Recently I've been looking at errors in the matplotlib test suite when building on Fedora's ppc64 arch, which is big-endian. I've sent a couple of PRs:

#7768
#7781

to fix some integer type problems in the freetype wrapper, which cuts the error count down from hundreds to 38, but I don't think I can fix the remaining issues, so I'm filing tickets for them.

Several of the remaining failed tests fall into the general category 'colorbar generation does not work right at all' - the following tests all fail:

matplotlib.tests.test_axes.test_contour_colorbar.test
matplotlib.tests.test_colorbar.test_colorbar_closed_patch.test
matplotlib.tests.test_colorbar.test_colorbar_extension_length.test
matplotlib.tests.test_colorbar.test_colorbar_extension_shape.test
matplotlib.tests.test_colorbar.test_colorbar_positioning.test
matplotlib.tests.test_colorbar.test_colorbar_single_scatter.test
matplotlib.tests.test_colorbar.test_gridspec_make_colorbar.test

All the failures are image comparison failures. I'll attach some of the failures here; the failure mode seems to be basically the same in all cases. The colorbar box is drawn the right size and shape and in the right position and with the right marks, but no actual colors are drawn inside the box - it's just white inside.

This occurs building from the 2.0.0rc2 source using the Fedora spec file, numpy will be 1.11.2 I believe. No warning messages are logged during the test run that relate to this issue (I confirmed by cutting the test run down to just test_colorbar_tests and when I do that, there are no error/warning messages logged).

cbar_locationing GOT:
cbar_locationing EXPECTED:
cbar_sharing GOT:
cbar_sharing EXPECTED:
cbar_with_orientation GOT:
cbar_with_orientation EXPECTED:
cbar_with_subplots_adjust GOT:
cbar_with_subplots_adjust EXPECTED:

That's not all of them, but I guess enough to get the idea.

@QuLogic (who seems interested in this area)