[SR-11920] Bogus Warning When Invalid Protocol is Used in Existential Position

[CodaFi]

Previous ID	SR-11920
Radar	rdar://problem/57715257
Original Reporter	@CodaFi
Type	Bug
Status	Resolved
Resolution	Done

Additional Detail from JIRA

Votes	0
Component/s	Compiler
Labels	Bug, DiagnosticsQoI, StarterBug, TypeChecker
Assignee	None
Priority	Medium

md5: 6a0184667d6dc91a903498b162ead4ab

Issue Description:

protocol P { associatedtype T }
class C : P {}

func i<T: C>(_ x : T?) -> Bool {
  //  warning: checking a value with optional type 'T?' against dynamic type 'P' succeeds whenever the value is non-nil; did you mean to use '!= nil'?
  return x is P
}

This warning's text makes no sense. `P` is not a dynamic protocol and may not be used in existential position. The wording of this diagnostic needs to be amended not to make reference to dynamism when a protocol may not be used in existential position.

[CodaFi]

@swift-ci Create

[swift-ci]

Comment by Tapan Thaker (JIRA)

@CodaFi: I am not clear on the meaning of "dynamic protocol" here. Could you please explain or link me to a reference, if possible.

[CodaFi]

To be honest, I’m not sure why that particular description is used. The a protocol without self or associated type constraints has nothing to do with dynamism... That said, there is a dynamic cast involved, so perhaps that was the original intent.

[swift-ci]

Comment by Ziyuan Zhao (JIRA)

I'll try to fix this

[slavapestov]

I think this behaves correctly. Since the class C conforms to P, it is true that you can replace the ‘is’ check with ‘!= nil’.

[CodaFi]

The diagnostic's wording is still bogus. Its text needs to reflect the the actual situation that's occurred in source. Re-opening this.

[LucianoPAlmeida]

Resolved by #40811
CodaFi (JIRA User) I'm resolving, but if you still think the wording can improve we can reopen again =]