Remove unnecessary binder from `get_impl_future_output_ty`
We never construct an `async fn` with a higher-ranked `impl Future` bound anyways, and basically all the call-sites already skip the binder.
Do not record unresolved const vars in generator interior
Don't record types in the generator interior when we see unresolved const variables.
We already do this for associated types -- this is important to avoid unresolved inference variables in the generator results during writeback, since the writeback results get stable hashed in incremental mode.
Fixes#104787
Add `ConstKind::Expr`
Starting to implement `ty::ConstKind::Abstract`, most of the match cases are stubbed out, some I was unsure what to add, others I didn't want to add until a more complete implementation was ready.
r? `@lcnr`
Initial pass at expr/abstract const/s
Address comments
Switch to using a list instead of &[ty::Const], rm `AbstractConst`
Remove try_unify_abstract_consts
Update comments
Add edits
Recurse more
More edits
Prevent equating associated consts
Move failing test to ui
Changes this test from incremental to ui, and mark it as failing and a known bug.
Does not cause the compiler to ICE, so should be ok.
Rollup of 9 pull requests
Successful merges:
- #104420 (Fix doc example for `wrapping_abs`)
- #104499 (rustdoc JSON: Use `Function` everywhere and remove `Method`)
- #104500 (`rustc_ast`: remove `ref` patterns)
- #104511 (Mark functions created for `raw-dylib` on x86 with DllImport storage class)
- #104595 (Add `PolyExistentialPredicate` type alias)
- #104605 (deduplicate constant evaluation in cranelift backend)
- #104628 (Revert "Update CI to use Android NDK r25b")
- #104662 (Streamline deriving on packed structs.)
- #104667 (Revert formatting changes of a test)
Failed merges:
r? `@ghost`
`@rustbot` modify labels: rollup
Add `PolyExistentialPredicate` type alias
Wrapping `ExistentialPredicate`s in a binder is very common, and this alias already exists for the `PolyExistential{TraitRef,Projection}` types.
Support using `Self` or projections inside an RPIT/async fn
I reuse the same idea as https://github.com/rust-lang/rust/pull/103449 to use variances to encode whether a lifetime parameter is captured by impl-trait.
The current implementation of async and RPIT replace all lifetimes from the parent generics by `'static`. This PR changes the scheme
```rust
impl<'a> Foo<'a> {
fn foo<'b, T>() -> impl Into<Self> + 'b { ... }
}
opaque Foo::<'_a>::foo::<'_b, T>::opaque<'b>: Into<Foo<'_a>> + 'b;
impl<'a> Foo<'a> {
// OLD
fn foo<'b, T>() -> Foo::<'static>::foo::<'static, T>::opaque::<'b> { ... }
^^^^^^^ the `Self` becomes `Foo<'static>`
// NEW
fn foo<'b, T>() -> Foo::<'a>::foo::<'b, T>::opaque::<'b> { ... }
^^ the `Self` stays `Foo<'a>`
}
```
There is the same issue with projections. In the example, substitute `Self` by `<T as Trait<'b>>::Assoc` in the sugared version, and `Foo<'_a>` by `<T as Trait<'_b>>::Assoc` in the desugared one.
This allows to support `Self` in impl-trait, since we do not replace lifetimes by `'static` any more. The same trick allows to use projections like `T::Assoc` where `Self` is allowed. The feature is gated behind a `impl_trait_projections` feature gate.
The implementation relies on 2 tweaking rules for opaques in 2 places:
- we only relate substs that correspond to captured lifetimes during TypeRelation;
- we only list captured lifetimes in choice region computation.
For simplicity, I encoded the "capturedness" of lifetimes as a variance, `Bivariant` vs `Invariant` for unused vs captured lifetimes. The `variances_of` query used to ICE for opaques.
Impl-trait that do not reference `Self` or projections will have their variances as:
- `o` (invariant) for each parent type or const;
- `*` (bivariant) for each parent lifetime --> will not participate in borrowck;
- `o` (invariant) for each own lifetime.
Impl-trait that does reference `Self` and/or projections will have some parent lifetimes marked as `o` (as the example above), and participate in type relation and borrowck. In the example above, `variances_of(opaque) = ['_a: o, '_b: *, T: o, 'b: o]`.
r? types
cc `@compiler-errors` , as you asked about the issue with `Self` and projections.
Use `ErrorGuaranteed::unchecked_claim_error_was_emitted` less
there are only like 3 or 4 call sites left after this but it wasnt obvious to me how to remove them
nll: correctly deal with bivariance
fixes#104409
when in a bivariant context, relating stuff should always trivially succeed. Also changes the mir validator to correctly deal with higher ranked regions.
r? types cc ``@RalfJung``
Remove allow(rustc::potential_query_instability) in rustc_trait_selection
Related to https://github.com/rust-lang/rust/issues/84447
This PR needs to be benchmarked to check for regressions.
