stabilize `-Znext-solver=coherence`
r? `@compiler-errors`
---
This PR stabilizes the use of the next generation trait solver in coherence checking by enabling `-Znext-solver=coherence` by default. More specifically its use in the *implicit negative overlap check*. The tracking issue for this is https://github.com/rust-lang/rust/issues/114862. Closes#114862.
## Background
### The next generation trait solver
The new solver lives in [`rustc_trait_selection::solve`](https://github.com/rust-lang/rust/blob/master/compiler/rustc_trait_selection/src/solve/mod.rs) and is intended to replace the existing *evaluate*, *fulfill*, and *project* implementation. It also has a wider impact on the rest of the type system, for example by changing our approach to handling associated types.
For a more detailed explanation of the new trait solver, see the [rustc-dev-guide](https://rustc-dev-guide.rust-lang.org/solve/trait-solving.html). This does not stabilize the current behavior of the new trait solver, only the behavior impacting the implicit negative overlap check. There are many areas in the new solver which are not yet finalized. We are confident that their final design will not conflict with the user-facing behavior observable via coherence. More on that further down.
Please check out [the chapter](https://rustc-dev-guide.rust-lang.org/solve/significant-changes.html) summarizing the most significant changes between the existing and new implementations.
### Coherence and the implicit negative overlap check
Coherence checking detects any overlapping impls. Overlapping trait impls always error while overlapping inherent impls result in an error if they have methods with the same name. Coherence also results in an error if any other impls could exist, even if they are currently unknown. This affects impls which may get added to upstream crates in a backwards compatible way and impls from downstream crates.
Coherence failing to detect overlap is generally considered to be unsound, even if it is difficult to actually get runtime UB this way. It is quite easy to get ICEs due to bugs in coherence.
It currently consists of two checks:
The [orphan check] validates that impls do not overlap with other impls we do not know about: either because they may be defined in a sibling crate, or because an upstream crate is allowed to add it without being considered a breaking change.
The [overlap check] validates that impls do not overlap with other impls we know about. This is done as follows:
- Instantiate the generic parameters of both impls with inference variables
- Equate the `TraitRef`s of both impls. If it fails there is no overlap.
- [implicit negative]: Check whether any of the instantiated `where`-bounds of one of the impls definitely do not hold when using the constraints from the previous step. If a `where`-bound does not hold, there is no overlap.
- *explicit negative (still unstable, ignored going forward)*: Check whether the any negated `where`-bounds can be proven, e.g. a `&mut u32: Clone` bound definitely does not hold as an explicit `impl<T> !Clone for &mut T` exists.
The overlap check has to *prove that unifying the impls does not succeed*. This means that **incorrectly getting a type error during coherence is unsound** as it would allow impls to overlap: coherence has to be *complete*.
Completeness means that we never incorrectly error. This means that during coherence we must only add inference constraints if they are definitely necessary. During ordinary type checking [this does not hold](https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=01d93b592bd9036ac96071cbf1d624a9), so the trait solver has to behave differently, depending on whether we're in coherence or not.
The implicit negative check only considers goals to "definitely not hold" if they could not be implemented downstream, by a sibling, or upstream in a backwards compatible way. If the goal is is "unknowable" as it may get added in another crate, we add an ambiguous candidate: [source](bea5bebf3d/compiler/rustc_trait_selection/src/solve/assembly/mod.rs (L858-L883)).
[orphan check]: fd80c02c16/compiler/rustc_trait_selection/src/traits/coherence.rs (L566-L579)
[overlap check]: fd80c02c16/compiler/rustc_trait_selection/src/traits/coherence.rs (L92-L98)
[implicit negative]: fd80c02c16/compiler/rustc_trait_selection/src/traits/coherence.rs (L223-L281)
## Motivation
Replacing the existing solver in coherence fixes soundness bugs by removing sources of incompleteness in the type system. The new solver separately strengthens coherence, resulting in more impls being disjoint and passing the coherence check. The concrete changes will be elaborated further down. We believe the stabilization to reduce the likelihood of future bugs in coherence as the new implementation is easier to understand and reason about.
It allows us to remove the support for coherence and implicit-negative reasoning in the old solver, allowing us to remove some code and simplifying the old trait solver. We will only remove the old solver support once this stabilization has reached stable to make sure we're able to quickly revert in case any unexpected issues are detected before then.
Stabilizing the use of the next-generation trait solver expresses our confidence that its current behavior is intended and our work towards enabling its use everywhere will not require any breaking changes to the areas used by coherence checking. We are also confident that we will be able to replace the existing solver everywhere, as maintaining two separate systems adds a significant maintainance burden.
## User-facing impact and reasoning
### Breakage due to improved handling of associated types
The new solver fixes multiple issues related to associated types. As these issues caused coherence to consider more types distinct, fixing them results in more overlap errors. This is therefore a breaking change.
#### Structurally relating aliases containing bound vars
Fixes https://github.com/rust-lang/rust/issues/102048. In the existing solver relating ambiguous projections containing bound variables is structural. This is *incomplete* and allows overlapping impls. These was mostly not exploitable as the same issue also caused impls to not apply when trying to use them. The new solver defers alias-relating to a nested goal, fixing this issue:
```rust
// revisions: current next
//[next] compile-flags: -Znext-solver=coherence
trait Trait {}
trait Project {
type Assoc<'a>;
}
impl Project for u32 {
type Assoc<'a> = &'a u32;
}
// Eagerly normalizing `<?infer as Project>::Assoc<'a>` is ambiguous,
// so the old solver ended up structurally relating
//
// (?infer, for<'a> fn(<?infer as Project>::Assoc<'a>))
//
// with
//
// ((u32, fn(&'a u32)))
//
// Equating `&'a u32` with `<u32 as Project>::Assoc<'a>` failed, even
// though these types are equal modulo normalization.
impl<T: Project> Trait for (T, for<'a> fn(<T as Project>::Assoc<'a>)) {}
impl<'a> Trait for (u32, fn(&'a u32)) {}
//[next]~^ ERROR conflicting implementations of trait `Trait` for type `(u32, for<'a> fn(&'a u32))`
```
A crater run did not discover any breakage due to this change.
#### Unknowable candidates for higher ranked trait goals
This avoids an unsoundness by attempting to normalize in `trait_ref_is_knowable`, fixing https://github.com/rust-lang/rust/issues/114061. This is a side-effect of supporting lazy normalization, as that forces us to attempt to normalize when checking whether a `TraitRef` is knowable: [source](47dd709bed/compiler/rustc_trait_selection/src/solve/assembly/mod.rs (L754-L764)).
```rust
// revisions: current next
//[next] compile-flags: -Znext-solver=coherence
trait IsUnit {}
impl IsUnit for () {}
pub trait WithAssoc<'a> {
type Assoc;
}
// We considered `for<'a> <T as WithAssoc<'a>>::Assoc: IsUnit`
// to be knowable, even though the projection is ambiguous.
pub trait Trait {}
impl<T> Trait for T
where
T: 'static,
for<'a> T: WithAssoc<'a>,
for<'a> <T as WithAssoc<'a>>::Assoc: IsUnit,
{
}
impl<T> Trait for Box<T> {}
//[next]~^ ERROR conflicting implementations of trait `Trait`
```
The two impls of `Trait` overlap given the following downstream crate:
```rust
use dep::*;
struct Local;
impl WithAssoc<'_> for Box<Local> {
type Assoc = ();
}
```
There a similar coherence unsoundness caused by our handling of aliases which is fixed separately in https://github.com/rust-lang/rust/pull/117164.
