Hello @shoyer! Thanks for updating this PR. We checked the lines you've touched for PEP 8 issues, and found:

There are currently no PEP 8 issues detected in this Pull Request. Cheers! :beers:

Comment last updated at 2019-07-04 20:02:40 UTC

Codecov Report

:exclamation: No coverage uploaded for pull request base (master@681ec0e). Click here to learn what that means. The diff coverage is n/a.

```diff @@ Coverage Diff @@

master #3070 +/-

======================================== Coverage ? 94.8%
======================================== Files ? 67
Lines ? 12876
Branches ? 0
======================================== Hits ? 12207
Misses ? 669
Partials ? 0 ```

I'm referring to the tests running on CI, e.g., "Linux py36-flakey" on Azure. If you look at the full logs from this PR (which I made verbose) and scroll down you see: xarray/tests/test_backends.py::TestPydapOnline::test_cmp_local_file SKIPPED [ 13%] xarray/tests/test_backends.py::TestPydapOnline::test_compatible_to_netcdf SKIPPED [ 13%] xarray/tests/test_backends.py::TestPydapOnline::test_dask SKIPPED [ 13%] xarray/tests/test_backends.py::TestPydapOnline::test_session SKIPPED [ 13%] So for some reason it's not doing flakey tests, even though in theory I'm passing the right flags for it: https://github.com/pydata/xarray/blob/681ec0e84af7c155b16ef46122e2a6cea64b585d/ci/azure/unit-tests.yml#L14

Actually, I definitely screwed this up. I should be using $PYTEST_EXTRA_FLAGS, not $EXTRA_FLAGS. Let me try fixing that...

Somewhat weirdly. I get different results; could you confirm yours?

``` pytest --run-network-tests

Results (175.99s): 7266 passed 5 xpassed 19 xfailed 1092 skipped ```

``` pytest

Results (165.58s): 7264 passed 5 xpassed 19 xfailed 1094 skipped ```

Also of note: --run-network also works, as does --run-n

That does sound suspicious! I'm on vacation with bad WiFi but let me confirm in the next few days, if that's ok

Pytest reports the exact same number of tests passed, both with and without --run-network-tests. That seems suspicious to me.

Let's def change to the most reliable & standard approach here. I think I originally got this from https://docs.pytest.org/en/latest/example/simple.html#control-skipping-of-tests-according-to-command-line-option but if this is minority let's make the switch

What makes us think the existing approach wasn't working? (not saying that it did work in CI! It does seem to work locally though)

--run-network works fine

``` I  test  ~/w/xarray   *  pytest --run-network xarray/tests/test_tutorial.py 192ms  Tue Jul 2 02:45:19 2019 Test session starts (platform: darwin, Python 3.7.3, pytest 4.6.3, pytest-sugar 0.9.2) rootdir: /Users/maximilian/workspace/xarray, inifile: setup.cfg plugins: xdist-1.29.0, forked-1.0.2, sugar-0.9.2, regtest-1.4.1, testmon-0.9.16, pycharm-0.5.0, celery-4.3.0 collecting ... xarray/tests/test_tutorial.py ✓✓ 100% ██████████ =============================================================================================== warnings summary =============================================================================================== xarray/tests/test_tutorial.py::TestLoadDataset::test_download_from_github /usr/local/Cellar/python/3.7.3/Frameworks/Python.framework/Versions/3.7/lib/python3.7/importlib/_bootstrap.py:219: RuntimeWarning: numpy.ufunc size changed, may indicate binary incompatibility. Expected 192 from C header, got 216 from PyObject return f(args, *kwds)

xarray/tests/test_tutorial.py::TestLoadDataset::test_download_from_github xarray/tests/test_tutorial.py::TestLoadDataset::test_download_from_github_load_without_cache /usr/local/lib/python3.7/site-packages/_pytest/python.py:174: RuntimeWarning: deallocating CachingFileManager(<function _open_scipy_netcdf at 0x114e29840>, '/Users/maximilian/.xarray_tutorial_data/tiny.nc', mode='r', kwargs={'mmap': None, 'version': 2}), but file is not already closed. This may indicate a bug. testfunction(**testargs)

-- Docs: https://docs.pytest.org/en/latest/warnings.html

Results (15.92s): 2 passed # <-- runs N  test  ~/w/xarray   *  pytest xarray/tests/test_tutorial.py 17.6s  Tue Jul 2 02:49:10 2019 Test session starts (platform: darwin, Python 3.7.3, pytest 4.6.3, pytest-sugar 0.9.2) rootdir: /Users/maximilian/workspace/xarray, inifile: setup.cfg plugins: xdist-1.29.0, forked-1.0.2, sugar-0.9.2, regtest-1.4.1, testmon-0.9.16, pycharm-0.5.0, celery-4.3.0 collecting ... xarray/tests/test_tutorial.py ss 100% ██████████

Results (3.06s): 2 skipped # <-- skips ```

Hmm... maybe it does work, at least in this. But we definitely aren't running the network tests, even with these changes (e.g., look at the TestPydapOnline tests). I'm not quite sure why.

Hmmm, this is my code. Checking what I missed. We use the similar code internally and it works (I think...)

Do we have any flaky tests? I only see one that's marked, and that's skipped anyway (maybe because of this problem):

``` rg flaky

azure-pipelines.yml 24: pytest_extra_flags: --run-flaky --run-network-tests

setup.cfg 13: flaky: flaky tests

conftest.py 8: parser.addoption("--run-flaky", action="store_true", 9: help="runs flaky tests") 16: if not config.getoption("--run-flaky"): 17: skip_flaky = pytest.mark.skip( 18: reason="set --run-flaky option to run flaky tests") 20: if "flaky" in item.keywords: 21: item.add_marker(skip_flaky)

doc/whats-new.rst 1847:- Enhanced tests suite by use of @slow and @flaky decorators, which are 1848: controlled via --run-flaky and --skip-slow command line arguments

xarray/tests/init.py 112:flaky = pytest.mark.flaky

xarray/tests/test_plot.py 30:@pytest.mark.flaky 31:@pytest.mark.skip(reason='maybe flaky') ```