-
Notifications
You must be signed in to change notification settings - Fork 12.8k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
More DefineOpaqueTypes::Yes #123794
More DefineOpaqueTypes::Yes #123794
Conversation
be168c7
to
4abbda6
Compare
@rfcbot merge See main post for an explanation of the new behaviour. Imo this is the expected behaviour, especially as we already do this for other types: fn k() -> impl Sized {
let x = match true {
true => {
// Get a value of an opaque type on stable
let t = k();
(t,)
}
false => ((),),
};
todo!()
} |
Team member @oli-obk has proposed to merge this. The next step is review by the rest of the tagged team members: No concerns currently listed. Once a majority of reviewers approve (and at most 2 approvals are outstanding), this will enter its final comment period. If you spot a major issue that hasn't been raised at any point in this process, please speak up! See this document for info about what commands tagged team members can give me. |
Is this just for consistency? Or was there some motivating example for this? I worry about continuing to expand the scope of what opaques can do, while we're currently trying to stabilise more of them (associated type impl trait). Maybe this is too pessimistic of a view, because this does seen fairly straightforward, but I'd hate to see us accidentally introduce another bug or subtlety that we have to consider. Also, I think at the very least @rust-lang/lang should be pinged on this, even if not part of the FCP. |
@@ -1145,7 +1145,7 @@ impl<'a, 'tcx> FnCtxt<'a, 'tcx> { | |||
// are the same function and their parameters have a LUB. | |||
match self.commit_if_ok(|_| { | |||
self.at(cause, self.param_env).lub( | |||
DefineOpaqueTypes::No, | |||
DefineOpaqueTypes::Yes, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
DefineOpaqueTypes::Yes
inside of a commit_if_ok
likely means that this can cause type inference breakage.
When coercing function definitions we may have previously used the lub
further down as this lub
here failed. This would have then caused us to return a function pointer (i think), whereas we now return an FnDef
.
I would like to see a crater run here to allow us to more eagerly detect any regressions from it
consistency with |
👍 perfect |
@bors try |
More DefineOpaqueTypes::Yes This accepts more code on stable. It is now possible to have match arms return a function item `foo::<ConcreteType>` and a function item `foo::<OpaqueTypeInDefiningScope>` in another, and that will constrain `OpaqueTypeInDefiningScope` to have the hidden type `ConcreteType`. So the following function will now compile, but on master it errors with a type mismatch on the second match arm ```rust // The function item whose generic params we want to merge. fn foo<T>(t: T) -> T { t } // Helper ensuring we can constrain `T` on `F` without explicitly specifying it fn bind<T, F: FnOnce(T) -> T>(_: T, f: F) -> F { f } fn k() -> impl Sized { let x = match true { true => { // `f` is `FnDef(foo, [infer_var])` let f = foo; // Get a value of an opaque type on stable let t = k(); // this returns `FnDef(foo, [k::return])` bind(t, f) } false => foo::<()>, }; todo!() } ``` r? `@compiler-errors`
☀️ Try build successful - checks-actions |
@craterbot check |
👌 Experiment ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more |
🔔 This is now entering its final comment period, as per the review above. 🔔 |
🚧 Experiment ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more |
🎉 Experiment
|
This comment has been minimized.
This comment has been minimized.
