In the process of rebuilding lots of gems for EL6, I've run across a a
couple areas where gems' test suites fail because of changes in Ruby's
hash ordering algorithms. I was able to apply upstream patches to fix
these issues in rubygem-activesupport and rubygem-activerecord. These
bugs are tricky to catch, because sometimes the test suite works, and
other times a bad test fails.
I've come across a FTBFS issue with rubygem-rack-mount on F16, and I'm
wondering if it could also be related to the hash ordering. These two
koji scratch builds fail:
I'm a Ruby newbie, so I'd appreciate another pair of eyes on these
Has anybody gone through this migration process yet? For my project I'm
struggling with things like updating the Rakefile, getting the RSpec and
Cucumber tests working. The app itself seems to be performing just fine.
Darryl L. Pierce, Sr. Software Engineer @ Red Hat, Inc.
Delivering value year after year.
Red Hat ranks #1 in value among software vendors.
here is a sumup of the last 2-hour FPC meeting, that me and Mo have attended:
- Virtual provides are going to be killed completely (e.g. new packages will now depend on ruby-foo or rubygem-foo, rather than on ruby(foo) or rubygem(foo); no provides will be therefore necessary, as the package name is enough for this).
- There has been some discussion about placement of Rubygems (/usr/share/rubygems) and that Toshio doesn't like that, because it's special casing. We didn't have enough time to solve that, so it's a matter for other discussion, that I have started on packaging list . I have summed up some arguments why things should stay the way they currently are and I'm waiting for Toshio's answer. This seems to be the last blocker (I hope) of the new guidelines.
I would really appreciate if as many of you as possible could join the discussion there and bring in some more arguments, so that we can persuade FPC that this is the right approach.
Thank you and special thanks goes to Mo for showing up on the FPC meeting.
Bohuslav "Slavek" Kabrda.