This change breaks the [`derive-visitor`](https://crates.io/crates/derive-visitor) crate. I have opened an issue in that repo: nikis05/derive-visitor#16.
### Evaluating goals to a fixpoint and applying inference constraints
In the old implementation of the implicit-negative check, each obligation is [checked separately without applying its inference constraints](bea5bebf3d/compiler/rustc_trait_selection/src/traits/coherence.rs (L323-L338)). The new solver instead [uses a `FulfillmentCtxt`](bea5bebf3d/compiler/rustc_trait_selection/src/traits/coherence.rs (L315-L321)) for this, which evaluates all obligations in a loop until there's no further inference progress.
This is necessary for backwards compatibility as we do not eagerly normalize with the new solver, resulting in constraints from normalization to only get applied by evaluating a separate obligation. This also allows more code to compile:
```rust
// revisions: current next
//[next] compile-flags: -Znext-solver=coherence
trait Mirror {
type Assoc;
}
impl<T> Mirror for T {
type Assoc = T;
}
trait Foo {}
trait Bar {}
// The self type starts out as `?0` but is constrained to `()`
// due to the where-clause below. Because `(): Bar` is known to
// not hold, we can prove the impls disjoint.
impl<T> Foo for T where (): Mirror<Assoc = T> {}
//[current]~^ ERROR conflicting implementations of trait `Foo` for type `()`
impl<T> Foo for T where T: Bar {}
fn main() {}
```
The old solver does not run nested goals to a fixpoint in evaluation. The new solver does do so, strengthening inference and improving the overlap check:
```rust
// revisions: current next
//[next] compile-flags: -Znext-solver=coherence
trait Foo {}
impl<T> Foo for (u8, T, T) {}
trait NotU8 {}
trait Bar {}
impl<T, U: NotU8> Bar for (T, T, U) {}
trait NeedsFixpoint {}
impl<T: Foo + Bar> NeedsFixpoint for T {}
impl NeedsFixpoint for (u8, u8, u8) {}
trait Overlap {}
impl<T: NeedsFixpoint> Overlap for T {}
impl<T, U: NotU8, V> Overlap for (T, U, V) {}
//[current]~^ ERROR conflicting implementations of trait `Foo`
```
### Breakage due to removal of incomplete candidate preference
Fixes#107887. In the old solver we incompletely prefer the builtin trait object impl over user defined impls. This can break inference guidance, inferring `?x` in `dyn Trait<u32>: Trait<?x>` to `u32`, even if an explicit impl of `Trait<u64>` also exists.
This caused coherence to incorrectly allow overlapping impls, resulting in ICEs and a theoretical unsoundness. See https://github.com/rust-lang/rust/issues/107887#issuecomment-1997261676. This compiles on stable but results in an overlap error with `-Znext-solver=coherence`:
```rust
// revisions: current next
//[next] compile-flags: -Znext-solver=coherence
struct W<T: ?Sized>(*const T);
trait Trait<T: ?Sized> {
type Assoc;
}
// This would trigger the check for overlap between automatic and custom impl.
// They actually don't overlap so an impl like this should remain possible
// forever.
//
// impl Trait<u64> for dyn Trait<u32> {}
trait Indirect {}
impl Indirect for dyn Trait<u32, Assoc = ()> {}
impl<T: Indirect + ?Sized> Trait<u64> for T {
type Assoc = ();
}
// Incomplete impl where `dyn Trait<u32>: Trait<_>` does not hold, but
// `dyn Trait<u32>: Trait<u64>` does.
trait EvaluateHack<U: ?Sized> {}
impl<T: ?Sized, U: ?Sized> EvaluateHack<W<U>> for T
where
T: Trait<U, Assoc = ()>, // incompletely constrains `_` to `u32`
U: IsU64,
T: Trait<U, Assoc = ()>, // incompletely constrains `_` to `u32`
{
}
trait IsU64 {}
impl IsU64 for u64 {}
trait Overlap<U: ?Sized> {
type Assoc: Default;
}
impl<T: ?Sized + EvaluateHack<W<U>>, U: ?Sized> Overlap<U> for T {
type Assoc = Box<u32>;
}
impl<U: ?Sized> Overlap<U> for dyn Trait<u32, Assoc = ()> {
//[next]~^ ERROR conflicting implementations of trait `Overlap<_>`
type Assoc = usize;
}
```
### Considering region outlives bounds in the `leak_check`
For details on the `leak_check`, see the FCP proposal in #119820.[^leak_check]
[^leak_check]: which should get moved to the dev-guide once that PR lands :3
In both coherence and during candidate selection, the `leak_check` relies on the region constraints added in `evaluate`. It therefore currently does not register outlives obligations: [source](ccb1415eac/compiler/rustc_trait_selection/src/traits/select/mod.rs (L792-L810)). This was likely done as a performance optimization without considering its impact on the `leak_check`. This is the case as in the old solver, *evaluatation* and *fulfillment* are split, with evaluation being responsible for candidate selection and fulfillment actually registering all the constraints.
This split does not exist with the new solver. The `leak_check` can therefore eagerly detect errors caused by region outlives obligations. This improves both coherence itself and candidate selection:
```rust
// revisions: current next
//[next] compile-flags: -Znext-solver=coherence
trait LeakErr<'a, 'b> {}
// Using this impl adds an `'b: 'a` bound which results
// in a higher-ranked region error. This bound has been
// previously ignored but is now considered.
impl<'a, 'b: 'a> LeakErr<'a, 'b> for () {}
trait NoOverlapDir<'a> {}
impl<'a, T: for<'b> LeakErr<'a, 'b>> NoOverlapDir<'a> for T {}
impl<'a> NoOverlapDir<'a> for () {}
//[current]~^ ERROR conflicting implementations of trait `NoOverlapDir<'_>`
// --------------------------------------
// necessary to avoid coherence unknowable candidates
struct W<T>(T);
trait GuidesSelection<'a, U> {}
impl<'a, T: for<'b> LeakErr<'a, 'b>> GuidesSelection<'a, W<u32>> for T {}
impl<'a, T> GuidesSelection<'a, W<u8>> for T {}
trait NotImplementedByU8 {}
trait NoOverlapInd<'a, U> {}
impl<'a, T: GuidesSelection<'a, W<U>>, U> NoOverlapInd<'a, U> for T {}
impl<'a, U: NotImplementedByU8> NoOverlapInd<'a, U> for () {}
//[current]~^ conflicting implementations of trait `NoOverlapInd<'_, _>`
```
### Removal of `fn match_fresh_trait_refs`
The old solver tries to [eagerly detect unbounded recursion](b14fd2359f/compiler/rustc_trait_selection/src/traits/select/mod.rs (L1196-L1211)), forcing the affected goals to be ambiguous. This check is only an approximation and has not been added to the new solver.
The check is not necessary in the new solver and it would be problematic for caching. As it depends on all goals currently on the stack, using a global cache entry would have to always make sure that doing so does not circumvent this check.
This changes some goals to error - or succeed - instead of failing with ambiguity. This allows more code to compile:
```rust
// revisions: current next
//[next] compile-flags: -Znext-solver=coherence
// Need to use this local wrapper for the impls to be fully
// knowable as unknowable candidate result in ambiguity.
struct Local<T>(T);
trait Trait<U> {}
// This impl does not hold, but is ambiguous in the old
// solver due to its overflow approximation.
impl<U> Trait<U> for Local<u32> where Local<u16>: Trait<U> {}
// This impl holds.
impl Trait<Local<()>> for Local<u8> {}
// In the old solver, `Local<?t>: Trait<Local<?u>>` is ambiguous,
// resulting in `Local<?u>: NoImpl`, also being ambiguous.
