NETCDF3 endian error in python 3.12 on windows server

[veenstrajelmer]

• netcdf4 version: 1.7.1.post2
• env: Windows server 2022, Python 3.12.7

Error report
In a poetry environment with numpy 2.0.2 and netcdf4 1.7.1.post2 I get the following error in my test suite on github: "RuntimeError: only endian='native' allowed for NETCDF3 files, got 'big' (variable 'network', group '/')". This also happens for other variables, e.g. "RuntimeError: only endian='native' allowed for NETCDF3 files, got 'big' (variable 'mesh2d', group '/')".

The environment details and tracebacks are available here:
https://github.com/Deltares/HYDROLIB-core/actions/runs/11290766627/job/31403293222

Would this be something that will be solved with netcdf4 1.7.2 (or a new 1.7.2 post release like mentioned in #1370) or is this issue not on the radar yet?

Additional information
What I tried:

• forcing NETCDF4_CLASSIC instead of NETCDF3_CLASSIC upon writing files only made three of the 21 tests pass. The rest is related to reading mostly and passing NETCDF4_CLASSIC did not change anything there. I would also like to prevent hardcoding that in my code anyway.
• What makes it complex is that I cannot reproduce this on my local Windows 10 machine in the same poetry environment (via the lockfile) and the same python version (3.12.7).
• when enforcing numpy 2.1.2 for python 3.12 (the same version as in your github tests), the errors still occur. I tried this since I have seem ValueError: numpy.dtype size changed when numpy and netcdf4 versions did not match up (Investigate numpy 2.0.0 support Deltares/HYDROLIB-core#657)

[chrishavlin]

Would this be something that will be solved with netcdf4 1.7.2 (or a new 1.7.2 post release like mentioned in #1370) or is this issue not on the radar yet?

@veenstrajelmer Looks like there is an unreleased fix already: #1355

[veenstrajelmer]

Ok that is amazing. In that case, I am very much looking forward to a new release. When is that expected?

[chrishavlin]

Not sure if the question was directed to me, but I'm not part of the netcdf4 dev team, just someone else who had the same error and was looking through linked issues and related pull requests :) So I couldn't tell you when the next netcdf4-python release is planned for.

[veenstrajelmer]

Fair enough, I was not paying attention. @jswhit could you please link this issue to PR #1355 (and close the issue) if that fix indeed resolves the problem in this issue? Also, now for several fixes, I am very much looking forward to a new release. When is that expected?

[jswhit]

Hope to create a new release this week

[jswhit]

there's a new release up on pypi. Feel free to re-open this issue if the problem persists.

[veenstrajelmer]

This issue was resolved with the 1.7.2 release, thanks a lot!