On Sat, Feb 10, 2018 at 05:07:03PM -0500, Mark H Weaver wrote: > reopen 23364 > thanks > > Andreas Enge writes: > > > This has probably been fixed in the following commit: > > > > commit 25b2c47d753efd0761a4e16519dce38d828789f5 > > Author: Hartmut Goebel > > Date: Sun Oct 16 18:40:58 2016 +0200 > > > > gnu: python-statsmodels: Fix build > > > > * gnu/packages/statistics.scm (python-statsmodels): [check] set > > PYTHONPATH prior to running tests. > > > > In any case the package did compile in 2017, and was updated later to 0.8.0. > > No, there have been over 200 nondeterministic build failures of > python-statsmodels since that commit, including over 140 failures of > 0.8.0. I usually have to ask Hydra to attempt to rebuild it several > times before it finally succeeds. I wonder if the test suite is really worth running in this case? Maybe we should just report the failures upstream and skip the tests.