Reproduced. I'm honestly not sure how this would have been working before—it seems like this really ought to be ambiguous?

@DougGregor, @xedin, have we seen something like this? It sounds a little familiar.

This is minimized example:

struct A {}
struct B {
  init(a: (A) -> Void) {}
  init(_ b: (A) -> Void) {}
}

_ = B { a in }

It used to work before only if the second overload doesn't have a label. I think this is ought to be ambiguous though because trailing closures do not require labels, @DougGregor WDYT?

stephan (JIRA User) If you form a proper call e.g. `_ B({ a in })` that would resolve ambiguity in favor of label-less overload, or you can do `_ = B(a: { a in })` to use the first one.

I agree that this should always have been ambiguous. Trailing closures mean that the label on the last parameter is uninteresting from a ranking perspective when it has function type.