Inconsistent type inference behavior of block-type vs expression-type closures

[stereotype441]

Consider the following code:

typedef num FunctionReturningNum();

main() {
  Object o = 1;
  FunctionReturningNum a = () => o;
  FunctionReturningNum b = () { return o; }
}

Type inference currently infers a return type of Object for the closure in a, and a return type of num for the closure in b.

The behavior of expression-type closures should be changed to match that of block-type closures.

[jensjoha]

This no longer seem to be the case. @stereotype441 can we close?

[stereotype441]

This issue was fixed in the front end in b755025. But this bug is about the analyzer. Since the analyzer hasn't yet been migrated to the front end, I am planning to replicate this bug fix in the analyzer. So I would like to leave this bug open until then.

[jensjoha]

OK, thanks for clearing that up.

[leafpetersen]

@stereotype441 The example above no longer reproduces on bleeding edge. Should we close this?

[stereotype441]

Confirmed that this has been fixed.