//
// In the new solver the first impl does not apply, constraining
// `?u` to `Local<()>`, causing `Local<()>: NoImpl` to error.
trait Indirect<T> {}
impl<T, U> Indirect<U> for T
where
T: Trait<U>,
U: NoImpl
{}
// Not implemented for `Local<()>`
trait NoImpl {}
impl NoImpl for Local<u8> {}
impl NoImpl for Local<u16> {}
// `Local<?t>: Indirect<Local<?u>>` cannot hold, so
// these impls do not overlap.
trait NoOverlap<U> {}
impl<T: Indirect<U>, U> NoOverlap<U> for T {}
impl<T, U> NoOverlap<Local<U>> for Local<T> {}
//~^ ERROR conflicting implementations of trait `NoOverlap<Local<_>>`
```
### Non-fatal overflow
The old solver immediately emits a fatal error when hitting the recursion limit. The new solver instead returns overflow. This both allows more code to compile and is results in performance and potential future compatability issues.
Non-fatal overflow is generally desirable. With fatal overflow, changing the order in which we evaluate nested goals easily causes breakage if we have goal which errors and one which overflows. It is also required to prevent breakage due to the removal of `fn match_fresh_trait_refs`, e.g. [in `typenum`](https://github.com/rust-lang/trait-system-refactor-initiative/issues/73).
#### Enabling more code to compile
In the below example, the old solver first tried to prove an overflowing goal, resulting in a fatal error. The new solver instead returns ambiguity due to overflow for that goal, causing the implicit negative overlap check to succeed as `Box<u32>: NotImplemented` does not hold.
```rust
// revisions: current next
//[next] compile-flags: -Znext-solver=coherence
//[current] ERROR overflow evaluating the requirement
trait Indirect<T> {}
impl<T: Overflow<()>> Indirect<T> for () {}
trait Overflow<U> {}
impl<T, U> Overflow<U> for Box<T>
where
U: Indirect<Box<Box<T>>>,
{}
trait NotImplemented {}
trait Trait<U> {}
impl<T, U> Trait<U> for T
where
// T: NotImplemented, // causes old solver to succeed
U: Indirect<T>,
T: NotImplemented,
{}
impl Trait<()> for Box<u32> {}
```
#### Avoiding hangs with non-fatal overflow
Simply returning ambiguity when reaching the recursion limit can very easily result in hangs, e.g.
```rust
trait Recur {}
impl<T, U> Recur for ((T, U), (U, T))
where
(T, U): Recur,
(U, T): Recur,
{}
trait NotImplemented {}
impl<T: NotImplemented> Recur for T {}
```
This can happen quite frequently as it's easy to have exponential blowup due to multiple nested goals at each step. As the trait solver is depth-first, this immediately caused a fatal overflow error in the old solver. In the new solver we have to handle the whole proof tree instead, which can very easily hang.
To avoid this we restrict the recursion depth after hitting the recursion limit for the first time. We also **ignore all inference constraints from goals resulting in overflow**. This is mostly backwards compatible as any overflow in the old solver resulted in a fatal error.
### sidenote about normalization
We return ambiguous nested goals of `NormalizesTo` goals to the caller and ignore their impact when computing the `Certainty` of the current goal. See the [normalization chapter](https://rustc-dev-guide.rust-lang.org/solve/normalization.html) for more details.This means we apply constraints resulting from other nested goals and from equating the impl header when normalizing, even if a nested goal results in overflow. This is necessary to avoid breaking the following example:
```rust
trait Trait {
type Assoc;
}
struct W<T: ?Sized>(*mut T);
impl<T: ?Sized> Trait for W<W<T>>
where
W<T>: Trait,
{
type Assoc = ();
}
// `W<?t>: Trait<Assoc = u32>` does not hold as
// `Assoc` gets normalized to `()`. However, proving
// the where-bounds of the impl results in overflow.
//
// For this to continue to compile we must not discard
// constraints from normalizing associated types.
trait NoOverlap {}
impl<T: Trait<Assoc = u32>> NoOverlap for T {}
impl<T: ?Sized> NoOverlap for W<T> {}
```
#### Future compatability concerns
Non-fatal overflow results in some unfortunate future compatability concerns. Changing the approach to avoid more hangs by more strongly penalizing overflow can cause breakage as we either drop constraints or ignore candidates necessary to successfully compile. Weakening the overflow penalities instead allows more code to compile and strengthens inference while potentially causing more code to hang.
While the current approach is not perfect, we believe it to be good enough. We believe it to apply the necessary inference constraints to avoid breakage and expect there to not be any desirable patterns broken by our current penalities. Similarly we believe the current constraints to avoid most accidental hangs. Ignoring constraints of overflowing goals is especially useful, as it may allow major future optimizations to our overflow handling. See [this summary](https://hackmd.io/ATf4hN0NRY-w2LIVgeFsVg) and the linked documents in case you want to know more.
### changes to performance
In general, trait solving during coherence checking is not significant for performance. Enabling the next-generation trait solver in coherence does not impact our compile time benchmarks. We are still unable to compile the benchmark suite when fully enabling the new trait solver.
There are rare cases where the new solver has significantly worse performance due to non-fatal overflow, its reliance on fixpoint algorithms and the removal of the `fn match_fresh_trait_refs` approximation. We encountered such issues in [`typenum`](https://crates.io/crates/typenum) and believe it should be [pretty much as bad as it can get](https://github.com/rust-lang/trait-system-refactor-initiative/issues/73).
Due to an improved structure and far better caching, we believe that there is a lot of room for improvement and that the new solver will outperform the existing implementation in nearly all cases, sometimes significantly. We have not yet spent any time micro-optimizing the implementation and have many unimplemented major improvements, such as fast-paths for trivial goals.
TODO: get some rough results here and put them in a table
### Unstable features
#### Unsupported unstable features
The new solver currently does not support all unstable features, most notably `#![feature(generic_const_exprs)]`, `#![feature(associated_const_equality)]` and `#![feature(adt_const_params)]` are not yet fully supported in the new solver. We are confident that supporting them is possible, but did not consider this to be a priority. This stabilization introduces new ICE when using these features in impl headers.
#### fixes to `#![feature(specialization)]`
- fixes#105782
- fixes#118987
#### fixes to `#![feature(type_alias_impl_trait)]`
- fixes#119272
- https://github.com/rust-lang/rust/issues/105787#issuecomment-1750112388
- fixes#124207
## This does not stabilize the whole solver
While this stabilizes the use of the new solver in coherence checking, there are many parts of the solver which will remain fully unstable. We may still adapt these areas while working towards stabilizing the new solver everywhere. We are confident that we are able to do so without negatively impacting coherence.
### goals with a non-empty `ParamEnv`
Coherence always uses an empty environment. We therefore do not depend on the behavior of `AliasBound` and `ParamEnv` candidates. We only stabilizes the behavior of user-defined and builtin implementations of traits. There are still many open questions there.
### opaque types in the defining scope
The handling of opaque types - `impl Trait` - in both the new and old solver is still not fully figured out. Luckily this can be ignored for now. While opaque types are reachable during coherence checking by using `impl_trait_in_associated_types`, the behavior during coherence is separate and self-contained. The old and new solver fully agree here.
### normalization is hard
This stabilizes that we equate associated types involving bound variables using deferred-alias-equality. We also stop eagerly normalizing in coherence, which should not have any user-facing impact.
We do not stabilize the normalization behavior outside of coherence, e.g. we currently deeply normalize all types during writeback with the new solver. This may change going forward
### how to replace `select` from the old solver
We sometimes depend on getting a single `impl` for a given trait bound, e.g. when resolving a concrete method for codegen/CTFE. We do not depend on this during coherence, so the exact approach here can still be freely changed going forward.
