On Sat, 2007-03-03 at 00:13 -0600, Tom 'spot' Callaway wrote: > On Sat, 2007-03-03 at 06:20 +0100, Ralf Corsepius wrote: > > On Fri, 2007-03-02 at 11:28 -0500, Jesse Keating wrote: > > > On Friday 02 March 2007 11:17:24 Tom 'spot' Callaway wrote: > > > > I don't see any fault in adding perl-devel to the FC7 buildroot, with > > > > the caveat that it will not be there in FC8+. Thoughts? > > > > > > I'd rather not change things multiple times. > > Well, exactly the reason why I would like to see perl-devel in FC7's > > buildroot. It avoids forcing packagers to change their packages NOW. > > > > Remember: Adding perl-devel to the buildroot affects one person (you). > > Not adding it and affects 100s of packages and dozens of packagers. > > > > > Why does Ralf think it hasn't been baked yet? I'm missing that context. > > > > This change causes most (all?) perl modules and applications using perl > > (comprising noarch packages) to depend on perl's config.h at build time. > > > > The primary question to clarify would be: > > > > What is the cause, why (esp. noarch) perl-modules now require a c-header > > (perl's config.h) and (thereby gcc) to build? [1] > > Well, fwiw, they seem to have required this before, the -devel split > just made it more apparent. I just tried to investigate on of my currently broken packages, and see a file dependency on config.h in ExtUtils::MakeMaker generated Makefiles. i.e. no compilation is involved, and no dependency on gcc or any library, this Makefile simply depends on the _file_ config.h. I haven't yet checked what it is used for. > When I rewrote the perl spec, I didn't make > any code changes, just spec file cleanups. Well, yes, you made an implicit dep visible. The question is "which dependency" and is kind of package split useful (There is no doubt that a split between a "c-devel environment" and a run-time environment is useful.) Ralf
