On 2-nov-2005, at 23:14, Bob Ippolito wrote: > > On Nov 2, 2005, at 2:06 PM, Ronald Oussoren wrote: > >> On 2-nov-2005, at 22:17, Bob Ippolito wrote: >> >>> Speaking of which, the svn trunk of py2app should be able to >>> analyze and write Mach-O headers of any combination of supported >>> architectures now (32-bit of both endians, 64-bit of both >>> endians, and the universal header that wraps them). It's not >>> really tested yet, though. >> >> Cool. Does that mean that if I'd have a universal binary version >> of Python and addon packages I'd end up with a universal binary >> application bundle after running py2app? > > Almost, the bootstrap is still PPC only for now. That's another > issue, because if I ship a universal bootstrap and someone uses it > with some subset of PPC-only modules, then the application will not > work on a x86 machine since it will run the bootstrap native. > > There may be some kind of Info.plist flag that can toggle which is > the preferred arch, that would be ideal, otherwise I will either > have to make py2app smart enough to do something like lipo -thin > and split off everything but the preferred arch.
The easy way out is to ship 3 precompiled bootstrap programs and select the right one based on the architectures supported by included binaries. You might end up with unused fat binaries for parts of the program, but at least it would work. > > Needless to say, universal binary support is going to be a pain! No need to tell me that :-) > > -bob > _______________________________________________ Pythonmac-SIG maillist - Pythonmac-SIG@python.org http://mail.python.org/mailman/listinfo/pythonmac-sig