@bors r- |
f4521f7
to
5c55d6a
Compare
@bors r=lcnr |
More DefineOpaqueTypes::Yes This accepts more code on stable. It is now possible to have match arms return a function item `foo::<ConcreteType>` and a function item `foo::<OpaqueTypeInDefiningScope>` in another, and that will constrain `OpaqueTypeInDefiningScope` to have the hidden type `ConcreteType`. So the following function will now compile, but on master it errors with a type mismatch on the second match arm ```rust // The function item whose generic params we want to merge. fn foo<T>(t: T) -> T { t } // Helper ensuring we can constrain `T` on `F` without explicitly specifying it fn bind<T, F: FnOnce(T) -> T>(_: T, f: F) -> F { f } fn k() -> impl Sized { let x = match true { true => { // `f` is `FnDef(foo, [infer_var])` let f = foo; // Get a value of an opaque type on stable let t = k(); // this returns `FnDef(foo, [k::return])` bind(t, f) } false => foo::<()>, }; todo!() } ``` r? `@compiler-errors` cc rust-lang#116652
Rollup of 5 pull requests Successful merges: - rust-lang#122500 (delegation: Support renaming, and async, const, extern "ABI" and C-variadic functions) - rust-lang#123316 (Test `#[unix_sigpipe = "inherit"]` with both `SIG_DFL` and `SIG_IGN`) - rust-lang#123794 (More DefineOpaqueTypes::Yes) - rust-lang#124282 (windows fill_utf16_buf: explain the expected return value) - rust-lang#124308 (Add diagnostic item for `std::iter::Enumerate`) Failed merges: - rust-lang#124136 (Provide more context and suggestions in borrowck errors involving closures) r? `@ghost` `@rustbot` modify labels: rollup
Rollup of 6 pull requests Successful merges: - rust-lang#123316 (Test `#[unix_sigpipe = "inherit"]` with both `SIG_DFL` and `SIG_IGN`) - rust-lang#123794 (More DefineOpaqueTypes::Yes) - rust-lang#123881 (Bump Fuchsia versions) - rust-lang#124281 (fix weak memory bug in TLS on Windows) - rust-lang#124282 (windows fill_utf16_buf: explain the expected return value) - rust-lang#124308 (Add diagnostic item for `std::iter::Enumerate`) r? `@ghost` `@rustbot` modify labels: rollup
Rollup of 6 pull requests Successful merges: - rust-lang#123316 (Test `#[unix_sigpipe = "inherit"]` with both `SIG_DFL` and `SIG_IGN`) - rust-lang#123794 (More DefineOpaqueTypes::Yes) - rust-lang#123881 (Bump Fuchsia versions) - rust-lang#124281 (fix weak memory bug in TLS on Windows) - rust-lang#124282 (windows fill_utf16_buf: explain the expected return value) - rust-lang#124308 (Add diagnostic item for `std::iter::Enumerate`) r? `@ghost` `@rustbot` modify labels: rollup
Rollup merge of rust-lang#123794 - oli-obk:define_opaque_types2, r=lcnr More DefineOpaqueTypes::Yes This accepts more code on stable. It is now possible to have match arms return a function item `foo::<ConcreteType>` and a function item `foo::<OpaqueTypeInDefiningScope>` in another, and that will constrain `OpaqueTypeInDefiningScope` to have the hidden type `ConcreteType`. So the following function will now compile, but on master it errors with a type mismatch on the second match arm ```rust // The function item whose generic params we want to merge. fn foo<T>(t: T) -> T { t } // Helper ensuring we can constrain `T` on `F` without explicitly specifying it fn bind<T, F: FnOnce(T) -> T>(_: T, f: F) -> F { f } fn k() -> impl Sized { let x = match true { true => { // `f` is `FnDef(foo, [infer_var])` let f = foo; // Get a value of an opaque type on stable let t = k(); // this returns `FnDef(foo, [k::return])` bind(t, f) } false => foo::<()>, }; todo!() } ``` r? ``@compiler-errors`` cc rust-lang#116652
…ackh726 Allow coercing functions whose signature differs in opaque types in their defining scope into a shared function pointer type r? `@compiler-errors` This accepts more code on stable. It is now possible to have match arms return a function item `foo` and a different function item `bar` in another, and that will constrain OpaqueTypeInDefiningScope to have the hidden type ConcreteType and make the type of the match arms a function pointer that matches the signature. So the following function will now compile, but on master it errors with a type mismatch on the second match arm ```rust fn foo<T>(t: T) -> T { t } fn bar<T>(t: T) -> T { t } fn k() -> impl Sized { fn bind<T, F: FnOnce(T) -> T>(_: T, f: F) -> F { f } let x = match true { true => { let f = foo; bind(k(), f) } false => bar::<()>, }; todo!() } ``` cc rust-lang#116652 This is very similar to rust-lang#123794, and with the same rationale: > this is for consistency with `-Znext-solver`. the new solver does not have the concept of "non-defining use of opaque" right now and we would like to ideally keep it that way. Moving to `DefineOpaqueTypes::Yes` in more cases removes subtlety from the type system. Right now we have to be careful when relating `Opaque` with another type as the behavior changes depending on whether we later use the `Opaque` or its hidden type directly (even though they are equal), if that later use is with `DefineOpaqueTypes::No`*
