Fix issue #4002 #4003
Fix issue #4002 #4003
Conversation
| @@ -1206,7 +1206,7 @@ object IO extends IOCompanionPlatform with IOLowPriorityImplicits { | |||
| lift(k(resume)) flatMap { | |||
| case Right(a) => G.pure(a) | |||
| case Left(Some(fin)) => G.onCancel(poll(get), lift(fin)) | |||
| case Left(None) => poll(get) | |||
| case Left(None) => get | |||
There was a problem hiding this comment.
Will this cause the same kind of semantic break as the async_ change in 3.5.0?
There was a problem hiding this comment.
Yes.
As far as I can tell, asyncCheckAttempt should have been added like this in 3.5.0 (and for Async, it was). It's only due to an unlucky git merge that it wasn't. (And I guess due to the lack of a unit test.)
There was a problem hiding this comment.
I suspect the impact of this is much lower than the async_ change, since this is a far less widely used constructor. Also we override the Async implementation in Async[IO], so this bug fix is actually entirely shadowed unless you're using monad transformers or a third-party effect. I think it's fair to consider this PR to be a bug fix rather than a breaking change.
Edit: Oh my shadowing argument is backwards. Rip. Either way, bug fix IMO.
This PR makes
IO#asyncCheckAttemptwork the same asAsync#asyncCheckAttempt. So both will be uncancelable when suspended with aNonefinalizer. Closes #4002.