[have_receive] (not have_receive"d") will return true no matter what

[github0013]

Subject of the issue

Calling have_receive always become true. (not have_receive[d])

have_receive, does this matcher method exist?
https://relishapp.com/rspec/rspec-mocks/docs/basics/spies
I only see have_receive[d]

Your environment

• ruby 2.6.2p47 (2019-03-13 revision 67232) [x86_64-darwin16]
• rspec-mocks (3.8.1)

Steps to reproduce

describe :have_receive do
  let(:object) { spy(:object) }

  it do
    object.some_method
    # This is expected to be true 
    expect(object).to have_received(:some_method)
  end

  # This is expected to be false, but is true, and it doesn't even raise an error  for `have_receive`
  it { expect(object).to have_receive(:method_never_called) }
end

Expected behavior

    # This should not pass or should raise an error for the method missing?
    it { expect(object).to have_receive(:never_exist_method) } 

Actual behavior

    # This passes, when it shouldn't
    it { expect(object).to have_receive(:never_exist_method) } 

[JonRowe]

👋 Hi, you've fallen victim to a typo leading you to a strange inconsistency.

Whats happening here...

1. A spy returns true for any method by default.
2. have_... is a special matcher checking for the method name after the _ as a predicate, e.g. recieve? in this case.
3. The spy returns true for receive?, the matcher is therefore satisfied, the test passes.

A work arounds to avoid this is to use a verifying double rather than a normal spy, which would then recognise the non existant method and help highlight the issue.

I think a proper fix is needed here too, but that will require us detecting spys in the magic matcher and special casing them, WDYT @benoittgt

[benoittgt]

Thanks for your issue @github0013

I think a proper fix is needed here too, but that will require us detecting spys in the magic matcher and special casing them, WDYT @benoittgt

I don't see exactly what you mean here. Isn't breaking the contract of spy?

rspec-mocks/lib/rspec/mocks/example_methods.rb

Lines 117 to 122 in d0cd8df

	# Constructs a test double that is optimized for use with
	# `have_received`. With a normal double one has to stub methods in order
	# to be able to spy them. A spy automatically spies on all methods.
	def spy(*args)
	double(*args).as_null_object
	end

[JonRowe]

@benoittgt have_receive is a typo, but its picked up as have_<predicate> using RSpec::Expectations::Has, the spy by design responds to all methods in order to work, but in this case it cases a false positive pass.

[nevinera]

As with #1132, I think this might be closeable now - the actual behavior is still there, but the problem case now causes:

expected #<Double "thing">.has_recieveddddddd?(:foo) to return false, got #<Double "thing">

At some point in the last four years, the matcher started checking for true/false, instead of truthy/falsey values.

Edit: As pointed out here, this was from the strict_predicate_matchers, which is enabled in this repo, but not by default. I have what I think is a resolution in rspec/rspec-expectations#1455

[JonRowe]

I'm closing this one in favour of #1132