## Acknowledgements
This work would not have been possible without `@compiler-errors.` He implemented large chunks of the solver himself but also and did a lot of testing and experimentation, eagerly discovering multiple issues which had a significant impact on our approach. `@BoxyUwU` has also done some amazing work on the solver. Thank you for the endless hours of discussion resulting in the current approach. Especially the way aliases are handled has gone through multiple revisions to get to its current state.
There were also many contributions from - and discussions with - other members of the community and the rest of `@rust-lang/types.` This solver builds upon previous improvements to the compiler, as well as lessons learned from `chalk` and `a-mir-formality`. Getting to this point would not have been possible without that and I am incredibly thankful to everyone involved. See the [list of relevant PRs](https://github.com/rust-lang/rust/pulls?q=is%3Apr+is%3Amerged+label%3AWG-trait-system-refactor+-label%3Arollup+closed%3A%3C2024-03-22+).
Make `Ty::boxed_ty` return an `Option`
Looks like a good place to use Rust's type system.
---
Most of 4ac7bcbaad/compiler/rustc_middle/src/ty/sty.rs (L971-L1963) looks like it could be moved to `TyKind` (then I guess `Ty` should be made to deref to `TyKind`).
explain why Rvalue::Len still exists
I just spent a bit of time trying to remove this until I realized why that's non-trivial. Let's document that for the next person. :)
Rename dump of coroutine by-move-body to be more consistent, fix ICE in dump_mir
First, we add a missing match for `DefKind::SyntheticCoroutineBody` in `dump_mir`. Fixes#129703. The second commit (directly below) serves as a test.
Second, we reorder the `dump_mir` in `coroutine_by_move_body_def_id` to be *after* we adjust the body source, and change the disambiguator so it reads more like any other MIR body. This also serves as a test for the ICE, since we're dumping the MIR of a body with `DefKind::SyntheticCoroutineBody`.
Third, we change the parenting of the synthetic MIR body to have the *coroutine-closure* (i.e. async closure) as its parent, so we don't have long strings of `{closure#0}-{closure#0}-{closure#0}`.
try-job: test-various
enable -Zrandomize-layout in debug CI builds
This builds rustc/libs/tools with `-Zrandomize-layout` on *-debug CI runners.
Only a handful of tests and asserts break with that enabled, which is promising. One test was fixable, the rest is dealt with by disabling them through new cargo features or compiletest directives.
The config.toml flag `rust.randomize-layout` defaults to false, so it has to be explicitly enabled for now.
Move `SanityCheck` and `MirPass`
They are currently in `rustc_middle`. This PR moves them to `rustc_mir_transform`, which makes more sense.
r? ``@cjgillot``
Non-exhaustive structs may be empty
This is a follow-up to a discrepancy noticed in https://github.com/rust-lang/rust/pull/122792: today, the following struct is considered inhabited (non-empty) outside its defining crate:
```rust
#[non_exhaustive]
pub struct UninhabitedStruct {
pub never: !,
// other fields
}
```
`#[non_exhaustive]` on a struct should mean that adding fields to it isn't a breaking change. There is no way that adding fields to this struct could make it non-empty since the `never` field must stay and is inconstructible. I suspect this was implemented this way due to confusion with `#[non_exhaustive]` enums, which indeed should be considered non-empty outside their defining crate.
I propose that we consider such a struct uninhabited (empty), just like it would be without the `#[non_exhaustive]` annotation.
Code that doesn't pass today and will pass after this:
```rust
// In a different crate
fn empty_match_on_empty_struct<T>(x: UninhabitedStruct) -> T {
match x {}
}
```
This is not a breaking change.
r? ``@compiler-errors``
Because that's now the only crate that uses it.
Moving stuff out of `rustc_middle` is always welcome.
I chose to use `impl crate::MirPass`/`impl crate::MirLint` (with
explicit `crate::`) everywhere because that's the only mention of
`MirPass`/`MirLint` used in all of these files. (Prior to this change,
`MirPass` was mostly imported via `use rustc_middle::mir::*` items.)
Rewrite lint_expectations in a single pass.
This PR aims at reducing the perf regression from https://github.com/rust-lang/rust/pull/120924#issuecomment-2202486203 with drive-by simplifications.
Basically, instead of using the lint level builder, which is slow, this PR splits `lint_expectations` logic in 2:
- listing the `LintExpectations` is done in `shallow_lint_levels_on`, on a per-owner basis;
- building the unstable->stable expectation id map is done by iterating on attributes.
r? ghost for perf
Rollup of 11 pull requests
Successful merges:
- #128523 (Add release notes for 1.81.0)
- #129605 (Add missing `needs-llvm-components` directives for run-make tests that need target-specific codegen)
- #129650 (Clean up `library/profiler_builtins/build.rs`)
- #129651 (skip stage 0 target check if `BOOTSTRAP_SKIP_TARGET_SANITY` is set)
- #129684 (Enable Miri to pass pointers through FFI)
- #129762 (Update the `wasm-component-ld` binary dependency)
- #129782 (couple more crash tests)
- #129816 (tidy: say which feature gate has a stability issue mismatch)
- #129818 (make the const-unstable-in-stable error more clear)
- #129824 (Fix code examples buttons not appearing on click on mobile)
- #129826 (library: Fix typo in `core::mem`)
r? `@ghost`
`@rustbot` modify labels: rollup
Enable Miri to pass pointers through FFI
Following https://github.com/rust-lang/rust/pull/126787, the purpose of this PR is to now enable Miri to execute native calls that make use of pointers.
> <details>
>
> <summary> Simple example </summary>
>
> ```rust
> extern "C" {
> fn ptr_printer(ptr: *mut i32);
> }
>
> fn main() {
> let ptr = &mut 42 as *mut i32;
> unsafe {
> ptr_printer(ptr);
> }
> }
> ```
> ```c
> void ptr_printer(int *ptr) {
> printf("printing pointer dereference from C: %d\n", *ptr);
> }
> ```
> should now show `printing pointer dereference from C: 42`.
>
> </details>
Note that this PR does not yet implement any logic involved in updating Miri's "analysis" state (byte initialization, provenance) upon such a native call.
r? ``@RalfJung``
Expand NLL MIR dumps
This PR is a first step to clean up and expand NLL MIR dumps:
- by restoring the "mir-include-spans" comments which are useful for `-Zdump-mir=nll`
- by adding the list of borrows to NLL MIR dumps, where they are introduced in the CFG and in which region
Comments in MIR dumps were turned off in #112346, but as shown in #114652 they were still useful for us working with NLL MIR dumps. So this PR pulls `-Z mir-include-spans` into its own options struct, so that passes dumping MIR can override them if need be. The rest of the compiler is not affected, only the "nll" pass dumps have these comments enabled again. The CLI still has priority when specifying the flag, so that we can explicitly turn them off in the `mir-opt` tests to keep blessed dumps easier to work with (which was one of the points of #112346).
Then, as part of a couple steps to improve NLL/polonius MIR dumps and `.dot` visualizations, I've also added the list of borrows and where they're introduced. I'm doing all this to help debug some polonius scope issues in my prototype location-sensitive analysis :3. I'll probably add member constraints soon.
const fn stability checking: also check declared language features
Fixes https://github.com/rust-lang/rust/issues/129656
`@oli-obk` I assume it is just an oversight that this didn't use `features().declared()`? Or is there a deep reason that this must only check `declared_lib_features`?
