On Fri, Sep 07, 2018 at 01:05:56PM +0100, Stuart Henderson wrote:
> On 2018/09/07 01:17, Ayaka Koshibe wrote:
> > Hi,
> > 
> > Removing MODPY_DEFAULT_VERSION_2 since that seems to be preferred at this
> > point in time.
> > 
> > OK?
> > 
> > Thanks,
> > Ayaka
> > 
> > Index: Makefile
> > ===================================================================
> > RCS file: /cvs/ports/net/mininet/Makefile,v
> > retrieving revision 1.10
> > diff -u -p -u -r1.10 Makefile
> > --- Makefile        6 Sep 2018 14:12:04 -0000       1.10
> > +++ Makefile        7 Sep 2018 08:02:05 -0000
> > @@ -21,8 +21,6 @@ MASTER_SITES =    https://github.com/akoshi
> >  
> >  MODULES =          lang/python
> >  MODPY_SETUPTOOLS = Yes
> > -# Does not run with Python 3
> > -MODPY_VERSION =            ${MODPY_DEFAULT_VERSION_2}
> >  BUILD_DEPENDS =            devel/help2man
> >  RUN_DEPENDS =              net/socat \
> >                     net/iperf
> > 
> 
> OK sthen@
> 
> As suggested by jca I'll explain more -
> 
> The reason I don't like this "python 2-only marker" is that ports may
> get updated and add py3 support, but either upstream forgets to add to
> changelog, or the porter doesn't notice it. Having the entry in the port
> Makefile acts as a prompt that it's already been checked and is correct,
> but really it might be long out of date.
> 
> That doesn't apply so much to mininet (as you are involved with the port
> and work on mininet itself I think you will notice!) but when people see
> something in one part of the tree, it gets copied ...
> 
> The time when this setting will be useful is when we switch
> python.port.mk to using 3.x by default, but this will be a lot of work
> in the ports tree (partly due to py2 modules being "unflavoured", partly
> due to the testing needed with the versions in-tree at the time, partly
> due to the whole-tree port changes/bumps needed) so I don't think it
> will happen very soon, and we would need to check the whole tree when
> we make that change.
> 
> For now: I do think that where a particular "standalone" python port
> offers a choice of either python 2 or 3, we should choose python 3
> unless there's a good reason not to. And that having an out-of-sync
> "py2-only" marker gets in the way of this happening in the future.
> 
> (For ports providing python modules, it usually makes sense to
> provide both py2 and py3 versions).

Amen!

-- 
Antoine

Reply via email to