CI breaks with netCDF4 1.7.3 wheel

[kmuehlbauer]

• See breaking CI run https://github.com/NCAS-CMS/pyfive/actions/runs/18482160354 with netCDF4 1.7.3
• Latest working run https://github.com/NCAS-CMS/pyfive/actions/runs/18452134189 with netCDF4 1.7.2

[kmuehlbauer]

It might not be netCDF4 1.7.3 actually, but obviously pip is backtracking s3fs vs aiobotocore. So we end up with ancient versions of some packages. Not sure, if this is intended.

[valeriupredoi]

no no you are right, Kai - netcdf4=1.7.3 from PyPI is causing it. I'll pin it now. We really ought to get our testing dependencies from conda-forge, all manners of broken or missing wheels on PyPI (see also compat with Python 3.14 - all fine off conda-forge, issues from PyPI) 🍻

[valeriupredoi]

pin in #124 - they released out of their arse, the new version is not even Python 3.14 compat; let's wait a bit until someone will invariably raise a red flag at netCDF4-python 😁

[davidhassell]

Our good friends at Unidata might not see it like that. Maybe find out what's going on, first ...

[valeriupredoi]

Our good friends at Unidata might not see it like that. Maybe find out what's going on, first ...

as Shakespeake said, something's not well in DenmarkCDF4 Unidata/netcdf4-python#1438