pytest fails to load pooch in developer setup

Describe the bug

The following Traceback is obtained when attempting to run any form of pytest, such as pytest mne/tests from the project root directory, when the current main is cloned on my system with the recommended YML-based environment creation and an editable pip install, with all requirements satisfied.

ImportError while loading conftest '/Users/tevesjb/repositories/mne-python/mne/conftest.py'.
mne/conftest.py:31: in <module>
    test_path = testing.data_path(download=False)
mne/datasets/testing/_testing.py:22: in data_path
    return _data_path(path=path, force_update=force_update,
mne/datasets/utils.py:213: in _data_path
    pooch = _soft_import('pooch', 'dataset downloading', strict=True)
mne/utils/check.py:273: in _soft_import
    raise RuntimeError(f'For {purpose} to work, the {name} module is '
E   RuntimeError: For dataset downloading to work, the pooch module is needed, but it could not be imported.

Steps to reproduce

pooch is available in the environment that the test is being run from:

>>> import pooch
>>> pooch.__version__
'v1.5.1'

Just as bizarrely, if I use the same facility as check.py, the load is successful:

>>> import mne
>>> from mne.utils.check import _soft_import
>>> p = _soft_import('pooch', 'dataset downloading', strict=True)
>>> print(p)
<module 'pooch' from '/Users/tevesjb/miniconda3/envs/mne-dev/lib/python3.9/site-packages/pooch/__init__.py'>

However, running pytest triggers the above RuntimeError and traceback, regardless of which test (presumably because the error occurs in test configuration).

Expected results

The tests at least run, even if they fail.

Actual results

The test configuration fails, so I can’t even use the debugger via pytest --trace.

Additional information

Platform:      macOS-10.15.7-x86_64-i386-64bit
Python:        3.9.7 | packaged by conda-forge | (default, Sep 23 2021, 07:31:23)  [Clang 11.1.0 ]
Executable:    /Users/tevesjb/miniconda3/envs/mne-dev/bin/python
CPU:           i386: 8 cores
Memory:        16.0 GB

mne:           0.24.dev0
numpy:         1.21.2 {blas=NO_ATLAS_INFO, lapack=lapack}
scipy:         1.7.0
matplotlib:    3.4.2 {backend=MacOSX}

sklearn:       0.24.2
numba:         0.53.1
nibabel:       3.2.1
nilearn:       0.8.0
dipy:          1.4.1
cupy:          Not found
pandas:        1.3.0
mayavi:        4.7.2
pyvista:       0.32.1 {pyvistaqt=0.5.0, OpenGL 4.1 INTEL-14.7.18 via Intel(R) Iris(TM) Plus Graphics OpenGL Engine}
vtk:           9.0.1
PyQt5:         5.12.3

[drammock]

apparently I was a few days behind and missed that; with the actually up-to-date main this bug does not exist

closing because already fixed

[larsoner]

I’m up for making a short PR; is there a relevant issue or should I just add it to the referenced tuple?

You can just add it to the tuple I think, and update toward the top of the Enhancements section of doc/changes/latest.inc (with the special “new contributor” substitution + bolding), and also doc/changes/names.inc to add your name!