Stop using `ty::GenericPredicates` for non-predicates_of queries
`GenericPredicates` is a struct of several parts: A list of of an item's own predicates, and a parent def id (and some effects related stuff, but ignore that since it's kinda irrelevant). When instantiating these generic predicates, it calls `predicates_of` on the parent and instantiates its predicates, and appends the item's own instantiated predicates too:
acb4e8b625/compiler/rustc_middle/src/ty/generics.rs (L407-L413)
Notice how this should result in a recursive set of calls to `predicates_of`... However, `GenericPredicates` is *also* misused by a bunch of *other* queries as a convenient way of passing around a list of predicates. For these queries, we don't ever set the parent def id of the `GenericPredicates`, but if we did, then this would be very easy to mistakenly call `predicates_of` instead of some other intended parent query.
Given that footgun, and the fact that we don't ever even *use* the parent def id in the `GenericPredicates` returned from queries like `explicit_super_predicates_of`, It really has no benefit over just returning `&'tcx [(Clause<'tcx>, Span)]`.
This PR additionally opts to wrap the results of `EarlyBinder`, as we've tended to use that in the return type of these kinds of queries to properly convey that the user has params to deal with, and it also gives a convenient way of iterating over a slice of things after instantiating.
We want to allow setting this on the CLI, override it only in MIR
passes, and disable it altogether in mir-opt tests.
The default value is "only for NLL MIR dumps", which is considered off
for all intents and purposes, except for `rustc_borrowck` when an NLL
MIR dump is requested.
Implement a first version of RFC 3525: struct target features
This PR is an attempt at implementing https://github.com/rust-lang/rfcs/pull/3525, behind a feature gate `struct_target_features`.
There's obviously a few tasks that ought to be done before this is merged; in no particular order:
- add proper error messages
- add tests
- create a tracking issue for the RFC
- properly serialize/deserialize the new target_features field in `rmeta` (assuming I even understood that correctly :-))
That said, as I am definitely not a `rustc` expert, I'd like to get some early feedback on the overall approach before fixing those things (and perhaps some pointers for `rmeta`...), hence this early PR :-)
Here's an example piece of code that I have been using for testing - with the new code, the calls to intrinsics get correctly inlined:
```rust
#![feature(struct_target_features)]
use std::arch::x86_64::*;
/*
// fails to compile
#[target_feature(enable = "avx")]
struct Invalid(u32);
*/
#[target_feature(enable = "avx")]
struct Avx {}
#[target_feature(enable = "sse")]
struct Sse();
/*
// fails to compile
extern "C" fn bad_fun(_: Avx) {}
*/
/*
// fails to compile
#[inline(always)]
fn inline_fun(_: Avx) {}
*/
trait Simd {
fn do_something(&self);
}
impl Simd for Avx {
fn do_something(&self) {
unsafe {
println!("{:?}", _mm256_setzero_ps());
}
}
}
impl Simd for Sse {
fn do_something(&self) {
unsafe {
println!("{:?}", _mm_setzero_ps());
}
}
}
struct WithAvx {
#[allow(dead_code)]
avx: Avx,
}
impl Simd for WithAvx {
fn do_something(&self) {
unsafe {
println!("{:?}", _mm256_setzero_ps());
}
}
}
#[inline(never)]
fn dosomething<S: Simd>(simd: &S) {
simd.do_something();
}
fn main() {
/*
// fails to compile
Avx {};
*/
if is_x86_feature_detected!("avx") {
let avx = unsafe { Avx {} };
dosomething(&avx);
dosomething(&WithAvx { avx });
}
if is_x86_feature_detected!("sse") {
dosomething(&unsafe { Sse {} })
}
}
```
Tracking:
- https://github.com/rust-lang/rust/issues/129107
LLVM uses the word "code" to refer to a particular kind of coverage mapping.
This unrelated usage of the word is confusing, and makes it harder to introduce
types whose names correspond to the LLVM classification of coverage kinds.
Get rid of `predicates_defined_on`
This is the uncontroversial part of #129532. This simply inlines the `predicates_defined_on` into into `predicates_of`. Nothing should change here logically.
Stop storing a special inner body for the coroutine by-move body for async closures
...and instead, just synthesize an item which is treated mostly normally by the MIR pipeline.
This PR does a few things:
* We synthesize a new `DefId` for the by-move body of a closure, which has its `mir_built` fed with the output of the `ByMoveBody` MIR transformation, and some other relevant queries.
* This has the `DefKind::ByMoveBody`, which we use to distinguish it from "real" bodies (that come from HIR) which need to be borrowck'd. Introduce `TyCtxt::is_synthetic_mir` to skip over `mir_borrowck` which is called by `mir_promoted`; borrowck isn't really possible to make work ATM since it heavily relies being called on a body generated from HIR, and is redundant by the construction of the by-move-body.
* Remove the special `PassManager` hacks for handling the inner `by_move_body` stored within the coroutine's mir body. Instead, this body is fed like a regular MIR body, so it's goes through all of the `tcx.*_mir` stages normally (build -> promoted -> ...etc... -> optimized) ✨.
* Remove the `InstanceKind::ByMoveBody` shim, since now we have a "regular" def id, we can just use `InstanceKind::Item`. This also allows us to remove the corresponding hacks from codegen, such as in `fn_sig_for_fn_abi` ✨.
Notable remarks:
* ~~I know it's kind of weird to be using `DefKind::Closure` here, since it's not a distinct closure but just a new MIR body. I don't believe it really matters, but I could also use a different `DefKind`... maybe one that we could use for synthetic MIR bodies in general?~~ edit: We're doing this now.
Detect `*` operator on `!Sized` expression
The suggestion is new:
```
error[E0277]: the size for values of type `str` cannot be known at compilation time
--> $DIR/unsized-str-in-return-expr-arg-and-local.rs:15:9
|
LL | let x = *"";
| ^ doesn't have a size known at compile-time
|
= help: the trait `Sized` is not implemented for `str`
= note: all local variables must have a statically known size
= help: unsized locals are gated as an unstable feature
help: references to `!Sized` types like `&str` are `Sized`; consider not dereferencing the expression
|
LL - let x = *"";
LL + let x = "";
|
```
Fix#128199.
Retroactively feature gate `ConstArgKind::Path`
This puts the lowering introduced by #125915 under a feature gate until we fix the regressions introduced by it. Alternative to whole sale reverting the PR since it didn't seem like a very clean revert and I think this is generally a step in the right direction and don't want to get stuck landing and reverting the PR over and over :)
cc #129137 ``@camelid,`` tests taken from there. beta is branching soon so I think it makes sense to not try and rush that fix through since it wont have much time to bake and if it has issues we can't simply revert it on beta.
Fixes#128016
Pretty-print own args of existential projections (dyn-Trait w/ GAT constraints)
Previously we would just drop them. This bug isn't that significant as it can only be triggered by user code that constrains GATs inside trait object types which is currently gated under the interim feature `generic_associated_types_extended` (whose future is questionable) or on stable if the GATs are 'disabled' in dyn-Trait via `where Self: Sized` (in which case the assoc type bindings get ignored anyway (and trigger the warn-by-default lint `unused_associated_type_bounds`)), so yeah.
Affects diagnostic output and output of `std::any::type_name{_of_val}`.
Use `bool` in favor of `Option<()>` for diagnostics
We originally only supported `Option<()>` for optional notes/labels, but we now support `bool`. Let's use that, since it usually leads to more readable code.
I'm not removing the support from the derive macro, though I guess we could error on it... 🤔
Added "copy" to Debug fmt for copy operands
In MIR's debug mode (--emit mir) the printing for Operands is slightly inconsistent.
