Decide whether v3.0.0 can be released

[floehopper]

• In a comment on one of the issues, I asked a few people to try out the release candidate.
• I opened this PR on alphagov/whitehall to try out Mocha v3.0.0.pre.rc.1 and all the tests pass and I can't see any Mocha-related deprecation warnings.

[floehopper]

@lastgabs @rafaelfranca @adrianna-chang-shopify I'd really appreciate it if one of you could try out the Mocha release candidate (v3.0.0.pre.rc.1) in one of your bigger projects when you have time? (@byroot suggested I contacted you).

[adrianna-chang-shopify]

👋 Hey @floehopper , we're unfortunately a couple versions behind in our monolith at Shopify (v2.5.0), but I'll bring this up with the team and see if we can get back on top of these + help test out the RC.

[floehopper]

@adrianna-chang-shopify Oh, that would be great - many thanks! ❤ Let me know if you run into any issues.

[floehopper]

@adrianna-chang-shopify Have you had a chance to try the RC yet?

[adrianna-chang-shopify]

ccing @etiennebarrie who started looking at this. We had a lot of failures related to the 2.6 => 2.7 upgrade that we were still working through last I'd heard :(

[floehopper]

@adrianna-chang-shopify Many thanks for the update.

@etiennebarrie let me know if there's anything I can do to help.

[etiennebarrie]

I'm still off until next week, but the main thing I remember is this kind of pattern:

require "bundler/inline"

gemfile do
  source "https://rubygems.org"
  gem "mocha", ARGV.fetch(0, "~> 2.5.0")
  gem "minitest"
end

require "minitest/autorun"
require "mocha/minitest"

module SomeModule
end

class SomeTest < Minitest::Test
  def test_passes
    SomeModule.stubs(:name).returns("Other")
    SomeModule.expects(:name).never # incorrect expectation
    assert_equal "Other", SomeModule.name
    SomeModule.name # doesn't raise
  end
end

To be clear this is an incorrectly written test, the never didn't have any effect, as evidenced by the last line of the test here.
But it passed with 2.5, warned with 2.6 (#681) and now fails with 2.7, so as far as we're concerned, 2.7.0 is the version that includes breaking changes.

We have a few of tests written like that, and I wanted to actually fix those tests (add expectations about invocations not happening) and it's a bit more involved, so I didn't have a chance to test 3.0 yet. I should have just skipped those tests in a branch and tested the release candidate directly, sorry about that.

[floehopper]

@etiennebarrie Ah, OK - thanks for interrupting your time off to provide that update. I quite understand your desire to fix those tests.