Prefer explict closure sig types over expected ones
fixes#100800
Previously we only checked that given closure arguments are equal to expected closure arguments, but now we choose the given closure arguments for the signature that is used when type checking the closure body, and keep the other signature for the type of the closure as seen outside of it.
Initial implementation of dyn*
This PR adds extremely basic and incomplete support for [dyn*](https://smallcultfollowing.com/babysteps//blog/2022/03/29/dyn-can-we-make-dyn-sized/). The goal is to get something in tree behind a flag to make collaboration easier, and also to make sure the implementation so far is not unreasonable. This PR does quite a few things:
* Introduce `dyn_star` feature flag
* Adds parsing for `dyn* Trait` types
* Defines `dyn* Trait` as a sized type
* Adds support for explicit casts, like `42usize as dyn* Debug`
* Including const evaluation of such casts
* Adds codegen for drop glue so things are cleaned up properly when a `dyn* Trait` object goes out of scope
* Adds codegen for method calls, at least for methods that take `&self`
Quite a bit is still missing, but this gives us a starting point. Note that this is never intended to become stable surface syntax for Rust, but rather `dyn*` is planned to be used as an implementation detail for async functions in dyn traits.
Joint work with `@nikomatsakis` and `@compiler-errors.`
r? `@bjorn3`
Emit a note that static bounds from HRTBs are a bug
This note isn't perfect, but opening this to either 1) land as is or 2) get some feedback on how to improve it
Let r? `@compiler-errors` and cc. `@nikomatsakis`
Make `compare_predicate_entailment` no longer a query
Make `compare_predicate_entailment` so it's no longer a query (again), and splits out the new logic (that equates the return types to infer RPITITs) into its own query. This means that this new query (now called `collect_trait_impl_trait_tys`) is no longer executed for non-RPITIT cases.
This should improve perf (https://github.com/rust-lang/rust/pull/101224#issuecomment-1241682203), though in practice we see that these some crates remain from the primary regressions list on the original report... They are all <= 0.43% regression and seemingly only on the incr-full scenario for all of them.
I am at a loss for what might be causing this regression other than what I fixed here, since we don't introduce much new non-RPITIT logic except for some `def_kind` query calls in some places, for example, like projection. Maybe that's it?
----
Originally this PR was opened to test enabling `cache_on_disk` (62164aaaa11) but that didn't turn out to be very useful (https://github.com/rust-lang/rust/pull/101615#issuecomment-1242403205), so that led me to just split the query (and rename the PR).
Check that the types in return position `impl Trait` in traits are well-formed
This effectively duplicates `check_associated_type_bounds`, but that shouldn't be for long, since we're going to remove it once we refactor RPITITs into regular associated items.
Fixes#101663
---
We don't check
```rust
trait Foo {
fn bar() -> impl ?Sized;
}
```
currently, but that's for a different reason, which is that we don't currently check that a trait function's return type is sized (i.e. `fn bar() -> [u8]` also works in a trait).
Fix ICE in opt_suggest_box_span
We were _totally_ mishandling substs and obligations in `opt_suggest_box_span`, so I reworked that function pretty heavily.
Also some drive-by changes, namely removing `ret_type_span`.
Fixes#101465
Equate fn outputs when inferring RPITIT hidden types
When we are trying to infer the hidden types for RPITITs, we need to equate the output tys instead of just subtyping them. For example:
```rust
trait Foo { fn bar() -> impl Sized {} }
impl Foo for () { fn bar() -> &'static str { "" } }
```
If we just subtype the signatures `fn() -> &'static str <: fn() -> _#1t` (where `_#1t` is the variable we've used to infer `impl Sized`), we'll end up `&'static str <: _#1t`, which causes us to infer `_#1t = #'_#2r str`, where `'_#2r` is unconstrained, which gets fixed up to `ReEmpty`, and which is certainly not what we want.
I can't actually think of a way to make this fail to compile, because during borrowck we've already done the method probe, and so we just look at the `impl` method signature and see the `&'static str` any time we call `<() as Foo>::bar()`. But this _does_ cause the ICE [here](https://github.com/rust-lang/rust/pull/98559#issuecomment-1241891994) in `@jackh726's` "Remove ReEmpty" PR (#98559) to stop ICEing, because after that PR we were leaking unconstrained region variables into the typeck results.
r? types
The `<*const T>::guaranteed_*` methods now return an option for the unknown case
cc https://github.com/rust-lang/rust/issues/53020#issuecomment-1236932443
I chose `0` for "not equal" and `1` for "equal" and left `2` for the unknown case so backends can just forward to raw pointer equality and it works ✨
r? `@fee1-dead` or `@lcnr`
cc `@rust-lang/wg-const-eval`
Stabilize raw-dylib for non-x86
This stabilizes the `raw-dylib` and `link_ordinal` features (#58713) for non-x86 architectures (i.e., `x86_64`, `aarch64` and `thumbv7a`):
* Marked the `raw_dylib` feature as `active`.
* Marked the `link_ordinal` attribute as `ungated`.
* Added new errors if either feature is used on x86 targets without the `raw_dylib` feature being enabled.
* Updated tests to only set the `raw_dylib` feature when building for x86.
Rollup of 7 pull requests
Successful merges:
- #98933 (Opaque types' generic params do not imply anything about their hidden type's lifetimes)
- #101041 (translations(rustc_session): migrates rustc_session to use SessionDiagnostic - Pt. 2)
- #101424 (Adjust and slightly generalize operator error suggestion)
- #101496 (Allow lower_lifetime_binder receive a closure)
- #101501 (Allow lint passes to be bound by `TyCtxt`)
- #101515 (Recover from typo where == is used in place of =)
- #101545 (Remove unnecessary `PartialOrd` and `Ord`)
Failed merges:
r? `@ghost`
`@rustbot` modify labels: rollup
Adjust and slightly generalize operator error suggestion
(in no particular order)
* Stop passing around a whole extra `ProjectionPredicate`
* Add spaces around `=` in `Trait<..., Output = Ty>` suggestion
* Some code clean-ups, including
* add `lang_item_for_op` to turn a `Op` into a `DefId`
* avoid `SourceMap` because we don't really need to render an expr
* Remove `TypeParamVisitor` in favor of just checking `ty.has_param_types_or_consts` -- this acts a bit differently, but shouldn't cause erroneous suggestions (actually might generalize them a bit)
* We now suggest `Output = Ty` in the `where` clause suggestion when we fail to add `Struct<T>` and `T`.
I can split this out into more PRs if needed, but they're all just miscellaneous generalizations, changes, and nitpicks I saw when messing with this operator code.