Rollup merge of rust-lang#124297 - oli-obk:define_opaque_types13, r=jackh726 Allow coercing functions whose signature differs in opaque types in their defining scope into a shared function pointer type r? `@compiler-errors` This accepts more code on stable. It is now possible to have match arms return a function item `foo` and a different function item `bar` in another, and that will constrain OpaqueTypeInDefiningScope to have the hidden type ConcreteType and make the type of the match arms a function pointer that matches the signature. So the following function will now compile, but on master it errors with a type mismatch on the second match arm ```rust fn foo<T>(t: T) -> T { t } fn bar<T>(t: T) -> T { t } fn k() -> impl Sized { fn bind<T, F: FnOnce(T) -> T>(_: T, f: F) -> F { f } let x = match true { true => { let f = foo; bind(k(), f) } false => bar::<()>, }; todo!() } ``` cc rust-lang#116652 This is very similar to rust-lang#123794, and with the same rationale: > this is for consistency with `-Znext-solver`. the new solver does not have the concept of "non-defining use of opaque" right now and we would like to ideally keep it that way. Moving to `DefineOpaqueTypes::Yes` in more cases removes subtlety from the type system. Right now we have to be careful when relating `Opaque` with another type as the behavior changes depending on whether we later use the `Opaque` or its hidden type directly (even though they are equal), if that later use is with `DefineOpaqueTypes::No`*
Allow coercing functions whose signature differs in opaque types in their defining scope into a shared function pointer type r? `@compiler-errors` This accepts more code on stable. It is now possible to have match arms return a function item `foo` and a different function item `bar` in another, and that will constrain OpaqueTypeInDefiningScope to have the hidden type ConcreteType and make the type of the match arms a function pointer that matches the signature. So the following function will now compile, but on master it errors with a type mismatch on the second match arm ```rust fn foo<T>(t: T) -> T { t } fn bar<T>(t: T) -> T { t } fn k() -> impl Sized { fn bind<T, F: FnOnce(T) -> T>(_: T, f: F) -> F { f } let x = match true { true => { let f = foo; bind(k(), f) } false => bar::<()>, }; todo!() } ``` cc rust-lang/rust#116652 This is very similar to rust-lang/rust#123794, and with the same rationale: > this is for consistency with `-Znext-solver`. the new solver does not have the concept of "non-defining use of opaque" right now and we would like to ideally keep it that way. Moving to `DefineOpaqueTypes::Yes` in more cases removes subtlety from the type system. Right now we have to be careful when relating `Opaque` with another type as the behavior changes depending on whether we later use the `Opaque` or its hidden type directly (even though they are equal), if that later use is with `DefineOpaqueTypes::No`*
Allow coercing functions whose signature differs in opaque types in their defining scope into a shared function pointer type r? `@compiler-errors` This accepts more code on stable. It is now possible to have match arms return a function item `foo` and a different function item `bar` in another, and that will constrain OpaqueTypeInDefiningScope to have the hidden type ConcreteType and make the type of the match arms a function pointer that matches the signature. So the following function will now compile, but on master it errors with a type mismatch on the second match arm ```rust fn foo<T>(t: T) -> T { t } fn bar<T>(t: T) -> T { t } fn k() -> impl Sized { fn bind<T, F: FnOnce(T) -> T>(_: T, f: F) -> F { f } let x = match true { true => { let f = foo; bind(k(), f) } false => bar::<()>, }; todo!() } ``` cc rust-lang/rust#116652 This is very similar to rust-lang/rust#123794, and with the same rationale: > this is for consistency with `-Znext-solver`. the new solver does not have the concept of "non-defining use of opaque" right now and we would like to ideally keep it that way. Moving to `DefineOpaqueTypes::Yes` in more cases removes subtlety from the type system. Right now we have to be careful when relating `Opaque` with another type as the behavior changes depending on whether we later use the `Opaque` or its hidden type directly (even though they are equal), if that later use is with `DefineOpaqueTypes::No`*
This accepts more code on stable. It is now possible to have match arms return a function item
foo::<ConcreteType>
and a function itemfoo::<OpaqueTypeInDefiningScope>
in another, and that will constrainOpaqueTypeInDefiningScope
to have the hidden typeConcreteType
. So the following function will now compile, but on master it errors with a type mismatch on the second match armr? @compiler-errors
cc #116652