dev-ruby/rails:5.2 required by www-apps/redmine which stuck in ruby26 due unmet dependencies on ruby27 target. List of packages: dev-ruby/actioncable/actioncable-5.2.8.ebuild dev-ruby/actionmailer/actionmailer-5.2.8.ebuild dev-ruby/actionpack/actionpack-5.2.8.ebuild dev-ruby/actionview/actionview-5.2.8.ebuild dev-ruby/activejob/activejob-5.2.8.ebuild dev-ruby/activemodel/activemodel-5.2.8.ebuild dev-ruby/activerecord/activerecord-5.2.8.ebuild dev-ruby/activestorage/activestorage-5.2.8.ebuild dev-ruby/rails/rails-5.2.8.ebuild dev-ruby/railties/railties-5.2.8.ebuild
Forgot about: dev-ruby/arel/arel-9.0.0.ebuild
graaff and I discussed this and concluded the best way forward just going to redmine 5. He said he had a bunch of test failures with rails:5.2. I haven't tried it though. If you could get them passing, could reconsider.
(In reply to Sam James from comment #2) > graaff and I discussed this and concluded the best way forward just going to > redmine 5. Yes, that version depends on rails:6.1 so it should be good to go for a while. > He said he had a bunch of test failures with rails:5.2. I haven't tried it > though. Yes, tests are failing (I only tried activemodel), but they are also failing on ruby26 now so this seems more of a generic problem, rather than ruby27-specific. Still needs to be fixed before we can consider adding ruby27 here. Personally I'm not going to spend time on this, let's move forward (where there is already enough to do) rather than standing still.
(In reply to Hans de Graaff from comment #3) > Yes, tests are failing (I only tried activemodel), but they are also failing > on ruby26 now so this seems more of a generic problem, rather than > ruby27-specific. Still needs to be fixed before we can consider adding > ruby27 here. > > Personally I'm not going to spend time on this, let's move forward (where > there is already enough to do) rather than standing still. These test already failed back on ruby25, surprisingly enough why they not fixed back then. In https://github.com/gentoo/gentoo/pull/26365 I've added fixes for dev-ruby/arel and dev-ruby/activemodel :5.2 :6.0 :6.1 branches. Other packages runs test fine.
(In reply to Azamat H. Hackimov from comment #4) > These test already failed back on ruby25, surprisingly enough why they not > fixed back then. At least for Arel: the BigDecimal issue was deprecated in ruby26 but still worked. Only in ruby27 did the old syntax actually break. I've updated arel 9.0.0 with your patch. I did not cherry pick that commit since I also wanted to update EAPI and get rid of versionator. > In https://github.com/gentoo/gentoo/pull/26365 I've added fixes for > dev-ruby/arel and dev-ruby/activemodel :5.2 :6.0 :6.1 branches. I don't see a fix for activemodel (only for actionview).
Here's an example of the errors I see for activemodel: 76) validates numericality of for :odd on generated message given if condition ArgumentError: mocked method :call expects 3 arguments, got [:title, :odd] /var/tmp/portage/dev-ruby/activemodel-5.2.8/work/ruby26/rails-5.2.8/activemodel/lib/active_model/errors.rb:454:in `normalize_message' /var/tmp/portage/dev-ruby/activemodel-5.2.8/work/ruby26/rails-5.2.8/activemodel/lib/active_model/errors.rb:298:in `add' /var/tmp/portage/dev-ruby/activemodel-5.2.8/work/ruby26/rails-5.2.8/activemodel/lib/active_model/validations/numericality.rb:62:in
Created attachment 791246 [details] dev-ruby/activemodel-5.2.8 buildlog
(In reply to Hans de Graaff from comment #6) > Here's an example of the errors I see for activemodel: > > 76) validates numericality of for :odd on generated message given if > condition > ArgumentError: mocked method :call expects 3 arguments, got [:title, > :odd] > > /var/tmp/portage/dev-ruby/activemodel-5.2.8/work/ruby26/rails-5.2.8/ > activemodel/lib/active_model/errors.rb:454:in `normalize_message' > > /var/tmp/portage/dev-ruby/activemodel-5.2.8/work/ruby26/rails-5.2.8/ > activemodel/lib/active_model/errors.rb:298:in `add' > > /var/tmp/portage/dev-ruby/activemodel-5.2.8/work/ruby26/rails-5.2.8/ > activemodel/lib/active_model/validations/numericality.rb:62:in Can you please post full build log? I can't reproduce it on my system. All test for dev-ruby/activemodel on ruby27 runs without fails. Finished in 0.642573s, 1260.5567 runs/s, 3546.6774 assertions/s. 810 runs, 2279 assertions, 0 failures, 0 errors, 0 skips >>> Completed testing dev-ruby/activemodel-5.2.8
(In reply to Azamat H. Hackimov from comment #8) > Can you please post full build log? I can't reproduce it on my system. All > test for dev-ruby/activemodel on ruby27 runs without fails. It turns out that this is an incompatibility introduced in minitest-5.16. I'll update the minitest dependencies for testing accordingly. Note that there is a security release of rails so I'll need to do bumps for all slots. I'll try to include ruby27 for rails 5.2 as well.
OK, so here's regression https://github.com/minitest/minitest/issues/912. Seems this commit solves regressions with minitest 5.16 on recent rails: https://github.com/rails/rails/pull/45370/commits/9766eb4a833c26c64012230b96dd1157ebb8e8a2 Probably I can backport it to all branches of rails, but seems there more wide issues over other ruby packages and minitest. Maybe worth it to mask minitest 5.16 for a while?
Update: support got added (see https://github.com/gentoo/gentoo/pull/26365#issuecomment-1182852089) but minitest issue still needs handling.
(In reply to Sam James from comment #11) > Update: support got added (see > https://github.com/gentoo/gentoo/pull/26365#issuecomment-1182852089) but > minitest issue still needs handling. Ah, yes, I forgot to restrict the minitest version in the older slots. This has now been fixed.
(In reply to Azamat H. Hackimov from comment #10) > OK, so here's regression https://github.com/minitest/minitest/issues/912. > > Seems this commit solves regressions with minitest 5.16 on recent rails: > > https://github.com/rails/rails/pull/45370/commits/ > 9766eb4a833c26c64012230b96dd1157ebb8e8a2 > > Probably I can backport it to all branches of rails, but seems there more > wide issues over other ruby packages and minitest. Maybe worth it to mask > minitest 5.16 for a while? I've restricted the minitest version used for testing in affected packages, so there is no need to backport. We'll revisit that restriction when new versions are released. Reading the minitest upstream bug it looks like this new behaviour is intentional and meant to stay, so I'd rather not mask minitest 5.16.
The bug has been referenced in the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=8b31a809571ce8c127f0de8443361b6711ab8830 commit 8b31a809571ce8c127f0de8443361b6711ab8830 Author: Hans de Graaff <graaff@gentoo.org> AuthorDate: 2022-07-16 08:01:20 +0000 Commit: Hans de Graaff <graaff@gentoo.org> CommitDate: 2022-07-16 08:11:03 +0000 dev-ruby/activesupport: restrict minitest version Bug: https://bugs.gentoo.org/857759 Signed-off-by: Hans de Graaff <graaff@gentoo.org> dev-ruby/activesupport/activesupport-5.2.8.1.ebuild | 1 + dev-ruby/activesupport/activesupport-6.0.5.1.ebuild | 1 + dev-ruby/activesupport/activesupport-6.1.6.1.ebuild | 1 + 3 files changed, 3 insertions(+)
