On Mon, Jan 18, 2016 at 04:03:34PM +0100, Andreas Enge wrote: > On Mon, Jan 18, 2016 at 03:53:05PM +0100, Cyril Roelandt wrote: > > Yes, a bunch of packages will have to be rebuilt, but there is nothing > > we can do about this, I'm afraid. > > I think what Efraim meant was that package-with-python2 does not work when > dealing with packages depending on python2-jsonschema. They will then not > depend on python2-jsonschema (built with python2-functools32), but on the > recursively rewritten (package-with-python2 python-jsonschema), which does > not include python2-functools32 in its inputs. We already have this rather > unpleasant problem around the numpy/matplotlib universe.
Shall we open a bug for this? It was discussed in #22013 [0] but that specific issue was solved and the bug was closed in expectation of a more general bug report. > > A cheating workaround would be to also add python-functools32 to the inputs > of python-jsonschema, assuming that it would be superfluous, but not > impossible, that is, not creating a conflict. This "solution" is already in place throughout python.scm. I don't think that python-setuptools is really required for many of the python-3 packages that call it as a native-input, but it eases the python-2 translation. [0] http://debbugs.gnu.org/cgi/bugreport.cgi?bug=22013 > > Andreas > >
