Enforce Sendability in EventLoopFuture and EventLoopPromise methods
#2501
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
FranzBusch
commented
Aug 11, 2023
| } | ||
| return next.futureResult | ||
| public func flatMapWithEventLoop<NewValue: Sendable>(_ callback: @escaping @Sendable (Value, EventLoop) -> EventLoopFuture<NewValue>) -> EventLoopFuture<NewValue> { | ||
| // Is this the same thing and still fast? |
There was a problem hiding this comment.
I am not sure about this one. Probably the previous implementation was faster but I wasn't really sure here
# Motivation With apple#2496 we fixed a Sendability checking hole by removing the unconditional conformance of `EventLoopFuture/Promise` to `Sendable`. This was the correct thing to do; however, it has the fallout that a couple of methods are now rightly complaining that their values are send across isolation domains. # Modification This PR requires values on some `ELF/P` methods to be `Sendable` when we might potentially transfer the values across isolation domains/ELs. We have to be overly aggressive here because we do not know that some `ELF` method are staying on the same EL. For example `flatMap` gets a new `ELF` from the closure provided to it. If the `ELF` is on the same EL we do not need to hop; however, we can not guarantee this right now from a type level so we have to stay on the safe side and actually require the `NewValue` to be `Sendable`. # Result This PR makes us more correct from a Sendability perspective but produces warnings for some safe patterns that are currently in use.
FranzBusch
force-pushed
the
fb-sendable-el
branch
from
05ec0a9 to
9d9ac1c
Compare
weissi
reviewed
Aug 11, 2023
| let eventLoop = self.eventLoop | ||
| return self.flatMap { | ||
| callback($0, eventLoop) | ||
| }.hop(to: self.eventLoop) |
There was a problem hiding this comment.
you can delete the .hop(to:). self.flatMap will return a future on self.eventLoop
There was a problem hiding this comment.
and the previous implementation probably also allocates less, depends on the optimiser a bit. Why not just revert?
There was a problem hiding this comment.
Right. Well this method became a lot easier then :D
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Motivation
With #2496 we fixed a Sendability checking hole by removing the unconditional conformance of
EventLoopFuture/PromisetoSendable. This was the correct thing to do; however, it has the fallout that a couple of methods are now rightly complaining that their values are send across isolation domains.Modification
This PR requires values on some
ELF/Pmethods to beSendablewhen we might potentially transfer the values across isolation domains/ELs. We have to be overly aggressive here because we do not know that someELFmethod are staying on the same EL. For exampleflatMapgets a newELFfrom the closure provided to it. If theELFis on the same EL we do not need to hop; however, we can not guarantee this right now from a type level so we have to stay on the safe side and actually require theNewValueto beSendable.Result
This PR makes us more correct from a Sendability perspective but produces warnings for some safe patterns that are currently in use.