FCL building in Github Actions uses wrong Python version

[rayzchen]

https://github.com/humanoid-path-planner/hpp-fcl/actions/runs/5054289798/jobs/9069053579#step:5:61, but the action for setup-miniconda is given and correctly sets up Python 3.7 (https://github.com/humanoid-path-planner/hpp-fcl/actions/runs/5054289798/jobs/9069053579#step:3:466)

[jcarpent]

Thanks @rayzchen for raising this issue. I've just opened a new PR #427 fixing it.

[rayzchen]

I don't think that is going to fix the problem, since it is currently using Python 3.11 and disregarding the python-version option.

[jcarpent]

Yes, you're right. I've updated the version of goanpeca/setup-miniconda.

[rayzchen]

Is there a particular reason why goanpeca/setup-miniconda is used instead of conda-incubator/setup-miniconda?

[jcarpent]

Except for historical reasons, no. I've just changed it.

[jcarpent]

Fixed via #427.

[rayzchen]

I believe it still is creating a Python 3.11 module, since https://github.com/humanoid-path-planner/hpp-fcl/actions/runs/5122067821/jobs/9210618055#step:5:8654 shows that it cannot find hppfcl.hppfcl, which should point to a file called hppfcl.cp310-win_amd64.pyd, but the file found there (from my fork of this repo that saves the generated Python files) is called hppfcl.cp311-win_amd64.pyd. This would cause a ModuleNotFoundError in Python.

[jcarpent]

Perfect. Thanks for the report. I was only looking at the cmake configuration.
I will look at it.

[jcarpent]

@rayzchen I've found the issue which is not due to Conda and the CI, but rather the way we proceed for the packaging of the project. I will investigate it later with @fabinsch.

[rayzchen]

Sorry, my bad, my fork had merge conflicts which weren't resolved correctly and still used v1 of the setup-miniconda action. Once fixed, the correct version of Python was used. It seems that I tried to link two separate problems together.