In meth(*foo, bar) both foo and bar are splatted

[whitequark]

Here is an example of a CI failure I'm experiencing on rbx: https://travis-ci.org/whitequark/parser/jobs/6120901.

All of the failures follow the same pattern: a method of form s(:foo, *bar, baz) is called, where both bar and baz implement #to_a; in particular, bar is an Array and baz is an AST::Node which implements #to_a, returning an Array.

The problem is that both bar and baz are splatted, instead of just bar.

I also made a minimal repro here: https://github.com/whitequark/ast/blob/master/test/test_ast.rb#L112 and it also fails the same way: https://travis-ci.org/whitequark/ast/builds/6121338.

[whitequark]

Even more minimal repro, courtesy of @tsion:

def foo(*args); p args; end
foo(*[1,2,3], [4,5,6]) # ok
a = [1,2,3]; b = [4,5,6]; foo(*a, b) # fail

[solson]

Here's a few more curious cases that might make it easier to track this down:

>> a = [1,2,3]; foo(*a, [4,5,6])
=> [1, 2, 3, 4, 5, 6]
>> a = [1,2,3]; foo(*a, [4,5,6], [7,8,9])
=> [1, 2, 3, [4, 5, 6], [7, 8, 9]]
>> a = [1,2,3]; foo([0,0,0], *a, [4,5,6])
=> [[0, 0, 0], 1, 2, 3, 4, 5, 6]
>> a = [1,2,3]; foo([0,0,0], *a, [4,5,6], [7,8,9])
=> [[0, 0, 0], 1, 2, 3, [4, 5, 6], [7, 8, 9]]

[leocassarani]

I'm having exactly the same issue. This was particularly sneaky in my own code, as one of the non-splatted arguments was a Struct instance, which the splat will erroneously transform into its attributes. Before realising this had already been reported, I'd come up with the following to illustrate the problem:

def bar(one, two)
  puts "this will not be reached"
end

a = ['foo']
b = [1, 2]
bar(*a, b) # method 'bar': given 3, expected 2 (ArgumentError)

This is not an issue in MRI 1.9 and above.