The RValues - values on the right side of an Assign - are usually printed with their Operand when they are Places.
Example:
_2 = move _3
But for arguments, the operand is omitted.
_2 = _1
I propose a change be made, to display the place with the operand.
_2 = copy _1
Move and copy have different semantics, meaning this difference is important and helpful to the user. It also adds consistency to the pretty printing.
-- EDIT --
Consider this example Rust program and its MIR output with the **updated pretty printer.**
This was generated with the arguments --emit mir --crate-type lib -Zmir-opt-level=0 (Otherwise, it's optimised away since it's a junk program).
```rust
fn main(foo: i32) {
let v = 10;
if v == 20 {
foo;
}
else {
v;
}
}
```
```MIR
// WARNING: This output format is intended for human consumers only
// and is subject to change without notice. Knock yourself out.
fn main(_1: i32) -> () {
debug foo => _1;
let mut _0: ();
let _2: i32;
let mut _3: bool;
let mut _4: i32;
let _5: i32;
let _6: i32;
scope 1 {
debug v => _2;
}
bb0: {
StorageLive(_2);
_2 = const 10_i32;
StorageLive(_3);
StorageLive(_4);
_4 = copy _2;
_3 = Eq(move _4, const 20_i32);
switchInt(move _3) -> [0: bb2, otherwise: bb1];
}
bb1: {
StorageDead(_4);
StorageLive(_5);
_5 = copy _1;
StorageDead(_5);
_0 = const ();
goto -> bb3;
}
bb2: {
StorageDead(_4);
StorageLive(_6);
_6 = copy _2;
StorageDead(_6);
_0 = const ();
goto -> bb3;
}
bb3: {
StorageDead(_3);
StorageDead(_2);
return;
}
}
```
In this example program, we can see that when we move a place, it is preceded by "move". e.g. ``` _3 = Eq(move _4, const 20_i32);```. However, when we copy a place such as ```_5 = _1;```, it is not preceded by the operand in the original printout. I propose to change the print to include the copy ```_5 = copy _1``` as in this example.
Regarding the arguments part. When I originally submitted this PR, I was under the impression this only affected the print for arguments to a function, but actually, it affects anything that uses a copy. This is preferable anyway with regard to consistency. The PR is about making ```copy``` explicit.
Use cnum for extern crate data key
Noticed this when fixing #129184. I still have yet to put up a fix for that (mostly because I'm too lazy to minimize a test, that will come soon though).
Use `FnSig` instead of raw `FnDecl` for `ForeignItemKind::Fn`, fix ICE for `Fn` trait error on safe foreign fn
Let's use `hir::FnSig` instead of `hir::FnDecl + hir::Safety` for `ForeignItemKind::Fn`. This consolidates some handling code between normal fns and foreign fns.
Separetly, fix an ICE where we weren't handling `Fn` trait errors for safe foreign fns.
If perf is bad for the first commit, I can rework the ICE fix to not rely on it. But if perf is good, I prefer we fix and clean up things all at once 👍
r? spastorino
Fixes#128764
Return correct HirId when finding body owner in diagnostics
Fixes#129145Fixes#128810
r? ```@compiler-errors```
```rust
fn generic<const N: u32>() {}
trait Collate<const A: u32> {
type Pass;
fn collate(self) -> Self::Pass;
}
impl<const B: u32> Collate<B> for i32 {
type Pass = ();
fn collate(self) -> Self::Pass {
generic::<{ true }>()
//~^ ERROR: mismatched types
}
}
```
When type checking the `{ true }` anon const we would error with a type mismatch. This then results in diagnostics code attempting to check whether its due to a type mismatch with the return type. That logic was implemented by walking up the hir until we reached the body owner, except instead of using the `enclosing_body_owner` function it special cased various hir nodes incorrectly resulting in us walking out of the anon const and stopping at `fn collate` instead.
This then resulted in diagnostics logic inside of the anon consts `ParamEnv` attempting to do trait solving involving the `<i32 as Collate<B>>::Pass` type which ICEs because it is in the wrong environment.
I have rewritten this function to just walk up until it hits the `enclosing_body_owner` and made some other changes since I found this pretty hard to read/understand. Hopefully it's easier to understand now, it also makes it more obvious that this is not implemented in a very principled way and is definitely missing cases :)
mir/pretty: use `Option` instead of `Either<Once, Empty>`
`Either` is wasteful for a one-or-none iterator, especially since `Once`
is already an `option::IntoIter` internally. We don't really need any of
the iterator mechanisms in this case, just a single conditional insert.
`Either` is wasteful for a one-or-none iterator, especially since `Once`
is already an `option::IntoIter` internally. We don't really need any of
the iterator mechanisms in this case, just a single conditional insert.
Special-case alias ty during the delayed bug emission in `try_from_lit`
This PR tries to fix#116308.
A delayed bug in `try_from_lit` will not be emitted so that the compiler will not ICE when it sees the pair `(ast::LitKind::Int, ty::TyKind::Alias)` in `lit_to_const` (called from `try_from_lit`).
This PR is related to an unstable feature `adt_const_params` (#95174).
r? ``@BoxyUwU``
`-Znext-solver` caching
This PR has two major changes while also fixing multiple issues found via fuzzing.
The main optimization is the ability to not discard provisional cache entries when popping the highest cycle head the entry depends on. This fixes the hang in Fuchsia with `-Znext-solver=coherence`.
It also bails if the result of a fixpoint iteration is ambiguous, even without reaching a fixpoint. This is necessary to avoid exponential blowup if a coinductive cycle results in ambiguity, e.g. due to unknowable candidates in coherence.
Updating stack entries pretty much exclusively happens lazily now, so `fn check_invariants` ended up being mostly useless and I've removed it. See https://gist.github.com/lcnr/8de338fdb2685581e17727bbfab0622a for the invariants we would be able to assert with it.
For a general overview, see the in-process update of the relevant rustc-dev-guide chapter: https://hackmd.io/1ALkSjKlSCyQG-dVb_PUHw
r? ```@compiler-errors```
Shrink `TyKind::FnPtr`.
By splitting the `FnSig` within `TyKind::FnPtr` into `FnSigTys` and `FnHeader`, which can be packed more efficiently. This reduces the size of the hot `TyKind` type from 32 bytes to 24 bytes on 64-bit platforms. This reduces peak memory usage by a few percent on some benchmarks. It also reduces cache misses and page faults similarly, though this doesn't translate to clear cycles or wall-time improvements on CI.
r? `@compiler-errors`
miri: make vtable addresses not globally unique
Miri currently gives vtables a unique global address. That's not actually matching reality though. So this PR enables Miri to generate different addresses for the same type-trait pair.
To avoid generating an unbounded number of `AllocId` (and consuming unbounded amounts of memory), we use the "salt" technique that we also already use for giving constants non-unique addresses: the cache is keyed on a "salt" value n top of the actually relevant key, and Miri picks a random salt (currently in the range `0..16`) each time it needs to choose an `AllocId` for one of these globals -- that means we'll get up to 16 different addresses for each vtable. The salt scheme is integrated into the global allocation deduplication logic in `tcx`, and also used for functions and string literals. (So this also fixes the problem that casting the same function to a fn ptr over and over will consume unbounded memory.)
r? `@saethlin`
Fixes https://github.com/rust-lang/miri/issues/3737
Store `do_not_recommend`-ness in impl header
Alternative to #128674
It's less flexible, but also less invasive. Hopefully it's also performant. I'd recommend we think separately about the design for how to gate arbitrary diagnostic attributes moving forward.
