Repository navigation
Additional trait restrictions prevent use of trait methods #44064
Description
Adding Send + Sync restrictions to an Any trait object causes the trait object to lose the downcast_ref method
https://play.rust-lang.org/?gist=9c665d721a9c6a2845f55e56d97ffe27&version=stable
use std::any::Any;
struct Foo;
// This compiles
fn main() {
let foo: Box<Any> = Box::new(Foo) as Box<Any>;
let foo = &foo;
let foo = foo.downcast_ref::<Foo>().unwrap();
}
// This compiles
fn main_2() {
let foo: Box<Any + Send + Sync> = Box::new(Foo) as Box<Any + Send + Sync>;
let foo = &foo;
let foo = Any::downcast_ref::<Foo>(foo).unwrap();
}
// This doesn't compile
fn main_3() {
let foo: Box<Any + Send + Sync> = Box::new(Foo) as Box<Any + Send + Sync>;
let foo = &foo;
let foo = foo.downcast_ref::<Foo>().unwrap();
}I would expect main_3 to compile, because main_2 compiles, however main_3 fails to compile with the error
error[E0599]: no method named `downcast_ref` found for type `&std::boxed::Box<std::any::Any + std::marker::Send + std::marker::Sync>` in the current scope
--> src/main.rs:23:19
|
23 | let foo = foo.downcast_ref::<Foo>().unwrap();
| ^^^^^^^^^^^^
error: aborting due to previous error(s)
error: Could not compile `playground`.
To learn more, run the command again with --verbose.Meta
$ rustc --version
rustc 1.19.0 (0ade339 2017-07-17)
Activity
kennytm commentedon Aug 23, 2017
Any + Sendalso works, since there is animplfor it. This needs a separateimpldue to #23423, in order to fix #18737.So to fix this we just need to forward
Any + SyncandAny + Send + Syncalso.Xaeroxe commentedon Aug 23, 2017
After thoughts on filing this issue:
I guess the question is "should adding a provided method on a trait be a breaking change for multi-trait trait objects in the event that ambiguity between methods for multiple traits is introduced by the provided method's presence?" Which I think the answer is no, so yeah it makes sense to prevent method calls in this way, but perhaps the compiler could give a better syntax hint as to what should be done instead.
I wouldn't be opposed to an impl for
Any + Send + Syncbut I also don't think it's terribly necessary.Xaeroxe commentedon Aug 23, 2017
So I'm going to close this issue.