Tests and testing of MRs before and after the merge.
Currently, we run a set of reduced tests for MRs and once merged then an extended set is run. This is not optimal because (as just happened) problems could be discovered sometimes only after the MR was merged leaving a potentially broken master behind. This is not a good situation.
In addition there are several tests covering the same set of features, e.g. in dune/fem/space/test: test-spaceinterpolation.cc and interpolation.cc Both do the same kind of testing but do not cover the same spaces and only one of them is run in normal testing mode. This is very problematic. Which one is the better one?