MNT: check CIs

[GuillaumeFavelier]

I am just doing a routine check of the CIs before the release. I noticed failures on LinuxConda and Linux Python37 jobs. Hopefully, it is just something minor like adjusting timeouts etc.

2021-12-14T13:29:57.0447323Z =================================== FAILURES ===================================
2021-12-14T13:29:57.0448358Z ____________________ test_background_plotting_add_callback _____________________
2021-12-14T13:29:57.0449063Z 
2021-12-14T13:29:57.0449873Z qtbot = <pytestqt.qtbot.QtBot object at 0x7fdc342a91f0>
2021-12-14T13:29:57.0451088Z monkeypatch = <_pytest.monkeypatch.MonkeyPatch object at 0x7fdc342a9880>
2021-12-14T13:29:57.0451952Z plotting = None
2021-12-14T13:29:57.0452429Z 
2021-12-14T13:29:57.0455482Z     def test_background_plotting_add_callback(qtbot, monkeypatch, plotting):
2021-12-14T13:29:57.0456299Z         class CallBack(object):
2021-12-14T13:29:57.0456782Z             def __init__(self, sphere):
2021-12-14T13:29:57.0457495Z                 self.sphere = sphere
2021-12-14T13:29:57.0457850Z     
2021-12-14T13:29:57.0458194Z             def __call__(self):
2021-12-14T13:29:57.0458591Z                 self.sphere.points *= 0.5
2021-12-14T13:29:57.0458929Z     
2021-12-14T13:29:57.0459246Z         update_count = [0]
2021-12-14T13:29:57.0459661Z         orig_update_app_icon = BackgroundPlotter.update_app_icon
2021-12-14T13:29:57.0460043Z     
2021-12-14T13:29:57.0460381Z         def update_app_icon(slf):
2021-12-14T13:29:57.0460776Z             update_count[0] = update_count[0] + 1
2021-12-14T13:29:57.0461195Z             return orig_update_app_icon(slf)
2021-12-14T13:29:57.0461537Z     
2021-12-14T13:29:57.0476893Z         monkeypatch.setattr(BackgroundPlotter, 'update_app_icon', update_app_icon)
2021-12-14T13:29:57.0477970Z         plotter = BackgroundPlotter(
2021-12-14T13:29:57.0478532Z             show=False,
2021-12-14T13:29:57.0479366Z             off_screen=False,
2021-12-14T13:29:57.0480181Z             title='Testing Window',
2021-12-14T13:29:57.0480924Z             update_app_icon=True,  # also does add_callback
2021-12-14T13:29:57.0482362Z         )
2021-12-14T13:29:57.0483846Z         assert plotter._last_update_time == -np.inf
2021-12-14T13:29:57.0484714Z         sphere = pyvista.Sphere()
2021-12-14T13:29:57.0485275Z         mycallback = CallBack(sphere)
2021-12-14T13:29:57.0485819Z         plotter.add_mesh(sphere)
2021-12-14T13:29:57.0486401Z         plotter.add_callback(mycallback, interval=200, count=3)
2021-12-14T13:29:57.0486986Z     
2021-12-14T13:29:57.0487506Z         # check that timers are set properly in add_callback()
2021-12-14T13:29:57.0488132Z         assert_hasattr(plotter, "app_window", MainWindow)
2021-12-14T13:29:57.0488861Z         assert_hasattr(plotter, "_callback_timer", QTimer)
2021-12-14T13:29:57.0489489Z         assert_hasattr(plotter, "counters", list)
2021-12-14T13:29:57.0490006Z     
2021-12-14T13:29:57.0490522Z         window = plotter.app_window  # MainWindow
2021-12-14T13:29:57.0491115Z         callback_timer = plotter._callback_timer  # QTimer
2021-12-14T13:29:57.0492261Z         counter = plotter.counters[-1]  # Counter
2021-12-14T13:29:57.0492851Z     
2021-12-14T13:29:57.0493360Z         # ensure that the window is showed
2021-12-14T13:29:57.0493911Z         assert not window.isVisible()
2021-12-14T13:29:57.0494476Z         with qtbot.wait_exposed(window):
2021-12-14T13:29:57.0495229Z             window.show()
2021-12-14T13:29:57.0495736Z         assert window.isVisible()
2021-12-14T13:29:57.0496311Z         assert update_count[0] in [0, 1]  # macOS sometimes updates (1)
2021-12-14T13:29:57.0497193Z         # don't check _last_update_time for non-inf-ness, won't be updated on Win
2021-12-14T13:29:57.0498159Z         plotter.update_app_icon()  # the timer doesn't call it right away, so do it
2021-12-14T13:29:57.0499032Z         assert update_count[0] in [1, 2]
2021-12-14T13:29:57.0499842Z         plotter.update_app_icon()  # should be a no-op
2021-12-14T13:29:57.0500475Z         assert update_count[0] in [2, 3]
2021-12-14T13:29:57.0501105Z         with pytest.raises(ValueError, match="ndarray with shape"):
2021-12-14T13:29:57.0501887Z             plotter.set_icon(0.)
2021-12-14T13:29:57.0502684Z         # Maybe someday manually setting "set_icon" should disable update_app_icon?
2021-12-14T13:29:57.0503351Z         # Strings also supported directly by QIcon
2021-12-14T13:29:57.0503975Z         plotter.set_icon(os.path.join(
2021-12-14T13:29:57.0504575Z             os.path.dirname(pyvistaqt.__file__), "data",
2021-12-14T13:29:57.0505324Z             "pyvista_logo_square.png"))
2021-12-14T13:29:57.0505815Z     
2021-12-14T13:29:57.0506343Z         # ensure that self.callback_timer send a signal
2021-12-14T13:29:57.0506997Z         callback_blocker = qtbot.wait_signals([callback_timer.timeout], timeout=300)
2021-12-14T13:29:57.0507813Z         callback_blocker.wait()
2021-12-14T13:29:57.0508540Z         # ensure that self.counters send a signal
2021-12-14T13:29:57.0509179Z         counter_blocker = qtbot.wait_signals([counter.signal_finished], timeout=700)
2021-12-14T13:29:57.0509803Z >       counter_blocker.wait()
2021-12-14T13:29:57.0510465Z E       pytestqt.exceptions.TimeoutError: Emitted signals: None. Missing: [signal_finished()]

