On Mon, 12 Jun 2006, yipdw wrote:
Michael Wybrow wrote:
I think having a Universal Binary, or at least an Mactel version of 0.44 is *really* important. The later is a lot more feasible as it only requires someone with an Intel Mac to follow the existing packaging insructions and create a package. A Universal version will realistically require one of us Mac packager to have access to both a PPC and Intel Mac to sort out the process of creating Universal binaries, as well as testing early packages.
I have access to an Intel Mac running Tiger and have managed to get working packages of Inkscape SVN checkouts. If you'd like these, just drop me a line.
If you could use the same process and make a package for 0.44pre4 that would be great! If it looks like it works then I think we should put it up with the other file releases. Then I'll have a go at working on a Universal binary using lipo.
So we can try and make sure the packages are built with the same dependencies, I am building the PPC builds with the following packages (in addition to the standard dependencies) with newest versions of everything from fink's 10.4 unstable tree: libgettext3-dev lcms lcms-shlibs gnome-vfs2-ssl gnome-vfs2-ssl-dev gnome-vfs2-ssl-shlibs pango1-xft2 pango1-xft2-dev pango1-xf2-shlibs boost1.33 boost1.33-shlibs And just using the same steps as mentioned on the CompilingMacOsX wiki page, with just the folowing configure options: ./configure --prefix=/path/to/install/dir --enable-osxapp
My interpretation of the Universal Binary tutorial at developer.apple.com (http://developer.apple.com/opensource/buildingopensourceuniversal.html) tells me that one can generate Universal Binaries with access to only one architecture provided that one also has appropriate Universal versions of all involved shared libraries. Am I reading this right?
Yes, the real problem is needing universal versions of all the dependencies. There is currently no easy way to do this with fink. Unless anyone has a *lot* of time to spend trying to get a full fink tree to compile universally, I think our best alternative is to build each version on a different machine and attempt to merge them.
Cheers, Michael