Test failing on some architectures

[juliangilbey]

mystic/tests/test_collapse.py fails at line 177 on several architectures on Debian (see https://tracker.debian.org/pkg/mystic): these include armel, armhf, i386 and ppc64el. Running the test manually, as the error message is quite uninformative, we find that on i386:

>>> ct.collapse_position(m, mask=((0,1,2),((0,1),(1,2),(0,2))))
((1, 1, 2), ((0, 2), (0, 1), (1, 2)))
>>> x = ct.collapse_position(m, mask=((0,1,2),((0,1),(1,2),(0,2))))
>>> x == ((1,1,2), ((0,1),(0,2),(1,2)))
False

The order of the elements in the second tuple has changed.

The test succeeds on other architectures (amd64, arm64, s390x), so it seems that there is something flaky about the order of the result here.

[mmckerns]

Hmmm... that's odd.

[juliangilbey]

Quite! I don't understand what the code's doing (or trying to do) well enough to have any idea what causing this issue, or whether it is a problem.

(If the tests were written using pytest or even unittest, I would be able to test all of the tests and report on any others that also fail. As it is, though, it gives up at the first error, so there may well be other failing tests hiding after this one...)

[mmckerns]

the order of the results is actually irrelevant, so I think I could compare a set so order is not important.