[codecov]

Codecov Report

Merging #136 (df8696a) into main (d81885c) will decrease coverage by 0.91%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##             main     #136      +/-   ##
==========================================
- Coverage   98.17%   97.26%   -0.92%     
==========================================
  Files           8        8              
  Lines         658      658              
==========================================
- Hits          646      640       -6     
- Misses         12       18       +6     

[GuillaumeFavelier]

Here is the situation:

1. CI / macos-latest / conda / PySide2 is stuck in limbo and I have no idea why.
2. From time to time, Linux Python36 cannot pass test_background_plotting_add_callback(). I tried this one locally and I cannot reproduce.
3. Occasionally, CI / macos-latest / pip / PySide2 / crashes. Still it happens too much to my liking. But I noticed that those crashes appears during the calls of qtbot.wait_signals() (not sure how it could be useful for now).

[GuillaumeFavelier]

To be precise, 3) does not happen only in qtbot.wait_signals() but on the explicit call to wait(). My first idea is then to refactor those to use the context manager instead.

[GuillaumeFavelier]

After trying multiple things:

• show=False
• off_screen=False
• auto_update=False
• force _render() or not
• without empty_scene()

It's not entirely clear why the crash does happen. Although I learned where it comes from or rather where it is detected:
the __exit__() function of the _WaitWidgetContextManager class in pytest-qt (I inspected it and I do not see anything wrong here).

I still think it is somehow related to QTimers and/or rendering paintEvent()/render().

---

To move forward with the PR:

• I changed the Qt binding to PyQt5 on MacOS. After all, PySide2 is not updated on PyPi since last year (11/20/2020). And this solves 1. and 3.
• 2. does still happens occasionally but restarting should fix this

Apart from that, I modified test_background_plotting_add_callback() to test only one timer at a time, added stages on Azure to support [skip azp] and updated the GitHub Actions workflow.

This is ready to go from my end @pyvista/developers

[larsoner]

Any way to keep testing pyside2 on Linux as well? Maybe the matrix should be macos-pyqt5, linux-pyqt5, linux-pyside2?

[GuillaumeFavelier]

PySide2 is still tested on linux and windows for PyPI jobs.
PyQt5 is set for conda jobs though.

[larsoner]

Oh okay, I misread the config -- great!

[larsoner]

Otherwise LGTM. Feel free to merge if you don't think PySide2 on linux is a good idea, or want to tackle it in another PR @GuillaumeFavelier !

[larsoner]

... on second thought, I'll go ahead and merge so you can push out a release @GuillaumeFavelier , let's tackle the linux+pyside2 in another PR