swig-python and python{24,25,31,32}

Michael Dickens michaelld at macports.org
Sun Jan 4 19:10:06 PST 2015

On Sun, Jan 4, 2015, at 09:58 PM, Lawrence Velázquez wrote:
> On Jan 4, 2015, at 3:46 PM, Michael Dickens <michaelld at macports.org>
> wrote:
> > That said, ports that -rely- on swig-python will probably need to be
> > rev-bumped, because they will have a library/libraries linked against
> > the Python library of choice. Once the old Python has been replaced_by
> > the new one, the linkage of the generated swig-python libraries will be
> > invalid; rebuilding those ports should do the trick, I think.
> But how will those dependent ports know what to link against? The old
> Pythons will be replaced, but users' `port select` selections are not
> going to be migrated.

I sincerely hope that -no- ports depend on the selection of "port
select" to work correctly; that would be poor Portfile design. At least
for GNU Radio (and all of its relatives), we specify which Python to
link to, and SWIG just follows what we tell it to do. I don't think SWIG
tries to be smart when determining which Python to use; I think they
require that the command line specify which one. Thus, when Python 2.7
is installed, its library should be specified for SWIG to link against
-- no matter the "port select" setting.

All of the above said, there are likely some ports for which the above
does not hold true, which do depend on "port select" somehow
inadvertently. Stuff happens; we try the best we can and then fix those
that do not work after the change. - MLD

More information about the macports-dev mailing list