Repository navigation
Possible regression on use of Self in generics ("mismatched types") #69611
Description
The following code (also on playground) compiles fine on Stable and Beta rustc, but has started no longer compiling on latest nightly (2020-02-28, 2020-02-29 both have this issue).
pub struct Container<V>(pub V);
impl<T, V> core::ops::Mul<T> for Container<V>
where
T: From<V> + core::ops::Mul<T, Output = T>,
{
type Output = Container<T>;
fn mul(self, scalar: T) -> Self::Output {
Self(scalar * self.0.into())
}
}I would expect this code snippet to compile with no errors as it currently does in stable and beta Rust.
Instead, on recent nightlies (02-28 and 02-29 were tested), an error is produced:
--> src/lib.rs:10:14
|
3 | impl<T, V> core::ops::Mul<T> for Container<V>
| - - expected type parameter
| |
| found type parameter
...
10 | Self(scalar * self.0.into())
| ^^^^^^^^^^^^^^^^^^^^^^ expected type parameter `V`, found type parameter `T`
|
= note: expected type parameter `V`
found type parameter `T`
= note: a type parameter was expected, but a different one was found; you might be missing a type parameter or trait bound
= note: for more information, visit https://doc.rust-lang.org/book/ch10-02-traits.html#traits-as-parameters
For what it's worth, the substitution
pub struct Container<V>(pub V);
impl<T, V> core::ops::Mul<T> for Container<V>
where
T: From<V> + core::ops::Mul<T, Output = T>,
{
type Output = Container<T>;
fn mul(self, scalar: T) -> Self::Output {
- Self(scalar * self.0.into())
+ Container(scalar * self.0.into())
}
}resolves the error.
I believe this is because rustc is erroneously picking Self to be Container<V> (the type for which the impl is being specified) instead of using Container and inferring the (formerly correct) output type of Container<T>.
Meta
rustc --version --verbose:
rustc 1.43.0-nightly (d3c79346a 2020-02-29)
binary: rustc
commit-hash: d3c79346a3e7ddbb5fb417810f226ac5a9209007
commit-date: 2020-02-29
host: x86_64-apple-darwin
release: 1.43.0-nightly
LLVM version: 9.0
Activity
rye commentedon Mar 1, 2020
Sure enough, this seems related to #69306, which was marked resolved by PR #69340. I guess I'm one of the lucky people affected by those changes. 😄
Centril commentedon Mar 1, 2020
Yeah, this was caused by a bug fix to type checking of
Selfpaths, which would previously wrongly introduce inference variables for type parameters in the impl. UsingSelfshould always refer to theSelftype, just as usingSelf {...}would. In other words, your code should never have compiled.Centril commentedon Mar 18, 2020
Closing as expected breakage and "wontfix".
Fix use of Self