On 2010/09/15 11:53, Eric Faurot wrote: > On Wed, Sep 15, 2010 at 10:18:01AM +0100, Federico G. Schwindt wrote: > > Hi, > > > > New day new diff. Move things to WRKDIR to avoid warnings during plist > > generation reported by r...@. > > While I was here also change how bad eggs are handled to avoid warnings > > and to allow make clean to run as a regular user. Tested with py-urwid, > > py-cheetah and py-distutils-extra. > > Comments? OK to individual pieces or the whole welcome. > > I am not fiercely opposed to these changes, but 2 comments though: > > About the MODPY_JSON thing. Shouldn't it be handled per-port where > needed? I am not quite sure what is the point of trying to add a > pseudo-"backward compatibility" hack in the infrastructure, when we > now moved to 2.6 anyway. Especially for something like json, which > sounds very anecdotal.
This would still be done per-port, you would set ${MODPY_JSON} in the dependency list, and the module would take care of whether to populate the variable. Saves doing the ".if MODPY_VERSION < 2.6" dance in every port that needs it. > About the egg thing (I hate them, I am disgusted at the farce python > has turned into wrt to packaging). How about simply adding setuptools > unconditionnally as a BUILD_DEPENDS for all python modules? Would that > really hurt? It needs to be RUN_DEPENDS too. It wouldn't really hurt but, well, the people who end up actually doing the package bumps got a bit bored of it by now... and this method gives a clear and simple error message.