Normalize struct tail properly for `dyn` ptr-to-ptr casting in new solver
Realized that the new solver didn't handle ptr-to-ptr casting correctly.
r? lcnr
Built on #128694
doing so requires overwriting global cache entries and
generally adds significant complexity to the solver. This is
also only ever done for root goals, so it feels easier to wrap
the `evaluate_canonical_goal` in an ordinary query if
necessary.
Differentiate between methods and associated functions in diagnostics
Accurately refer to assoc fn without receiver as assoc fn instead of methods. Add `AssocItem::descr` method to centralize where we call methods and associated functions.
Stabilize `min_exhaustive_patterns`
## Stabilisation report
I propose we stabilize the [`min_exhaustive_patterns`](https://github.com/rust-lang/rust/issues/119612) language feature.
With this feature, patterns of empty types are considered unreachable when matched by-value. This allows:
```rust
enum Void {}
fn foo() -> Result<u32, Void>;
fn main() {
let Ok(x) = foo();
// also
match foo() {
Ok(x) => ...,
}
}
```
This is a subset of the long-unstable [`exhaustive_patterns`](https://github.com/rust-lang/rust/issues/51085) feature. That feature is blocked because omitting empty patterns is tricky when *not* matched by-value. This PR stabilizes the by-value case, which is not tricky.
The not-by-value cases (behind references, pointers, and unions) stay as they are today, e.g.
```rust
enum Void {}
fn foo() -> Result<u32, &Void>;
fn main() {
let Ok(x) = foo(); // ERROR: missing `Err(_)`
}
```
The consequence on existing code is some extra "unreachable pattern" warnings. This is fully backwards-compatible.
### Comparison with today's rust
This proposal only affects match checking of empty types (i.e. types with no valid values). Non-empty types behave the same with or without this feature. Note that everything below is phrased in terms of `match` but applies equallly to `if let` and other pattern-matching expressions.
To be precise, a visibly empty type is:
- an enum with no variants;
- the never type `!`;
- a struct with a *visible* field of a visibly empty type (and no #[non_exhaustive] annotation);
- a tuple where one of the types is visibly empty;
- en enum with all variants visibly empty (and no `#[non_exhaustive]` annotation);
- a `[T; N]` with `N != 0` and `T` visibly empty;
- all other types are nonempty.
(An extra change was proposed below: that we ignore #[non_exhaustive] for structs since adding fields cannot turn an empty struct into a non-empty one)
For normal types, exhaustiveness checking requires that we list all variants (or use a wildcard). For empty types it's more subtle: in some cases we require a `_` pattern even though there are no valid values that can match it. This is where the difference lies regarding this feature.
#### Today's rust
Under today's rust, a `_` is required for all empty types, except specifically: if the matched expression is of type `!` (the never type) or `EmptyEnum` (where `EmptyEnum` is an enum with no variants), then the `_` is not required.
```rust
let foo: Result<u32, !> = ...;
match foo {
Ok(x) => ...,
Err(_) => ..., // required
}
let foo: Result<u32, &!> = ...;
match foo {
Ok(x) => ...,
Err(_) => ..., // required
}
let foo: &! = ...;
match foo {
_ => ..., // required
}
fn blah(foo: (u32, !)) {
match foo {
_ => ..., // required
}
}
unsafe {
let ptr: *const ! = ...;
match *ptr {} // allowed
let ptr: *const (u32, !) = ...;
match *ptr {
(x, _) => { ... } // required
}
let ptr: *const Result<u32, !> = ...;
match *ptr {
Ok(x) => { ... }
Err(_) => { ... } // required
}
}
```
#### After this PR
After this PR, a pattern of an empty type can be omitted if (and only if):
- the match scrutinee expression has type `!` or `EmptyEnum` (like before);
- *or* the empty type is matched by value (that's the new behavior).
In all other cases, a `_` is required to match on an empty type.
```rust
let foo: Result<u32, !> = ...;
match foo {
Ok(x) => ..., // `Err` not required
}
let foo: Result<u32, &!> = ...;
match foo {
Ok(x) => ...,
Err(_) => ..., // required because `!` is under a dereference
}
let foo: &! = ...;
match foo {
_ => ..., // required because `!` is under a dereference
}
fn blah(foo: (u32, !)) {
match foo {} // allowed
}
unsafe {
let ptr: *const ! = ...;
match *ptr {} // allowed
let ptr: *const (u32, !) = ...;
match *ptr {
(x, _) => { ... } // required because the matched place is under a (pointer) dereference
}
let ptr: *const Result<u32, !> = ...;
match *ptr {
Ok(x) => { ... }
Err(_) => { ... } // required because the matched place is under a (pointer) dereference
}
}
```
### Documentation
The reference does not say anything specific about exhaustiveness checking, hence there is nothing to update there. The nomicon does, I opened https://github.com/rust-lang/nomicon/pull/445 to reflect the changes.
### Tests
The relevant tests are in `tests/ui/pattern/usefulness/empty-types.rs`.
### Unresolved Questions
None that I know of.
try-job: dist-aarch64-apple
Cache supertrait outlives of impl header for soundness check
This caches the results of computing the transitive supertraits of an impl and filtering it to its outlives obligations. This is purely an optimization to improve https://github.com/rust-lang/rust/pull/124336.
Stop unnecessarily taking GenericPredicates by `&self`
This results in overcapturing in edition 2024, and is unnecessary since `GenericPredicates: Copy`.
Accurately refer to assoc fn without receiver as assoc fn instead of methods.
Add `AssocItem::descr` method to centralize where we call methods and associated functions.
Miscellaneous improvements to struct tail normalization
1. Make checks for foreign tails more accurate by normalizing the struct tail. I didn't write a test for this one.
2. Normalize when computing struct tail for `offset_of` for slice/str. This fixes the new solver only.
3. Normalizing when computing tails for disaligned reference check. This fixes both solvers.
r? lcnr
By splitting the `FnSig` within `TyKind::FnPtr` into `FnSigTys` and
`FnHeader`, which can be packed more efficiently. This reduces the size
of the hot `TyKind` type from 32 bytes to 24 bytes on 64-bit platforms.
This reduces peak memory usage by a few percent on some benchmarks. It
also reduces cache misses and page faults similarly, though this doesn't
translate to clear cycles or wall-time improvements on CI.
```
error[E0277]: the size for values of type `str` cannot be known at compilation time
--> $DIR/unsized-str-in-return-expr-arg-and-local.rs:15:9
|
LL | let x = *"";
| ^ doesn't have a size known at compile-time
|
= help: the trait `Sized` is not implemented for `str`
= note: all local variables must have a statically known size
= help: unsized locals are gated as an unstable feature
help: references are always `Sized`, even if they point to unsized data; consider not dereferencing the expression
|
LL - let x = *"";
LL + let x = "";
|
```
MIR required_consts, mentioned_items: ensure we do not forget to fill these lists
Bodies initially get created with empty required_consts and mentioned_items, but at some point those should be filled. Make sure we notice when that is forgotten.
The pattern-analysis code needs to print patterns, as part of its user-visible
diagnostics. But it never actually tries to print "real" patterns! Instead, it
only ever prints synthetic patterns that it has reconstructed from its own
internal represenations.
We can therefore simultaneously remove two obstacles to changing `thir::Pat`,
by having the pattern-analysis code use its own dedicated type for building
printable patterns, and then making `thir::Pat` not printable at all.
This reverts commit ae0ec731a8.
The original change in #128304 was intended to be a step towards being able to
print `thir::Pat` even after switching to `PatId`.
But because the only patterns that need to be printed are the synthetic ones
created by pattern analysis (for diagnostic purposes only), it makes more sense
to completely separate the printable patterns from the real THIR patterns.
