yay, all green :)

Had to remove pre-commit in tests/requirements.txt as not available for s390x... that can potentially be moved back after somebody converts the pre-commit-feedstock in conda-forge to be noarch (that should be possible with the right arguments). pre-commit anyway runs via the pre-commit-ci service with auto-commit and does not rely on the conda-ci containers.

@jezdez Regarding your windows container note: what would you expect from windows containers, just more speed? That should be also possible by 1. not installing miniconda (it is pre-installed on all public osx, linux, windows github action runners, see) and 2. using the cache action with a good cache key, e.g. key: ${{ runner.os }}-${{ matrix.python-version }}-${{ hashFiles('test\requirements') }}-${{ hashFiles('dev\windows\setup.bat') }}-${{ DAY_OF_WEEK }}

yay, all green :)

Had to remove pre-commit in tests/requirements.txt as not available for s390x... that can potentially be moved back after somebody converts the pre-commit-feedstock in conda-forge to be noarch (that should be possible with the right arguments). pre-commit anyway runs via the pre-commit-ci service with auto-commit and does not rely on the conda-ci containers.

Fascinating, might make sense to open a ticket in the feedstock?

FWIW we're relying on pre-commit being installed in the development environment, so

@jezdez Regarding your windows container note: what would you expect from windows containers, just more speed? That should be also possible by 1. not installing miniconda (it is pre-installed on all public osx, linux, windows github action runners, see) and 2. using the cache action with a good cache key, e.g. key: ${{ runner.os }}-${{ matrix.python-version }}-${{ hashFiles('test\requirements') }}-${{ hashFiles('dev\windows\setup.bat') }}-${{ DAY_OF_WEEK }}

Yeah, I've been meaning to look into whether we can use the pre-installed miniconda for conda as well and just worry that we end up in a cyclic situation where the CI environment needs to do something different than what the old miniconda did.

Besides that speed component I'm wondering what we can do to further improve the parity between development environments and the CI environment, to reduce the barriers of entry for contributors. A readily available base image might be worth exploring since it'd get Windows devs set up faster. Just a theory though, not really in scope of this PR :)

conda-forge/pre-commit-feedstock#54

Readded pre-commit and removed s390x from the platform list... s390x can be added later when the pre-commit noarch situation is resolved.

adding arm64/ppc64le already serves the purpose of a better Apple M1 experience and prove of consistency.

all green again. ready for another review / merge. thanks