Not working out of the box with cmd

[tdegeus]

Currently I need to activate the conda environment by hand on the cmd shell. Here are my instructions

    strategy:
      fail-fast: false
      matrix:
        runs-on: [windows-latest]

    name: ${{ matrix.runs-on }} • x64 ${{ matrix.args }} • py
    runs-on: ${{ matrix.runs-on }}

    steps:

    - name: Basic GitHub action setup
      uses: actions/checkout@v2

    - name: Set conda environment
      uses: mamba-org/provision-with-micromamba@main
      with:
        environment-file: environment.yaml
        environment-name: myenv

    - name: Build and install Python module
      shell: cmd /C CALL {0}
      run: |
        call micromamba activate myenv
        python -m pip install . -vv

[tdegeus]

I'm really not an expert on this, but I did create a reproducer that, when merged, should check for this in the future:

#43

[jonashaag]

Why does this GHA step fail and abort the rest of the run: code, but the job still continues and eventually succeeds? https://github.com/mamba-org/provision-with-micromamba/runs/6408495202?check_suite_focus=true

provision-with-micromamba/.github/workflows/test.yml

Line 34 in d7eeccc

- name: run python in cmd.exe

[xhochy]

I'm also hitting this issue in blue-yonder/turbodbc#380. From peeking at the source code of setup-miniconda, I can see that they append auto-activation to conda_hook.bat: https://github.com/conda-incubator/setup-miniconda/blob/7e642bb2e4ca56ff706818a0febf72bb226d348d/src/conda.ts#L259

Thus the fix should be that we add a similar statement to C:\Users\runneradmin\micromamba-root\condabin\mamba_hook.bat as part of the action.

[jonashaag]

Btw, I tend to not use activation anymore (in CI and on my local machines) because it simply causes too much trouble. I use micromamba run and micromamba shell (which we broke in 1.3.0 :-()

[xhochy]

I have a serious of commands which all need to run with CMD activation. There activation (and not mm run) is much more convenient to write.

[jdblischak]

Conveniently migrating to the newer mamba-org/setup-micromamba allowed me to drop the hacky workaround (call micromamba activate <env>) to manually activate the conda env in each step