Delegation: support generics for delegation from free functions
(The PR was split from https://github.com/rust-lang/rust/pull/123958, explainer - https://github.com/Bryanskiy/posts/blob/master/delegation%20in%20generic%20contexts.md)
This PR implements generics inheritance from free functions to free functions and trait methods.
#### free functions to free functions:
```rust
fn to_reuse<T: Clone>(_: T) {}
reuse to_reuse as bar;
// desugaring:
fn bar<T: Clone>(x: T) {
to_reuse(x)
}
```
Generics, predicates and signature are simply copied. Generic arguments in paths are ignored during generics inheritance:
```rust
fn to_reuse<T: Clone>(_: T) {}
reuse to_reuse::<u8> as bar;
// desugaring:
fn bar<T: Clone>(x: T) {
to_reuse::<u8>(x) // ERROR: mismatched types
}
```
Due to implementation limitations callee path is lowered without modifications. Therefore, it is a compilation error at the moment.
#### free functions to trait methods:
```rust
trait Trait<'a, A> {
fn foo<'b, B>(&self, x: A, y: B) {...}
}
reuse Trait::foo;
// desugaring:
fn foo<'a, 'b, This: Trait<'a, A>, A, B>(this: &This, x: A, y: B) {
Trait::foo(this, x, y)
}
```
The inheritance is similar to the previous case but with some corrections:
- `Self` parameter converted into `T: Trait`
- generic parameters need to be reordered so that lifetimes go first
Arguments are similarly ignored.
---
In the future, we plan to support generic inheritance for delegating from all contexts to all contexts (from free/trait/impl to free/trait /impl). These cases were considered first as the simplest from the implementation perspective.
Isolate the diagnostic code that expects `thir::Pat` to be printable
Currently, `thir::Pat` implements `fmt::Display` (and `IntoDiagArg`) directly, for use by a few diagnostics.
That makes it tricky to experiment with alternate representations for THIR patterns, because the patterns currently need to be printable on their own. That immediately rules out possibilities like storing subpatterns as a `PatId` index into a central list (instead of the current directly-owned `Box<Pat>`).
This PR therefore takes an incremental step away from that obstacle, by removing `thir::Pat` from diagnostic structs in `rustc_pattern_analysis`, and hiding the pattern-printing process behind a single public `Pat::to_string` method. Doing so makes it easier to identify and update the code that wants to print patterns, and gives a place to pass in additional context in the future if necessary.
---
I'm currently not sure whether switching over to `PatId` is actually desirable or not, but I think this change makes sense on its own merits, by reducing the coupling between `thir::Pat` and the pattern-analysis error types.
miri: fix offset_from behavior on wildcard pointers
offset_from wouldn't behave correctly when the "end" pointer was a wildcard pointer (result of an int2ptr cast) just at the end of the allocation. Fix that by expressing the "same allocation" check in terms of two `check_ptr_access_signed` instead of something specific to offset_from, which is both more canonical and works better with wildcard pointers.
The second commit just improves diagnostics: I wanted the "pointer is dangling (has no provenance)" message to say how many bytes of memory it expected to see (since if it were 0 bytes, this would actually be legal, so it's good to tell the user that it's not 0 bytes). And then I was annoying that the error looks so different for when you deref a dangling pointer vs an out-of-bounds pointer so I made them more similar.
Fixes https://github.com/rust-lang/miri/issues/3767
This gives a clearer view of the (diagnostic) code that expects to be able to
print THIR patterns, and makes it possible to experiment with requiring some
kind of context (for ID lookup) when printing patterns.
Let InstCombine remove Clone shims inside Clone shims
The Clone shims that we generate tend to recurse into other Clone shims, which gets very silly very quickly. Here's our current state: https://godbolt.org/z/E69YeY8eq
So I've added InstSimplify to the shims optimization passes, and improved `is_trivially_pure_clone_copy` so that it can delete those calls inside the shim. This makes the shim way smaller because most of its size is the required ceremony for unwinding.
This change also completely breaks the UI test added for https://github.com/rust-lang/rust/issues/104870. With this PR, that program ICEs in MIR type checking because `is_trivially_pure_clone_copy` and the trait solver disagree on whether `*mut u8` is `Copy`. And adding the requisite `Copy` impl to make them agree makes the test not generate any diagnostics. Considering that I spent most of my time on this PR fixing `#![no_core]` tests, I would prefer to just delete this one. The maintenance burden of `#![no_core]` is uniquely high because when they break they tend to break in very confusing ways.
try-job: x86_64-mingw
Switch from `derivative` to `derive-where`
This is a part of the effort to get rid of `syn 1.*` in compiler's dependencies: #109302
Derivative has not been maintained in nearly 3 years[^1]. It also depends on `syn 1.*`.
This PR replaces `derivative` with `derive-where`[^2], a not dead alternative, which uses `syn 2.*`.
A couple of `Debug` formats have changed around the skipped fields[^3], but I doubt this is an issue.
[^1]: https://github.com/mcarton/rust-derivative/issues/117
[^2]: https://lib.rs/crates/derive-where
[^3]: See the changes in `tests/ui`
Fix malformed suggestion for repeated maybe unsized bounds
Fixes#127441
Now when we encounter something like `foo(a : impl ?Sized + ?Sized)`, instead of suggesting removal of both bounds and leaving `foo(a: impl )` behind, we suggest changing the first bound to `Sized` and removing the second bound, resulting in `foo(a: impl Sized)`.
Although the issue was reported for impl trait types, it also occurred with regular param bounds. So if we encounter `foo<T: ?Sized + ?Sized>(a: T)` we now detect that all the bounds are `?Sized` and therefore emit the suggestion to remove the entire predicate `: ?Sized + ?Sized` resulting in `foo<T>(a: T)`.
Lastly, if we encounter a situation where some of the bounds are something other than `?Sized`, then we emit separate removal suggestions for each `?Sized` bound. E.g. if we see `foo(a: impl ?Sized + Bar + ?Sized)` or `foo<T: ?Sized + Bar + ?Sized>(a: T)` we emit suggestions such that the user will be left with `foo(a : impl Bar)` or `foo<T: Bar>(a: T)` respectively.
Try to fix ICE from re-interning an AllocId with different allocation contents
As far as I can tell, based on my investigation in https://github.com/rust-lang/rust/issues/126741, the racy decoding scheme implemented here was never fully correct, but the arrangement of Allocations that's required to ICE the compiler requires some very specific MIR optimizations to create. As far as I can tell, GVN likes to create the problematic pattern, which is why we're noticing this problem now.
So the solution here is to not do racy decoding. If two threads race to decoding an AllocId, one of them is going to sit on a lock until the other is done.
Remove unnecessary impl sorting in queries and metadata
Removes unnecessary impl sorting because queries already return their keys in HIR definition order: https://github.com/rust-lang/rust/issues/120371#issuecomment-1926422838
r? `@cjgillot` or `@lcnr` -- unless I totally misunderstood what was being asked for here? 😆fixes#120371
Add a hook for `should_codegen_locally`
This PR lifts the module-local function `should_codegen_locally` to `TyCtxt` as a hook.
In addition to monomorphization, this function is used for checking the dependency of `compiler_builtins` on other libraries. Moving this function to the hooks also makes overriding it possible for the tools that use the rustc interface.
Clean up a few minor refs in `format!` macro, as it has a performance cost. Apparently the compiler is unable to inline `format!("{}", &variable)`, and does a run-time double-reference instead (format macro already does one level referencing). Inlining format args prevents accidental `&` misuse.
