Some ports could use a "depends_test" phase, instead of sticking it in depends_build or even worse depends_lib.
This causes certain ports like py*-cairo to uselessly pull in heavy ports like py*-numpy (and atlas and gcc*) "py26-cairo depends on py26-numpy" distfiles: port py26-numpy: numpy-1.5.0.tar.gz 2276377 bytes py26-numpy: 2.2M port atlas: atlas3.8.3.tar.bz2 2527756 bytes port atlas: lapack-3.2.2.tgz 4819849 bytes atlas: 7.0M port gcc44: gcc-core-4.4.5.tar.bz2 26356188 bytes port gcc44: gcc-fortran-4.4.5.tar.bz2 1414484 bytes port gcc44: gcc-g++-4.4.5.tar.bz2 5946112 bytes port gcc44: gcc-java-4.4.5.tar.bz2 18488619 bytes port gcc44: gcc-objc-4.4.5.tar.bz2 240383 bytes gcc44: 50.0M packages: 2,5M py26-numpy-1.5.0_1+atlas+gcc44.x86_64.tgz 1,7M py26-numpy-1.5.0_1+atlas+gcc44.x86_64.txz 6,6M atlas-3.8.3_4+gcc44.x86_64.tgz 2,3M atlas-3.8.3_4+gcc44.x86_64.txz 88M gcc44-4.4.5_0.x86_64.tgz 63M gcc44-4.4.5_0.x86_64.txz Even though numpy is _only_ used to create test data! At least with build dependencies (instead of runtime), you can uninstall them again after building the ports... Matters more when installing from binaries, but anyway. http://trac.macports.org/ticket/27369 has the patches --anders _______________________________________________ macports-dev mailing list [email protected] http://lists.macosforge.org/mailman/listinfo.cgi/macports-dev
