Reproduction case for Ruby 3.2 and kwargs detection

[pirj]

Related to #1495

[pirj]

@eregon Do you happen to know what might have caused the failures on 3.2.0-preview2 and ruby-head while passed just fine with 3.1?

Is this still correct for Ruby 3.2?

        if RSpec::Support::RubyFeatures.distincts_kw_args_from_positional_hash? && expected_args == actual_args
          expected_hash = expectation.expected_args.last
          actual_hash = args_for_multiple_calls.last.last
          if Hash === expected_hash && Hash === actual_hash &&
            (Hash.ruby2_keywords_hash?(expected_hash) != Hash.ruby2_keywords_hash?(actual_hash))
            actual_args += Hash.ruby2_keywords_hash?(actual_hash) ? " (keyword arguments)" : " (options hash)"
            expected_args += Hash.ruby2_keywords_hash?(expected_hash) ? " (keyword arguments)" : " (options hash)"
          end
        end

2.2 failure is unrelated, to be fixed by #1498

[eregon]

Isn't this expected to fail? It expects a positional Hash but is given kwargs.
But I guess expect(dbl).to receive actually creates a method with no kwargs so then it should pass?

[pirj]

This indeed should have failed on earlier Ruby versions, too.

[eregon]

That code looks fine. I suspect it's a missing ruby2_keywords on the chain of calls, and so the flag is lost somewhere.
You can debug by printing Hash.ruby2_keywords_hash?(h) before the exception happens.

[eregon]

And 3.2 fixes a bug where some missing ruby2_keywords accidentally kept the flag: https://github.com/ruby/ruby/blob/master/NEWS.md

[pirj]

Thank you so much for the explanation, @eregon !

[pirj]

Superseded by #1514