On Mon, Jan 06, 2014 at 02:16:24PM +0100, Lucas Nussbaum wrote: > On 05/01/14 at 16:42 +0100, Francesco Poli wrote: > > On Sat, 4 Jan 2014 17:26:31 +0100 Lucas Nussbaum wrote: > > Specifically, it says: > > > > [...] > > | ### Dependencies > > | > > | All packages **must** depend either on > > | `ruby | ruby-interpreter` > > | if the package supports all implementations of Ruby or on the specific > > | implementations needed if that is not the case. > > [...] > > > > Hence, I am a bit hesitant about dropping something that is mandated by > > the currently recommended Debian Ruby practice. > > Yeah, I raised it on -ruby@ because I think that there might be a bug in > our recommended practice. But I'm not sure. I feel that I'm getting a > nice headache thanks to alternative depends + alternatives. :-) > > > Maybe ruby-defaults should provide a symlink to the default ruby version > > > that we could use in shebangs (#!/usr/bin/ruby-debian-default)? That > > > would allow to ensure that hcih really runs with ruby1.9.3 even if the > > > user used update-alternatives to switch to another version. > > > > Wouldn't that defeat the very purpose of the > > /usr/bin/ruby -> /etc/alternatives/ruby > > symlink?!? > > Thinking more about it, I think that the problem could be solved by > ruby1.8 dropping its Provides on ruby-interpreter. If we remove that > Provides, during a fresh install of hcih or apt-listbugs, 'ruby' will > get installed. Agreed. We want to strenghten out definition of `ruby` and `ruby-interpreter` as follows: ruby = the default Ruby ruby-interpreter = any *supported* Ruby (thus disqualifying ruby1.8) > I'm not sure what will happen during upgrades when ruby1.8 is > installed. That would have to be tested. It should just work™, but yeah, needs to be tested. -- Antonio Terceiro <terceiro@debian.org>
Attachment:
signature.asc
Description: Digital signature