Package: ruby2.0 Version: 2.0.0.484+really457-3 Severity: important I see ruby2.0 is M-A:foreign, which seems wrong to me.
Consider a package "ruby-multiarch-example" that contains a ruby script requiring an arch:any ruby module (ruby-bcrypt for instance), and a compiled binary. This would be Architecture:any and Depend on ruby2.0, ruby-bcrypt. Now it's possible to install ruby2.0:amd64, ruby-bcrypt:i386, and ruby-multiarch-example:i386. This results in the ruby script failing because the architectures of ruby-bcrypt and ruby2.0 don't match, even though all the dependencies are satisfied. I've put such a package available for testing at deb http://people.debian.org/~ntyni/multiarch-testing sid main deb-src http://people.debian.org/~ntyni/multiarch-testing sid main which is signed with the key at http://people.debian.org/~ntyni/multiarch-testing.asc (containing a signature from my DD key.) To reproduce, do something like this in a clean amd64 chroot: # dpkg --add-architecture i386 # apt-get update # apt-get install ruby2.0:amd64 ruby-multiarch-example:i386 # ruby-multiarch-example /usr/lib/ruby/2.0.0/rubygems/core_ext/kernel_require.rb:55:in `require': cannot load such file -- bcrypt_ext (LoadError) from /usr/lib/ruby/2.0.0/rubygems/core_ext/kernel_require.rb:55:in `require' from /usr/lib/ruby/vendor_ruby/bcrypt.rb:16:in `rescue in <top (required)>' from /usr/lib/ruby/vendor_ruby/bcrypt.rb:12:in `<top (required)>' from /usr/lib/ruby/2.0.0/rubygems/core_ext/kernel_require.rb:55:in `require' from /usr/lib/ruby/2.0.0/rubygems/core_ext/kernel_require.rb:55:in `require' from /usr/bin/ruby-multiarch-example:3:in `<main>' I believe 'Multi-Arch: allowed' was designed for this use case, and that's what python is using. However, there are concerns that this is still broken in some cases. See Guillem's message at https://lists.debian.org/debian-devel/2013/04/msg00578.html and the current thread on debian-devel, around https://lists.debian.org/debian-devel/2014/04/msg00464.html (I still haven't enabled multi-arch in perl mainly because of those concerns, but maybe python has showed now that it's working well enough.) Thanks for your work on ruby, -- Niko Tyni nt...@debian.org -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
