Some are producing ever-so-slightly nondeterministic results, and others demonstrate a Python 2 vs. 3 issue. While #5307 got most of these stuff, these were punted on because obvious solutions haven't yet been found.

• test_axes_grid1:test_twin_axes_empty_and_removed
• test_axes:test_boxplot
• test_axes:test_boxplot_rc_parameters
• test_axes:test_specgram_freqs
• test_axes:test_specgram_noise
• test_axes:test_specgram_magnitude_freqs
• test_axes:test_hist_steplog

appveyor/windows: as part of #5922 added or increased tolerance for failures on:

• matplotlib.tests.test_axes.test_specgram_freqs.test (RMS 0.042) (x64,35)
• matplotlib.tests.test_axes.test_specgram_freqs.test (RMS 0.042) (x64,35)
• matplotlib.tests.test_axes.test_specgram_magnitude_freqs.test (RMS 0.042) (x64,35)
• matplotlib.tests.test_axes.test_specgram_magnitude_freqs.test (RMS 0.042) (x64,35)

-> currently set to 0.03, set tolerance to 0.05 on windows

• matplotlib.tests.test_patheffects.test_collection.test (RMS 0.006) (x64,35)
• matplotlib.tests.test_patheffects.test_collection.test (RMS 0.008) (x86,27)
• matplotlib.tests.test_patheffects.test_collection.test (RMS 0.012) (x64,27)
• matplotlib.tests.test_patheffects.test_collection.test (RMS 0.012) (x64,34)

This has a black diff, so up the tolerance on windows to 0.013

• matplotlib.tests.test_patches.test_wedge_range.test (RMS 0.059) (x64,27)
• matplotlib.tests.test_patches.test_wedge_range.test (RMS 0.059) (x64,34)
• matplotlib.tests.test_patches.test_wedge_range.test (RMS 0.059) (x86,27)

This looks actually interesting: it seems that only the middle figure in the last row is different

• matplotlib.tests.test_axes.test_specgram_angle_freqs.test (RMS 0.002) (x86,27)

Also looks black, but only on py27/x86...?

• matplotlib.tests.test_triangulation.test_tri_smooth_gradient.test (RMS 0.014) (x64,35)