Conda package not compiling

[davide-f]

Describe the bug
The conda packages are unfortunately not compiling, see https://github.com/conda-forge/chaospy-feedstock

I've read in the comments that this issue is known, though believed a dedicated issue could help, sharing my experience too.
I'm trying to explore to fix the feedstock, see conda-forge/chaospy-feedstock#56 .

Just to understand, I've seen that the feedstock is only maintained by a single person: jscheller. Is he still available?
It would be great to have his support; potentially extending the maintainers of the recipe too.
Just need to list more codes here:
https://github.com/conda-forge/chaospy-feedstock/blob/4b33c44d1bcca6b5bb07342b5c26895348698f99/recipe/meta.yaml#L47-L48

Update: conda-forge/chaospy-feedstock#56 fixes the weeling but the CI is breaking and it does not pass.

[jonathf]

The Chaospy project does not maintain the Conda feedstock, so jschiller isn't here. We only target the normal pypa pip index from here.

But I am happy to contribute to get Conda builds working again if it is within reason.

[davide-f]

Many thanks @jonathf :)
I've proposed I've also proposed conda-forge/chaospy-feedstock#56 that should re-enable the conda installation [hopefully].

Elements of attention are:

• minimum python version: 3.7 in agreement to pypi. When this dependency changes, the feedstock should be changed as well, but it is a line in meta.yaml [see changes], the maintenance on that is minimal
• in that PR, the test command has been simplified into pip check instead of py.test [pytest]. The idea of this is that this tests should be already be done in github and before pushing to pypi. If a push to pypi contains few tests that fail, it should not be the reason to block only the conda installation. This is why probably having less checks on the conda feedstock may simplify the process

The feedstock is generally very light to maintain; there is the autobot merge and it relies on the pip installation. As long as those are fine, there should be minimal risks.

On the other hand, having it function can be very helpful for anaconda and miniconda users. Relying on mixed pip-conda installation can lead to problems, we have experienced it in the pypsa-earth repository and this could help many