[Libosinfo] [PATCH osinfo-db 7/7] tests: Add custom pytest-3 --network-tests option

Fabiano Fidêncio fidencio at redhat.com
Thu Mar 21 17:12:58 UTC 2019


On Thu, Mar 21, 2019 at 5:57 PM Cole Robinson <crobinso at redhat.com> wrote:
>
> On 3/21/19 11:21 AM, Fabiano Fidêncio wrote:
> > On Thu, Mar 21, 2019 at 4:07 PM Cole Robinson <crobinso at redhat.com> wrote:
> >>
> >> On 3/21/19 8:23 AM, Fabiano Fidêncio wrote:
> >>> On Wed, 2019-03-20 at 17:53 -0400, Cole Robinson wrote:
> >>>> This wires up a --network-tests option to save us from needing the
> >>>> OSINFO_DB_NETWORK_TESTS environment variable, though that method
> >>>> still
> >>>> works. It works a bit differently in that unless specified it
> >>>> entirely
> >>>> takes the tests out of the pool and doesn't list any SKIP options.
> >>>>
> >>>> IMO this is nicer because it makes the default test suite output less
> >>>> noisy. The network tests need to only be run periodically
> >>>>
> >>>> Signed-off-by: Cole Robinson <crobinso at redhat.com>
> >>>> ---
> >>>>  tests/conftest.py  | 17 +++++++++++++++++
> >>>>  tests/test_urls.py |  9 ---------
> >>>>  2 files changed, 17 insertions(+), 9 deletions(-)
> >>>>
> >>>> diff --git a/tests/conftest.py b/tests/conftest.py
> >>>> index 8bd7236..5f9cd5b 100644
> >>>> --- a/tests/conftest.py
> >>>> +++ b/tests/conftest.py
> >>>> @@ -13,6 +13,23 @@ def _setup_env():
> >>>>
> >>>>  _setup_env()
> >>>>
> >>>> +
> >>>> +def pytest_addoption(parser):
> >>>> +    parser.addoption("--network-tests", action="store_true",
> >>>> default=False,
> >>>> +            help=("Run osinfo-db network tests. Same tests as
> >>>> triggered "
> >>>> +                  "by setting env variable
> >>>> OSINFO_DB_NETWORK_TESTS"))
> >>>> +
> >>>> +
> >>>> +def pytest_ignore_collect(path, config):
> >>>> +    """
> >>>> +    Entirely skip loading test_urls.py if the option wasn't
> >>>> specified
> >>>> +    """
> >>>> +    run_network = bool(config.getoption("--network-tests") or
> >>>> +            os.environ.get("OSINFO_DB_NETWORK_TESTS"))
> >>>> +    if os.path.basename(str(path)) == "test_urls.py" and not
> >>>> run_network:
> >>>> +        return True
> >>>> +
> >>>> +
> >>>>  # This will trigger some DATA_DIR validation
> >>>>  from . import util
> >>>>  dummy = util
> >>>> diff --git a/tests/test_urls.py b/tests/test_urls.py
> >>>> index 552e6eb..5900f3d 100644
> >>>> --- a/tests/test_urls.py
> >>>> +++ b/tests/test_urls.py
> >>>> @@ -1,15 +1,10 @@
> >>>>  # This work is licensed under the GNU GPLv2 or later.
> >>>>  # See the COPYING file in the top-level directory.
> >>>>
> >>>> -import os
> >>>> -import pytest
> >>>> -
> >>>>  from . import util
> >>>>
> >>>>
> >>>>  @util.os_parametrize('_os', filter_images=True)
> >>>> - at pytest.mark.skipif(os.environ.get('OSINFO_DB_NETWORK_TESTS') is
> >>>> None,
> >>>> -                    reason='Network related tests are not enabled')
> >>>>  def test_images_url(_os):
> >>>>      broken = []
> >>>>      for image in _os.images:
> >>>> @@ -20,8 +15,6 @@ def test_images_url(_os):
> >>>>
> >>>>
> >>>>  @util.os_parametrize('_os', filter_trees=True)
> >>>> - at pytest.mark.skipif(os.environ.get('OSINFO_DB_NETWORK_TESTS') is
> >>>> None,
> >>>> -                    reason='Network related tests are not enabled')
> >>>>  def test_medias_url(_os):
> >>>>      broken = []
> >>>>      for media in _os.medias:
> >>>> @@ -32,8 +25,6 @@ def test_medias_url(_os):
> >>>>
> >>>>
> >>>>  @util.os_parametrize('_os', filter_media=True)
> >>>> - at pytest.mark.skipif(os.environ.get('OSINFO_DB_NETWORK_TESTS') is
> >>>> None,
> >>>> -                    reason='Network related tests are not enabled')
> >>>>  def test_trees_url(_os):
> >>>>      broken = []
> >>>>      for tree in _os.trees:
> >>>
> >>> Cole,
> >>>
> >>> This one didn't work for me at all. I've tried:
> >>> - `make check`
> >>> - `pytest3 tests/test_urls.py`
> >>>
> >>> In both cases the URL tests were not skipped.
> >>>
> >>> Is there some setup needed on my environment?
> >>>
> >>
> >> Ah I didn't test 'make check', sorry, it is indeed busted. I can fix it
> >> easy enough, but...
> >
> > Mind that I've failed to have the tests skipped even running a simple:
> > `pytest-3 tests/test_urls.py`
> > Seems that the methods added are never ever called.
> >
>
> As you mentioned elsewhere, just invoking 'pytest-3' does the right
> thing: plain invocation skips the tests, but calling with
> --network-scripts option or the env variable set invokes the tests.
> That's what I was testing with
>
> The conftest.py bits alter how pytest finds/collects test files. If a
> test file isn't explicitly specified, that code will kick in, and skip
> test_urls unless the option or env var is specified.
>
> pytest-3 tests/test_urls.py is an explicit request to run that file, so
> it skips the file collection logic entirely. That's why the tests are
> always run in your case. Whether that's a bug or a feature is debatable:
> IMO if a user explicitly requests 'hey run test_urls.py' it shouldn't
> also require a special environment variable to actually run, which is
> the current state. It's a minor point though

Hmm. I agree with you in here.
So, yes, we can have this patch in the way it's right now and ...

>
> >>
> >> By having 'make check' run each test file individually with pytest-3 we
> >> lose some of the flexibility of pytest. I know you implemented it that
> >> way to have individual test file logfiles, which doesn't seem to have
> >> first class support in pytest
> >>
> >> Is the multiple log files specifically something you like, or were you
> >> just aiming to reproduce standard 'make check' behavior? Would just one
> >> log file suit your needs?
> >
> > I was just trying to mimic`make check` behaviour. I don't have any
> > strong feelings for changing it to whatever is more natural with
> > regards to using pytest. So, go ahead. :-)
> >
>
> Okay I'll play with it and send a v2 including some version of this patch
>

... and just have our Makefile calling pytest-3 instead of calling
pytest-3 /path/to/each/file

> Thanks,
> Cole

Best Regards,
-- 
Fabiano Fidêncio




More information about the Libosinfo mailing list