Better error for HRTB error from generator interior
cc #100013
This is just a first pass at an error. It could be better, and shouldn't really be emitted in the first place. But this is better than what was being emitted before.
Make InferCtxtExt use a FxIndexMap
This should be faster, because the map is only being used to iterate,
which is supposed to be faster with the IndexMap
Make the user_computed_preds use an IndexMap
It is being used mostly for iteration, so the change shouldn't result in
a perf hit
Make the RegionDeps fields use an IndexMap
This change could be a perf hit. Both `larger` and `smaller` are used
for iteration, but they are also used for insertions.
Make types_without_default_bounds use an IndexMap
It uses extend, but it also iterates and removes items. Not sure if
this will be a perf hit.
Make InferTtxt.reported_trait_errors use an IndexMap
This change brought a lot of other changes. The map seems to have been
mostly used for iteration, so the performance shouldn't suffer.
Add FIXME to change ProvisionalEvaluationCache.map to use an IndexMap
Right now this results in a perf hit. IndexMap doesn't have
the `drain_filter` API, so in `on_completion` we now need to iterate two
times over the map.
Remove #![allow(rustc::potential_query_instability)] from rustc_infer
Related to #84447
This PR probably needs to be benchmarked to check for regressions.
Change #[suggestion_*] attributes to use style="..."
As discussed [on Zulip](https://rust-lang.zulipchat.com/#narrow/stream/336883-i18n/topic/.23100717.20tool_only_span_suggestion), this changes `#[(multipart_)suggestion_{short,verbose,hidden}(...)]` attributes to plain `#[(multipart_)suggestion(...)]` attributes with a `style = "{short,verbose,hidden}"` parameter.
It also adds a new style, `tool-only`, that corresponds to `tool_only_span_suggestion`/`tool_only_multipart_suggestion` and causes the suggestion to not be shown in human-readable output at all.
Best reviewed commit-by-commit, there's a bit of noise in there.
cc #100717 `@compiler-errors`
r? `@davidtwco`
(almost) Always use `ObligationCtxt` when dealing with canonical queries
Hope this is a step in the right direction. cc rust-lang/types-team#50.
r? `@lcnr`
spastorino noticed some silly expressions like `item_id.def_id.def_id`.
This commit renames several `def_id: OwnerId` fields as `owner_id`, so
those expressions become `item_id.owner_id.def_id`.
`item_id.owner_id.local_def_id` would be even clearer, but the use of
`def_id` for values of type `LocalDefId` is *very* widespread, so I left
that alone.
Change reported_violations to use IndexSet
It is being used to iterate and to insert, without a lot of lookups
so hopefully it won't be a perf hit
Change MiniGraph.nodes to use IndexSet
It is being used to iterate and to insert, without performing lookups
so hopefully it won't be a perf hit
Change RegionConstraintData.givens to a FxIndexSet
This might result in a perf hit. Remove was being used in `givens`,
and `FxIndexSet` doesn't allow calling remove without losing the
fixed iteration order. So it was necessary to change remove to
`shift_remove`, but this method is slower.
Change OpaqueTypesVisitor to use stable sets and maps
This could also be a perf hit.
Make TraitObject visitor use a stable set
Don't carry MIR location in `ConstraintCategory::CallArgument`
It turns out that `ConstraintCategory::CallArgument` cannot just carry a MIR location in it, since we may bubble them up to totally different MIR bodies.
So instead, revert the commit a6b5f95fb0, and instead just erase regions from the original `Option<Ty<'tcx>>` that it carried, so that it doesn't ICE with the changes in #103220.
Best reviewed in parts -- the first is just a revert, and the second is where the meaningful changes happen.
Fixes#103624
Clean up hidden type registration
work on https://github.com/rust-lang/rust/issues/101186
Actually passing down the relation and using it instead of `eq` for the hidden type comparison has *no* effect whatsoever and allows for no further improvements at the call sites. I decided the increased complexity was not worth it and thus did not include that change in this PR.
r? `@compiler-errors`
Flatten diagnostic slug modules
This makes it easier to grep for the slugs in the code.
See https://rust-lang.zulipchat.com/#narrow/stream/131828-t-compiler/topic/Localization.20infra.20interferes.20with.20grepping.20for.20error for more discussion about it.
