This partially address #1521. I still don't see the grouputils failures. Is that pandas 0.11.0? If so, we can close that because we haev 0.12.0 as minimum now.
TST: Don't pass DataFrame to numpy
TST: Dont set_index with PeriodIndex. Known bug.
TST: Not a failure if joblib is installed
Should the GMM tests marked with np.testing.decorators.knownfailureif give an error? They do for me locally. It looks like the decorator works, but instead of marking as a knownfail it marks as an error for me.
When I ran nosetests xxx on the commandline, then I always got the knownfailures as errors. IIRC
But they don't show up as errors in the python shell.
I haven't checked in a while but it was because of the way numpy.testing and nose interacted.
TST: Use NumpyVersion
TST: Don't use generator. Closes #1145.
Yeah, that's what I see too. Not ideal for local testing.
I removed the generator in the test_multi. All of the tests still run but it doesn't list the whole shebang if we run the tests in verbose mode.
Info on knownfailureif
I'm punting on that. We should probably do conditional skips. I could've sworn we had known failures before that didn't raise an error from the command line.
I think we only use skipif to avoid knownfailure.
TST: knownfailureif -> skipif
Changed knownfailure to skipif.