Martin, Do you think that a single package (melina) merits a change in the gcc42 and gcc43 packaging? I could move the symlinks in %p/bin for gcc42 and gcc43 out into a gcc42-dev and gcc43-dev package (allowing the gcc42 and gcc43 packages to co-exist). Then packages would BuildDepends on gcc4X-dev instead of gcc4X. Of course melina would have to be patched to directly access the compilers from %p/lib/gcc4.X/bin and not %/bin. Jack ps It will be complex to pull off cleanly as I need to somehow make gcc42-dev Provides gcc42 (<= 4.2.3-1000) so limit the breakage in building existing packages.
On Sat, Mar 08, 2008 at 11:07:56AM +0100, Martin Costabel wrote: > Jack Howarth wrote: >> The info files for melina, pgplot and r-base >> all contain incorrect Depends for gcc42. These packages > > I can only speak for melina, but I can tell you that the Depends is not > "incorrect". Melina needs a fortran compiler at runtime, and it needs the > same one it was built with. There are good reasons why compilers are not > "BuildDependsOnly: true". > >> should be using a BuildDepends instead and only a >> Depends for gcc42-shlibs. There is no reason to Depends >> on gcc42 unless the package will directly call the gcc42 >> compilers. This should be fixed so that users can install >> gcc43 with deinstalling these programs. The best course >> of action though would be to migrate all packages to >> gcc43 and obsolete the gcc42 package in unstable. >> I don't see a clean mechanism (or clear reason) to support both gcc42 and >> gcc43. Specifically, packages >> with libraries built with gcc42 or gcc43 would require >> variants (which are pointless since code generation in >> gcc43 is much improved). > > If gcc43 cannot coexist with gcc42, fine (although 99% of all files are in > versioned directories already). One just has to keep in mind that there are > a couple of packages that need to be updated with new incompatible > revisions of gcc4X. We have this situation with other compilers, too, ocaml > for example, where every new revision requires that a couple of other > packages need to be updated and rebuilt. > > I'll update melina correspondingly (after checking that it works, which > will take some time). > > The maintainer of pgplot and r-base has given up, unfortunately, so that I > don't know who will do the necessary for these. > > -- > Martin > ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ Fink-devel mailing list Fink-devel@lists.sourceforge.net http://news.gmane.org/gmane.os.apple.fink.devel