Closed github-actions[bot] closed 3 weeks ago
The nightly feedstock builds that run tests all failed. Specifically the import test failed for the Python 3.9 environment.
Fortunately/frustratingly I already started troubleshooting the source of this problem yesterday. anndata 0.11.0 dropped support for Python 3.9, but the feedstock PR was automerged before I could fix it.
xref: https://github.com/conda-forge/anndata-feedstock/pull/42, https://github.com/conda-forge/anndata-feedstock/pull/45
@jdblischak I think that our Python 3.9 things will simply "find" anndata
0.10 .....
I think that our Python 3.9 things will simply "find"
anndata
0.10 .....
You're half right :-) I just checked. anndata is pinned <0.11
in the UDF image but not the notebook image
No I mean that (for pip
at least) when a Python 3.9 system tries to install anndata >= 0.10.1
, the most recent version 0.10.9, and no 0.11.anything will even be offered. And I thought it was similar for Conda; I may be mistaken ...
pip yes, conda no
mamba create --yes -n py39 -c conda-forge python=3.9
mamba activate py39
mamba install --dry-run -c conda-forge anndata=0.11
## + anndata 0.11.0 pyhd8ed1ab_0 conda-forge 112kB
pip install --dry-run anndata==0.11
## ERROR: Ignored the following versions that require a different python version: 0.11.0 Requires-Python >=3.10; 0.11.0rc3 Requires-Python >=3.10
But once I fix the conda recipe and conda-forge metadata, then conda will behave the same (and the nightlies will be fixed)
I fully expect an explicit request for anndata=0.11
to fail on a Python 3.9 system. Absolutely.
I did expect anndata >= 0.10.1
to succeed on a Python 3.9 system (in pip
or conda
) ...
Awesome, thanks @jdblischak !! :)
Nightly feedstock build failure for tiledbsoma-feedstock at https://dev.azure.com/TileDB-Inc/CI/_build?definitionId=43&_a=summary