test_time failure in CI logs

[WillAyd]

The following failure has shown up a couple times in CI today. Haven't looked in detail but assume it to be an unreliable test.

https://dev.azure.com/pandas-dev/pandas/_build/results?buildId=9836&view=logs&jobId=a69e7846-138e-5465-0656-921e8964615b&taskId=56da51de-fd5a-5466-5244-b5f65d252624&lineStart=42&lineEnd=95&colStart=1&colEnd=64

=================================== FAILURES ===================================
_____________________________ TestTSPlot.test_time _____________________________
[gw0] linux -- Python 3.6.6 /home/vsts/miniconda3/envs/pandas-dev/bin/python

self = <pandas.tests.plotting.test_datetimelike.TestTSPlot object at 0x7f2889f69978>

    @pytest.mark.slow
    def test_time(self):
        t = datetime(1, 1, 1, 3, 30, 0)
        deltas = np.random.randint(1, 20, 3).cumsum()
        ts = np.array([(t + timedelta(minutes=int(x))).time() for x in deltas])
        df = DataFrame({'a': np.random.randn(len(ts)),
                        'b': np.random.randn(len(ts))},
                       index=ts)
        fig, ax = self.plt.subplots()
        df.plot(ax=ax)
    
        # verify tick labels
        fig.canvas.draw()
        ticks = ax.get_xticks()
        labels = ax.get_xticklabels()
        for t, l in zip(ticks, labels):
            m, s = divmod(int(t), 60)
            h, m = divmod(m, 60)
            rs = l.get_text()
            if len(rs) > 0:
                if s != 0:
                    xp = time(h, m, s).strftime('%H:%M:%S')
                else:
                    xp = time(h, m, s).strftime('%H:%M')
                assert xp == rs
    
        # change xlim
        ax.set_xlim('1:30', '5:00')
    
        # check tick labels again
        fig.canvas.draw()
        ticks = ax.get_xticks()
        labels = ax.get_xticklabels()
        for t, l in zip(ticks, labels):
            m, s = divmod(int(t), 60)
            h, m = divmod(m, 60)
            rs = l.get_text()
            if len(rs) > 0:
                if s != 0:
                    xp = time(h, m, s).strftime('%H:%M:%S')
                else:
                    xp = time(h, m, s).strftime('%H:%M')
>               assert xp == rs
E               AssertionError: assert '01:06:40' == '03:40'
E                 - 01:06:40
E                 + 03:40

[gfyoung]

Yeah...it seems to fail on and off.

[WillAyd]

@gfyoung can you reproduce locally by chance? Tried a few times on macOS with no luck curious if you fared any better

[gfyoung]

Of course not. Why would flaky CI make our lives easy? 😂

[gfyoung]

But actually, though, unless we can figure something out now, I think it's best skipping this test (for now) so that we can merge other PR's and return to this before the 0.25.0 release.

[WillAyd]

Hmm not sure about skipping the whole thing. I think we could at least salvage everything up until the xlim change and maybe skip everything thereafter for the time being.

Could also be explicit about the data to use rather than relying on random

[gfyoung]

I think we could at least salvage everything up until the xlim change and maybe skip everything thereafter for the time being.

That's fair.

Could also be explicit about the data to use rather than relying on random

Possibly, could try that and see what happens.

[mroeschke]

Looks like this test is no longer xfailed (and possibly the CI has recovered?) We can reopen if we see it again