Update for requirements

[dopplershift]

• Update pinning for pooch 0.1.* I think we can trust bug fix releases. 😁 Also, apparently our environment.yml was already doing this (by accident IMO), but now setup.py will allow too.
• Tweak tests to pass on matplotlib 3.0. Small bug in a certain path of MultipleLocator (BUG: Fix typo in view_limits() for MultipleLocator matplotlib/matplotlib#12313) broke the tests--but this path is only triggered with non-default matplotlib style (which happens to be used by pytest-mpl). Just adjusting our tests is good enough to avoid the bug.

I expect our tests to pass with this PR.

[jrleeman]

Looks like a cross section issue in the gallery:

Warning, treated as error:
/home/travis/build/Unidata/MetPy/examples/cross_section.py failed to execute correctly: Traceback (most recent call last):
  File "/home/travis/build/Unidata/MetPy/examples/cross_section.py", line 93, in <module>
    inline_spacing=8, fmt='%i', rightside_up=True, use_clabeltext=True, alpha=0.6)
  File "/home/travis/virtualenv/python3.6.3/lib/python3.6/site-packages/matplotlib/pyplot.py", line 2496, in clabel
    return gca().clabel(CS=CS, *args, **kwargs)
TypeError: clabel() got multiple values for argument 'CS'

[dopplershift]

Ah, that one is matplotlib breaking cartopy (matplotlib/matplotlib#12288). I think we can fix it up by not using pyplot.

[jrleeman]

Looks like a mpl 3 hold machine issue from scipy.spatial 2d tri plotting. Relevant scipy code:

 if tri.points.shape[1] != 2:
        raise ValueError("Delaunay triangulation is not 2-D")

    ax.plot(tri.points[:,0], tri.points[:,1], 'o')
    ax.triplot(tri.points[:,0], tri.points[:,1], tri.simplices.copy())

    _adjust_bounds(ax, tri.points)

[dopplershift]

Right, saw that last night--must have missed one. Maybe now it will pass? 🙏

[jrleeman]

[jrleeman]

Still a no go

[dopplershift]

Once more into the breach...

[jrleeman]

Got it!