Hi Rebecca > The binNMUs (to add python3.8 support) of pandas and statsmodels failed. > I think this will make them work (but haven't tried it): > > - Apply the (existing) #943418 fix to python-xlrd. > - Build matplotlib, pandas and statsmodels in that order. (ben can't > work this out because the declared dependencies are circular, but I > think the actual strict test dependencies only go that way.)
Thanks for the info. I was able to build matplotlib first, then python-xlrd, then pandas. Pandas has already finished on some of the faster architectures (amd64/i386/ppc64el/s390x) and I scheduled statsmodels there, but it still fails. I'm not sure if this is due to python 3.8, or the new numpy. Regards Graham