[ConstraintSystem] Allow fixing missing conformance failures for Void and uninhabited types.
#27123
Conversation
`Void` and uninhabited types.
and no return statements.
|
@swift-ci please smoke test |
| fatalError() | ||
| } | ||
|
|
||
| let _ = foo() {fatalError()} & nil // expected-error {{generic parameter 'T' could not be inferred}} | ||
| let _ = foo() {fatalError()} & nil // expected-error {{global function 'foo' requires that 'Never' conform to 'P'}} |
There was a problem hiding this comment.
This is a clear improvement!
However I can't help but wonder about 'Void'. The diagnostic would say "global function 'foo' requires that 'Void' conform to 'P'", which is a true statement, but 'Void' cannot ever conform to 'P'. Same if it's a function type or a metatype. As a follow-up you might want to add some test cases for non-nominal types and see if the diagnostic could be tailored a bit.
There was a problem hiding this comment.
I agree that there should be a tailored diagnostic. For example, this code:
protocol P {}
func genericFunc<T: P>(_ x: T) {}
genericFunc { return }
currently produces the error "Argument type '() -> ()' does not conform to expected type 'P'".
We already give a good error message for protocol types - "Protocol type 'P' cannot conform to 'P' because only concrete types can conform to protocols" - we could use a similar error message for non-nominal types.
There was a problem hiding this comment.
I'll do this in a separate PR
|
I think #27133 should have fixed the crash here... |
|
@swift-ci please smoke test |
1 similar comment
|
@swift-ci please smoke test |
|
Tests are still failing because of one validation test that I need to update the diagnostics for |
missing conformance diagnostics.
|
@swift-ci please smoke test |
…sing conformance diagnostics.
|
@swift-ci please smoke test |
No description provided.