In testing a bleeding edge conda install on my Mac, which includes numpy 1.14, astropy 3.0.3 and matplotlib 2.2.2, lsst_distrib builds fine but when I run lsst_dm_stack_demo I get nearly 100,000 failures, 8000 of which are failing at the 1e-06 level. I have no idea whether this is reasonable but it sounds very large. The detected-sources txt file is attached.
I have not tested ci_hsc or lsst_ci. I will see if I can run lsst_ci on this system.