🐞 Coverage report on new empty project fails hiding stderr of subproccess

[ckunki]

File exasol/toolbox/metrics.py in method total_coverage() calls subprocess.run():

        run(
            ["coverage", "json", f"--data-file={file}", "-o", f"{report}"],
            capture_output=True,
            check=True,
        )

A sample CI build displays the following error message.

subprocess.CalledProcessError: Command '['coverage', 'json', 
'--data-file=.coverage', 
'-o', '/tmp/tmpd9c4d11b/coverage.json']' 
returned non-zero exit status 2.
nox > Session report failed.

The current ticket proposes to

• surround the call by try/catch and
• When catching CalledProcessError then report stdout and stderr of the subprocess
• re-raise the exception then again

Additionally

• The template for file test/unit/unit_smoke_test.py uses import exasol.text_ai_extension but
• file exasol/text_ai_extension/init.py still contains no actual line of code
• which in result still makes coverage fail when in file pyproject.toml using threshold fail_under = 15

I was able to make coverage report terminate successfully by

• adding a function to __init__.py, e.g. (see below)
• importing the function into the test and verifying its value in the test:

File __init__.py:

def sample_function():
    return True

file test/unit/unit_smoke_test.py:

import {{ cookiecutter.import_package }}

def test_unit_smoke_test():
    assert {{cookiecutter.package_name}}.sample_function() 

[Nicoretti]

Importing the pre-existing version constant(s) would likely yield a more sensible test.

For example, a unit test that verifies the version in version.py is in sync with the version in pyproject.toml could simultaneously address two issues:

1. The explicit version check workflow step could be eliminated.
2. Some code from the module in question would be tested.