Revert "[SR-12033] [Sema] Do not allow inferring defaultable closure () -> $T for autoclosure arguments result"

[xedin]

Reverts #35503

@LucianoPAlmeida Unfortunately we have to revert this for now and make it more tailored since these changes caused a source compatibility regression in code like this:

protocol P {
  associatedtype F : Error
}

extension P {
  func test(error: F?) {
    accepts_autoclosure({
      if let failure = error {
        throw failure
      }
    })
  }

  func accepts_autoclosure<T>(_ expression: @autoclosure () throws -> T) {}
}

[xedin]

@swift-ci please smoke test

[xedin]

@swift-ci please smoke test Windows platform

[xedin]

@swift-ci please smoke test Linux platform

[xedin]

@swift-ci please smoke test Linux platform

[LucianoPAlmeida]

@LucianoPAlmeida Unfortunately we have to revert this for now and make it more tailored since these changes caused a source compatibility regression

hum, ok sorry for that and thank you for reverting :))
I'll try to take a look on that to see if we can make an adjustment

[xedin]

@LucianoPAlmeida No worries! The fix is correct since otherwise the behavior is not obvious to users it's just the source compatibility impact :/ Could you please open a PR to avoid transitive inference of DefaultClosureType removed which used to be a part of this one?

[LucianoPAlmeida]

it's just the source compatibility impact :/

That's unfortunate... do you think its possible to fix this without source compatibility break?

Could you please open a PR to avoid transitive inference of DefaultClosureType removed which used to be a part of this one?

For sure :)

[xedin]

That's unfortunate... do you think its possible to fix this without source compatibility break?

I don't think so since there are projects which rely on this behavior. What we can do though is to rank solutions with @autcolosure lower than solutions where parameters accept regular closures e.g.

func test<T>(_: () -> T) {}
func test<T>(_: @autoclosure () -> T) {}

test { }

This example is ambiguous today but it shouldn't be.

[LucianoPAlmeida]

What we can do though is to rank solutions with @autcolosure lower than solutions where parameters accept regular closures

That makes sense!
Do we have a JIRA ticket for that already? If not I can create it just to track this

[xedin]

We might, it might be worth a search before creating a new one.

[LucianoPAlmeida]

We might, it might be worth a search before creating a new one.

Right, I'll look it up. Thanks!

[LucianoPAlmeida]

Found two actually that seems to address the same issue SR-2705 and SR-6725

[xedin]

SR-2705 is what we want here.