On Thu, 2006-07-20 at 15:59 +0200, Nicolas Mailhot wrote: > Le Jeu 20 juillet 2006 15:25, Erwin Rol a écrit : > > On Thu, 2006-07-20 at 15:20 +0200, Nicolas Mailhot wrote: > >> Believe me you're far better off with overlapping packages than a > >> multiplication of "common" packages. > > > > What is the technical reason that overlapping packages are better than > > common packages ? They still would be created from the same source rpm, > > just like the gcc source rpm "creates" a number of rpms. > > common subpackages wouldn't help your conflict a little bit as sanity > would demand i386 and x86_64 require a common subpackage with the same > nevr. So instead of yum barfing because i386 and x86_64 are not in sync > you'd get yum barfing because i386 demands one common package and x86_64 > another. > > rpm does not support producing subpackages with differing arches from the > same srpm, you have to workaround this by launching several builds from > the same srpm (plus spec ugly logic...) this is only done for very special > packages in the buildsys > > you'd get a lot of packages with one or two files in them and needlessly > bloat the package number (which would bloat repodata, make more spec > creation and translation manual work, bloat the dependency graph your rpm > has to manage...) > > and probably a vouple other reasons I forget. > > All this for no win, since your problem is not the way we share files > accross multiarch, but that at a given time i386 and x86_64 package sets > are not consistent with each other. OK that was a long but clear technical explanation, thanks. - Erwin -- fedora-devel-list mailing list fedora-devel-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/fedora-devel-list
