Hi Andreas, just a few comments -- I will not have time at the weekend to fix it myself.
The problems seem simple: First, buth RC bugs could be merged IMO since they point to the same problem. The two arrays are only "nearly identical"; unclear why, but I would just change the equal to a "almost-equal" and report this upstream. #849368 documents another failure as well; IMO the according test could be extended to compare either with "blue", or with the 4-tuple. The latter seems to be returned by matplotlib 2 (where the get_axis_bg() is deprecated anyway). IMO both changes could be done quickly, but my sun celebrates 3rd birthday tomorrow, that is more important :-) Cheers Ole
