-mesh13] - ValueError: cannot res...

I am not sure what should be done about these, but they appear to share a common root cause, in which this

meshio/src/meshio/gmsh/common.py

Lines 68 to 71 in b2ee998

	if is_ascii:
	data = np.fromfile(f, count=num_items * (1 + num_components), sep=" ").reshape(
	(num_items, 1 + num_components)
	)

raises an exception of the form

E           ValueError: cannot reshape array of size 1 into shape (4,4)

where the “target” shape varies. I spot-checked one failing case and found that the array indeed contained only one scalar value.

FAILED tests/test_ugrid.py::test_io[1e-07-.ugrid-mesh0] - meshio._exceptions....
FAILED tests/test_ugrid.py::test_io[1e-07-.ugrid-mesh1] - meshio._exceptions....
FAILED tests/test_ugrid.py::test_io[1e-07-.ugrid-mesh2] - meshio._exceptions....
FAILED tests/test_ugrid.py::test_io[1e-07-.ugrid-mesh3] - meshio._exceptions....
FAILED tests/test_ugrid.py::test_io[1e-07-.ugrid-mesh4] - meshio._exceptions....
FAILED tests/test_ugrid.py::test_io[1e-07-.ugrid-mesh5] - meshio._exceptions....

All of these are happening here:

meshio/src/meshio/ugrid/_ugrid.py

Lines 81 to 82 in b2ee998

	if not nitems.size == 7:
	raise ReadError("Header of ugrid file is ill-formed")

but, again, I am not sure what should be done.

All of these errors can be reproduced in a fresh git checkout by simply running tox.