GWT passes locally, but fails on Travis

[emorway-usgs]

(This issue relates to #74) @rbwinst-usgs discovered that the .sft input is bad in the GWT problem. Upon fixing the input in commit 3628918 (here), rerunning locally, and uploading the newly minted .ucn file as the new .ucn.cmp file, Travis started failing. However, GWT passed locally. As a test, I cloned a fresh copy of the repo that includes both the updated .sft input and .ucn.cmp files, and it passed locally no problem. I'm hoping you have some ideas on what this could be, @langevin-usgs? Have you run into this sort of thing before?

[langevin-usgs]

In compare_concs, try setting verbose=True to see if we can get some indication for what's going on.

[emorway-usgs]

Thanks for the suggestion, @langevin-usgs, the job log is no showing which cells have a difference (search "Maximum concentration difference," it appears there are 4 instances). Would you mind cloning a fresh copy of the repo into a scratch directory and running t000_test.py and t001_test.py on your local machine? The only test active in t001_test.py is GWT, which takes about 12 sec on my machine. I have updated flopy and pymake as I was behind on both. However, I still seem to be passing locally, so knowing how it behaves on a different machine would be helpful.