Travis tests are failing

Travis tests have started to fail. It looks like this is due to a string-formatting issue:

======================================================================
ERROR: Testing reference/alchemical overlap for no alchemical dispersion LennardJonesFluid with dispersion correction, annihilated sterics
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/travis/miniconda/envs/test/lib/python3.6/site-packages/nose/case.py", line 197, in runTest
    self.test(*self.arg)
  File "/home/travis/build/choderalab/openmmtools/openmmtools/tests/test_alchemy.py", line 1164, in overlap_check
    'du mean {:.3f} kT stddev {:.3f} kT').format(DeltaF, dDeltaF, Neff, g, du_n.mean(), du_n.std())
ValueError: Unknown format code 'f' for object of type 'str'
-------------------- >> begin captured logging << --------------------

[jchodera]

@jaimergp : Thanks for catching this!

Yikes! There’s no way we should have broken the API in a non-backward-compatible manner in a minor bugfix release! I’ll create an issue on the pymbar issue tracker and we can revert that ASAP.

[jaimergp]

Yes, I think that could be indeed the reason.

This is the test code that is failing. As suspected, this will unpack the dictionary keys, instead of the values. That’s why you get str objects:

In [2]: d = {'a': 0, 'b': 1}

In [3]: x, y = d

In [4]: x
Out[4]: 'a'

In [5]: y
Out[5]: 'b'