pythonloc | Python replacement that imports packages

{0} is replaced by the value of the run element from your workflow definition

https://docs.github.com/en/actions/reference/workflow-syntax-for-github-actions#using-a-specific-shell

You can also change the error handling if you use e.g.

Everytime once you change the ENV["PYTHON"] you need to run Pkg.build("PyCall"). It is the only moment in time when Julia actually reads Python's configuration - otherwise it just does not care.

If you want to test a matrix of Python configuration, I still advise you to have a miniconda Python for each of them (although it does not need to be installed by Julia - it can miniconda external to Julia). When running Pkg.build("PyCall") with a non-Conda Python various "issues" tend to come out and you may end up fine-tunnining Julia for each of those installation.

UPDATE

Per comment I have updated the response to provide the proper way of adding timeout and still succeed when timing out while also supporting normal failure.

Basically we check for error codes 124 (timeout) and 0 (success) and make sure that we don't exit on those codes. But if we receive anything else then we exit to match what github actions normally does on failures

DEMO SUCCESS ON TIMEOUT

DEMO FAILURE ON ERROR

Code Snippet

The problem is the line

/opt/hostedtoolcache/Python/3.8.5/x64/lib/python3.8/site-packages/setuptools/dist.py:454: UserWarning: Normalizing '0.2.3-dev26' to '0.2.3.dev26' warnings.warn(tmpl.format(**locals()))

in your Publish python package step.

This causes a mismatch (. vs - after 0.2.3):

Actual file: dist/wraplite-0.2.3.dev26-py3-none-any.whl

File that you're trying to upload: dist/wraplite-0.2.3-dev26-py3-none-any.whl

You may have better luck using the already existing codecov action.

In any case, you are running your action in a windows environment, so I would suggest not relying too much on unix-specific behaviour (i.e. process substitution <()) because this may require support from the underlying OS to work. Try this:

The Python 3 executable is usually available with python3 on Linux and Mac, where the default python might the old Python 2. This is not the case on Windows so the solution is just to use python.