Drop py3.6, add 3.10; switch CI from macos-10.15 to macos-latest #1507
Conversation
There was a problem hiding this comment.
thanks @kanderso-nrel. Should we handle the uncommenting in another PR or try it here?
I think we've cleared out all of the documentation's explicit references to python version compatibility. And setup.py classifier is just python 3. ✅
We should update the upload to anaconda.org part of the release procedures once this is merged.
Do we need to do anything with pvlib-benchmarker? I don't think so but not confident.
Mild preference for another PR
+1, and in hindsight I shouldn't have added 3.10 to that list before adding it to the test matrix. But per your suggestion a few months ago, maybe it's time to stop updating the
I also don't think so, at least not immediately. The only thing is that changing the python version will create yet another disconnect/variant in the plots. I wonder if we should periodically re-run the benchmarks using the current environments so that comparisons across time are apples to apples. |
I think we should first resolve conda-forge/pvlib-python-feedstock#30 @pvlib/pvlib-maintainer please merge if you also approve |
[ ] Closes #xxxx[ ] Tests added[ ] Updates entries indocs/sphinx/source/referencefor API changes.docs/sphinx/source/whatsnewfor all changes. Includes link to the GitHub Issue with:issue:`num`or this Pull Request with:pull:`num`. Includes contributor name and/or GitHub username (link with:ghuser:`user`).[ ] New code is fully documented. Includes numpydoc compliant docstrings, examples, and comments where necessary.remote-data) and Milestone are assigned to the Pull Request and linked Issue.I expect Windows tests to fail until mamba 0.25.1 is released; see conda-forge/micromamba-feedstock#90 and mamba-org/mamba#1828.
I'll leave this PR as a draft until that issue is resolved and everything is passing here.The mamba fix is deployed, so those tests are passing. However the 3.6 remote-data tests will continue to fail (and the 3.10 won't run at all) because it's apull_request_targetaction for security reasons and still using the configuration onmaster. There's nothing to "fix" there, it just means we need to manually inspect the changes to the remote-data configuration instead of relying on the CI.We should probably also take a look at uncommenting some items in the conda environment files. I've not done that here. The new 3.10 environment is essentially just a copy of the 3.9 environment.