Mocks on nameless splats

[kaukas]

Hi! Thanks for your library! It makes me, an RSpec lover, feel more at home. :-)

After a while I tried to run my project specs (that used to work) and something is failing now. I don't really know what caused the failure; perhaps a newer Crystal version.

My tail of shards.lock:

  spectator:
    git: https://gitlab.com/arctic-fox/spectator.git
    version: 0.11.5

$ crystal --version
Crystal 1.7.2 (2023-01-23)

LLVM: 14.0.6
Default target: aarch64-apple-darwin22.1.0

$ git clone https://github.com/kaukas/crystal-cassandra
$ shards install
$ crystal spec spec/cassandra/dbapi_spec.cr --error-trace
There was a problem expanding macro 'default_stub'

Called macro defined in lib/spectator/src/spectator/mocks/stubbable.cr:108:13

 108 | private macro default_stub(method)

Which expanded to:

 >  1 |
 >  2 |
 >  3 |
 >  4 |
 >  5 |
 >  6 |        def should(
 >  7 |         expectation : BeAExpectation(T), failure_message : String | ::Nil = nil, *, file = __FILE__, line = __LINE__,
 >  8 |
 >  9 |
 > 10 |       ) : T forall T
 > 11 |         super(expectation, failure_message, file: file, line: line)
 > 12 |       end
 > 13 |
 > 14 |
 > 15 |
 > 16 |
 > 17 |        def should(
 > 18 |         expectation : BeAExpectation(T), failure_message : String | ::Nil = nil, *, file = __FILE__, line = __LINE__,
 > 19 |
 > 20 |
 > 21 |       ) : T forall T
 > 22 |
 > 23 |         # Capture information about the call.
 > 24 |         __temp_153 = ::Spectator::MethodCall.build(
 > 25 |           :should,
 > 26 |           ::NamedTuple.new(
 > 27 |             "expectation": expectation, "failure_message": failure_message,
 > 28 |           ),
 > 29 |           :"", ,
 > 30 |           ::NamedTuple.new(
 > 31 |             "file": file, "line": line,
 > 32 |           ).merge()
 > 33 |         )
 > 34 |         _spectator_record_call(__temp_153)
 > 35 |
 > 36 |         # Attempt to find a stub that satisfies the method call and arguments.
 > 37 |         # Finding a suitable stub is delegated to the type including the `Stubbable` module.
 > 38 |         if __temp_154 = _spectator_find_stub(__temp_153)
 > 39 |           # Cast the stub or return value to the expected type.
 > 40 |           # This is necessary to match the expected return type of the original method.
 > 41 |           _spectator_cast_stub_value(__temp_154, __temp_153, typeof(previous_def),
 > 42 |           :raise)
 > 43 |         else
 > 44 |           # Delegate missing stub behavior to concrete type.
 > 45 |           _spectator_stub_fallback(__temp_153, typeof(previous_def)) do
 > 46 |             # Use the default response for the method.
 > 47 |             previous_def
 > 48 |           end
 > 49 |         end
 > 50 |       end
 > 51 |
Error: unterminated call

From what I gather splat names are expected to either be non empty strings on nils. However, this particular should definition has an unnamed splat which ends up as an empty string when rendered in the macro (see :"", , on line 29).

Unfortunately, my quick attempt to reproduce it with a spec in spectator failed. Perhaps you'd have better luck?..

Thank you!

[icy-arctic-fox]

Luckily this has been fixed on a branch and will be merged soon.

[kaukas]

Thank you so much!