Closed pirj closed 3 years ago
For some reason, builds stopped triggering. Going to switch to "ready for review" temporarily just to figure out if it's the case (quite unlikely though).
For the build issue it is because of travisci config. https://github.com/rspec/rspec-support/pull/436/commits/112973040cbab3bfb69b23bc21b03bc084c93e00
@JonRowe Thanks for the review.
I've outlined the things to discuss in the description, including:
[?] extract rubocop changes into a separate pull request
I'll extract this into a separate follow-up pull request to 4-0-dev
.
@JonRowe @benoittgt Appreciate if you can take a look at Discuss section of the description, there are some important questions, let me cite below:
[?] introduce https://github.com/ruby/ruby2_keywords dependency and remove respond_to?(:ruby2_keywords, true) check
No, we can't introduce dependencies.
[?] drop Ruby 2.3 - problematic'ish due to a dependency on outdated OpenSSL, EOL'ed, not supported by RuboCop, still used by 8% of users
What Rubocop supports does not drive us. 2.3 is not really problematic.
[?] tweak clone_repo to accept a branch, use maintenance-branch's branch for main repos, and 4-0-maintenance for cloning rspec-rails since it's alienated anyway?
It already does this? Its just hardcoded based on the branch?
[?] extract rubocop changes into a separate pull request
Just remove them, they're unneeded.
and 4-0-maintenance for cloning rspec-rails since it's alienated anyway?
It already does this? Its just hardcoded based on the branch?
Doesn't seem to. The build was failing without drop-old-rubies
rspec-rails
branch. Also, it failed with default version.rb
:
Resolving dependencies...
Bundler could not find compatible versions for gem "rspec-core":
In Gemfile:
rspec-core
rspec-rails was resolved to 4.1.0.pre, which depends on
rspec-core (= 3.11.0.pre)
Changed it to 4.0.0.pre
to trick this build.
I'm getting stuck with removing WithKeywordsWhenNeeded
and dealing with kwargs delegation related failures in combination with dropping support for old rubies.
Removal of WithKeywordsWhenNeeded
is tangential to the removal of rubies, and I'm going to remove all eval
hacks, but not kwargs delegation hacks.
I intend to address the removal of WithKeywordsWhenNeeded
and kwargs delegation issues and hacks in a separate series of pull requests.
@JonRowe I intend to add back a jruby-9.1
build job, ~and make an attempt to remove ripper_supported?
after that~.
No CI-related files included in this change! Suddenly all-green CI π
@benoittgt @JonRowe welcome for a review.
PS and huge thanks for your work on JRuby/Windows/CI etc, guys. Your work has made this cleanup possible β€οΈ
Updates:
jruby_version
removal and its usagesffi
at ~> 1.12.0
Update: combined ReentrantMutex with Mutex.
Update: reset the temporary maintenance-branch
commit.
Green! π
π
Better viewed ignoring indentation changes https://github.com/rspec/rspec-support/pull/436/files?diff=split&w=1
As per https://github.com/rspec/rspec/issues/61#issuecomment-687290045
Sibling PRs:
Minimal changes to
rspec-core
andrspec-mocks
were required to runrspec-support
's own spec suite, pushed to branches with the same name.What has been done
Done:
(class << self; self; end)
usages for possibility to usesingleton_class
(some can't be replaced assingleton_class
does create a singleton class if it doesn't exist yet)eval
gsub
/unindent
where squiggly heredoc can be usedripper_supported?
~ - it has poor support in JRuby 9.1