On 2013/01/09 01:27, frantisek holop wrote: > hmm, on Tue, Jan 08, 2013 at 09:09:44PM +0000, Stuart Henderson said that > > > searching for it in e.g. devel/ > > > reveals only a handful of py-* that uses it. > > > > > > the python documentation seems to imply that > > > bytecode is arch independent.. > > > > I would generally avoid it for python things really, they are awkward > > enough as it is without having to think about more things ;) > > care to go into this a bit more? :] what's awkward?
current peeves: the continual plist fiddling, and some things are starting to show up with things like this in setup.py... import distribute_setup distribute_setup.use_setuptools() ...which doesn't work nicely with --single-version-externally-managed. the reliance on timestamps for pyc files is a pain (though that's hacked around in pkg_add tie_files now so it's just an inefficiency rather than causing a visible problem any more). hmm what else.. oh yes, all the REVISION bumps when we switch version.... > i was thinking, if it was that useful, it would be > already in pyhon.port.mk... there are quite a few ports which have a mixture of python and other things, I don't think there are really any good heuristics to avoid using it incorrectly on these.. I'd rather get rid of PKG_ARCH=* completely (and I'm probably about the only person who made serious use of this recently [to help with arm package builds]) than have it used wrongly and produce broken packages.