Failing tests for v3.3.0 on Alpine Linux

[OTLabs]

I am building updated package for v3.3.0 and giving followup to all your kind feedback about failing tests.

Now I got following tests failing:

=================================== FAILURES ===================================
__________________________ test_selocale_unsupported ___________________________

tmp_path = PosixPath('/tmp/pytest-of-builder/pytest-0/test_selocale_unsupported0')

    def test_selocale_unsupported(tmp_path):
>       with pytest.raises(locale.Error):
E       Failed: DID NOT RAISE <class 'locale.Error'>

tests/i18n/test_i18n.py:27: Failed
__________________ test_ogvjs_installed_script_missing_param ___________________

    def test_ogvjs_installed_script_missing_param():
        # run from installed script to check real conditions
        script = subprocess.run(
            ["/usr/bin/env", "fix_ogvjs_dist"],
            text=True,
            capture_output=True,
            check=False,
        )
>       assert script.returncode == 1
E       assert 127 == 1
E        +  where 127 = CompletedProcess(args=['/usr/bin/env', 'fix_ogvjs_dist'], returncode=127, stdout='', stderr="env: can't execute 'fix_ogvjs_dist': No such file or directory\n").returncode

tests/ogvjs/test_ogvjs.py:55: AssertionError
________________________ test_ogvjs_installed_script_ok ________________________

tmp_path = PosixPath('/tmp/pytest-of-builder/pytest-0/test_ogvjs_installed_script_ok0')
videojs_url = 'https://github.com/videojs/video.js/releases/download/v7.6.4/video-js-7.6.4.zip'
ogvjs_url = 'https://github.com/brion/ogv.js/releases/download/1.6.1/ogvjs-1.6.1.zip'
videojs_ogvjs_url = 'https://github.com/hartman/videojs-ogvjs/archive/v1.3.1.zip'

    @pytest.mark.slow
    def test_ogvjs_installed_script_ok(tmp_path, videojs_url, ogvjs_url, videojs_ogvjs_url):
        # run from installed script to check real conditions
    
        prepare_ogvjs_folder(tmp_path, videojs_url, ogvjs_url, videojs_ogvjs_url)
    
        script = subprocess.run(
            ["/usr/bin/env", "fix_ogvjs_dist", str(tmp_path)],
            text=True,
            capture_output=True,
            check=False,
        )
>       assert script.returncode == 0
E       assert 127 == 0
E        +  where 127 = CompletedProcess(args=['/usr/bin/env', 'fix_ogvjs_dist', '/tmp/pytest-of-builder/pytest-0/test_ogvjs_installed_script_ok0'], returncode=127, stdout='', stderr="env: can't execute 'fix_ogvjs_dist': No such file or directory\n").returncode

tests/ogvjs/test_ogvjs.py:77: AssertionError



...



=========================== short test summary info ============================
FAILED tests/i18n/test_i18n.py::test_selocale_unsupported - Failed: DID NOT RAISE <class 'locale.Error'>
FAILED tests/ogvjs/test_ogvjs.py::test_ogvjs_installed_script_missing_param - assert 127 == 1
FAILED tests/ogvjs/test_ogvjs.py::test_ogvjs_installed_script_ok - assert 127 == 0
=================== 3 failed, 367 passed in 80.17s (0:01:20) ===================

The first test as expected fails, so I will continue to ignoring it.

Please, let me know if you need a full check log for ogvjs related errors.

Currently I build the wheel and:

python3 -m venv --clear --without-pip --system-site-packages .testenv   
.testenv/bin/python3 -m installer .dist/*.whl
.testenv/bin/python3 -m pytest \
                --runslow \
                --cov=zimscraperlib \
                --cov-report=term --cov-report term-missing

[kelson42]

@rgaudin @benoit74 Is that a kind of regression? A bit concerned that @OTLabs seems to keep discovering compilation/testing issues post-release. Should we do something to harden our CI?

[rgaudin]

We don't test on Alpine on purpose because we don't support it. We have a blocker (and a ticket) on i18n.

[rgaudin]

@OTLabs I suppose you're following APKBUILD_examples:Python.

Could you share the grep517-built wheel?

The issue is that the tests can't find the generated script in the PATH. Looking at what you run, that seems normal 😅 We should use sys.executable in the test instead of looking for PATH although we were doing it on purpose because we want to make sure this gets installed properly. The solution might be to make this test optional as for --runslow.

Please send the wheel so we can test. I've checked already that grep517 and installer both support entry_points so your not reliance on pip shouldn't be an issue.

[OTLabs]

Here it goes! zimscraperlib-3.3.0-py3-none-any.whl

[OTLabs]

I would like to share with you the APKBUILD I use to build the package.

[benoit74]

@rgaudin Maybe we should just restore the --runinstalled flag we removed because we thought we would always be "installed", I think it would be enough?

[benoit74]

We don't test on Alpine on purpose because we don't support it. We have a blocker (and a ticket) on i18n.

Here it is even worse, I exceptionally ran tests on Alpine (since there was an issue on this) but not with the same procedure than the one use by @OTLabs (I relied on pip to install stuff).

I don't think that having full support of Alpine (and all Linux distro in fact, and all build chains) is a priority for scrapers, please speak up otherwise.

[rgaudin]

@rgaudin Maybe we should just restore the --runinstalled flag we removed because we thought we would always be "installed", I think it would be enough?

Yep