Note: the version shipped in xcode10 is 3.7.

Coverage decreased (-0.03%) to 94.547% when pulling 57ba55b on rht:macOS into 6ba4ba0 on QuantEcon:master.

The first commit setup works fine (by install hdf5 via homebrew instead of conda), however, the machine was too slow to run the timing test. using conda is much faster

thanks @rht -- this is great. I will check through the diff tomorrow and merge.

OK, I found the test_rotated_hyper_ellipsoid to be non-deterministic though, in that sometimes on darwin it is fulfilled/failed, in that I can't write it as a test-expect-fail.

hey @rht we have had tests running on linux for quite some time. I'm not sure why you're getting non-deterministic results. Are you sure its not just a tolerance issue?

Otherwise, I can't explain the error in the commit "TestTicTacToc: Skip test_loop and test_timer for darwin". There was an error of ~1e-5.

So, the test flake disappeared in the dust just like that.

thanks @rht for the PR! I will merge this and take a look at tic tac toc issue for darwin when I get a chance. It's tested on the linux side anyway.