tis 2016-01-26 klockan 10:18 +0000 skrev Peter Robinson: > > > > So it appears this thread was probably not enough. Which keeps us with > > > > interesting state where mock by default does not install weak > > > > dependencies where Koji installs them. It causes interesting issues already. > > > > > > mock/koji not installing weak dependencies == anything wanting ruby > > > being broken. > > > > > > Reason: "ruby" suggests "rubypick" which suggests "ruby". > > > > > > Packages buildrequire "ruby" but do not get "rubypick" installed (or if > > > they are lucky they get) so are unable to find Ruby because there is no > > > "/bin/ruby" executable. > > > > If ruby needs ruby-pick to work, then ruby-pick must not be a weak > > dependency of ruby, but a hard one. > > > > The koji buildroot really should only install hard dependencies. The > > buildroot is supposed to be the minimal possible set needed to build > > the package. If a package that would be installed as a weak dependency > > of one of the build dependencies is needed to build the package, that > > packa is a build dependency too. > > That's incorrect. EG a package has optional Ruby bindings, it doesn't > need ruby to build but if ruby isn't present the ruby binding sub > package is empty. The buildroot should have all that is needed to > build the desired functionality, nothing more nothing less. > > Peter I agree that it is a reasonable expectation that installing "ruby" should make a /usr/bin/ruby binary available. The fact that it currently doesn't when weak dependencies are not installed is a packaging bug in ruby, which should be addressed. It is not a reason to bloat every koji build root for every build of every package in Fedora with every weak dependency of every build dependency of the package being built. Mattias
<<attachment: smime.p7s>>
-- devel mailing list devel@xxxxxxxxxxxxxxxxxxxxxxx http://lists.fedoraproject.org/admin/lists/devel@xxxxxxxxxxxxxxxxxxxxxxx
