Fix documentation in tokio::sync::watch
#5261
Conversation
tokio/src/sync/watch.rs
Outdated
| /// to keep the borrow as short lived as possible. Additionally, if you are | ||
| /// running in an environment that allows `!Send` futures, you must ensure that | ||
| /// running in an environment that disallows `!Send` futures, you must ensure that | ||
| /// the returned `Ref` type is never held alive across an `.await` point. |
There was a problem hiding this comment.
This is correct as written.
There was a problem hiding this comment.
But if the environment allows !Send futures, then there is no restriction for holding a reference to Ref across an .await point. This restriction applies only when !Send futures are disallowed (e.g. multi-thread runtime).
So I still think this is a valid correction
There was a problem hiding this comment.
Holding a Ref across an .await is always bad because it can lead to a deadlock, but if !Send futures are disallowed, then you can't do it by accident since the compiler will catch it.
There was a problem hiding this comment.
Aha, I see this may lead to a panic or a deadlock... That's the piece I was missing. Maybe I should extend this comment to make it more explicit about what the aftermath is? WDYT?
There was a problem hiding this comment.
That's fine. (It can only lead to a deadlock, not a panic.)
There was a problem hiding this comment.
Added a small clarification about the deadlock, and added a dash to long-lived and short-lived words as per the grammar.
I've noticed that this paragraph, which warns about holding a ref across an .await point, is repeated in several places. Some of them slightly differed from the doc comment on the struct Ref. I've updated them all to the same wording for consistency ✔️
There was a problem hiding this comment.
@Darksonn waiting for your review
Fixed just a few logical and grammar mistakes in the docs