This was mostly done with a few regexes and a bunch of manual work. This also exposes a pretty annoying inconsistency for the extra labels. Some of the extra labels are defined as additional properties in the fluent message (which makes them not prefixed with the crate name) and some of them are new fluent messages themselves (which makes them prefixed with the crate name). I don't know whether we want to clean this up at some point but it's useful to know.
r? `@davidtwco`
Handle return-position `impl Trait` in traits properly in `register_hidden_type`
The bounds that we get by calling `bound_explicit_item_bounds` from an RPITIT have projections, not opaques, but when we're *registering* an opaque, we want to treat it like an opaque.
Coincidentally fixes#102688 as well, which makes sense, since that was failing because we were inferring an opaque type to be equal to itself (opaque cycle error => "cannot resolve opaque type").
Fixes#103352
r? ```@oli-obk```
translation: doc comments with derives, subdiagnostic-less enum variants, more derive use
- Adds support for `doc` attributes in the diagnostic derives so that documentation comments don't result in the derive failing.
- Adds support for enum variants in the subdiagnostic derive to not actually correspond to an addition to a diagnostic.
- Made use of the derive in more places in the `rustc_ast_lowering`, `rustc_ast_passes`, `rustc_lint`, `rustc_session`, `rustc_infer` - taking advantage of recent additions like eager subdiagnostics, multispan suggestions, etc.
cc #100717
`AddToDiagnostic::add_to_diagnostic_with` is similar to the previous
`AddToDiagnostic::add_to_diagnostic` but takes a function that can be
used by the caller to modify diagnostic messages originating from the
subdiagnostic (such as performing translation eagerly).
`add_to_diagnostic` now just calls `add_to_diagnostic_with` with an
empty closure.
Signed-off-by: David Wood <david.wood@huawei.com>
rename `ImplItemKind::TyAlias` to `ImplItemKind::Type`
The naming of this variant seems inconsistent given that this is not really a "type alias", and the associated type variant for `TraitItemKind` is just called `Type`.
Suggest `.into()` when all other coercion suggestions fail
Also removes some bogus suggestions because we now short-circuit when offering coercion suggestions(instead of, for example, suggesting every one that could possibly apply)
Fixes#102415
`Res::SelfTy` currently has two `Option`s. When the second one is `Some`
the first one is never consulted. So we can split it into two variants,
`Res::SelfTyParam` and `Res::SelfTyAlias`, reducing the size of `Res`
from 24 bytes to 12. This then shrinks `hir::Path` and
`hir::PathSegment`, which are the HIR types that take up the most space.
Neither require nor imply lifetime bounds on opaque type for well formedness
The actual hidden type can live arbitrarily longer than any individual lifetime and arbitrarily shorter than all but one of the lifetimes.
fixes#86218fixes#84305
This is a **breaking change** but it is a necessary soundness fix
implied_bounds: deal with inference vars
fixes#101951
while computing implied bounds for `<<T as ConstructionFirm>::Builder as BuilderFn<'_>>::Output` normalization replaces a projection with an inference var (adding a `Projection` obligation). Until we prove that obligation, this inference var remains unknown, which caused us to miss an implied bound necessary to prove that the unnormalized projection from the trait method signature is wf.
r? types
fix a ui test
use `into`
fix clippy ui test
fix a run-make-fulldeps test
implement `IntoQueryParam<DefId>` for `OwnerId`
use `OwnerId` for more queries
change the type of `ParentOwnerIterator::Item` to `(OwnerId, OwnerNode)`
Introduce mir::Unevaluated
Previously the distinction between unevaluated constants in the type-system and in mir was not explicit and a little confusing. Probably better to introduce its own type for that.
r? `@lcnr`
Rollup of 8 pull requests
Successful merges:
- #100734 (Split out async_fn_in_trait into a separate feature)
- #101664 (Note if mismatched types have a similar name)
- #101815 (Migrated the rustc_passes annotation without effect diagnostic infrastructure)
- #102042 (Distribute rust-docs-json via rustup.)
- #102066 (rustdoc: remove unnecessary `max-width` on headers)
- #102095 (Deduplicate two functions that would soon have been three)
- #102104 (Set 'exec-env:RUST_BACKTRACE=0' in const-eval-select tests)
- #102112 (Allow full relro on powerpc64-unknown-linux-gnu)
Failed merges:
r? `@ghost`
`@rustbot` modify labels: rollup
Note if mismatched types have a similar name
If users get a type error between similarly named types, it will point out that these are actually different types, and where they were defined.
Const unification is already infallible, remove the error handling logic
r? `@lcnr`
is this expected to be used in the future? Right now it is dead code.
Rollup of 8 pull requests
Successful merges:
- #101598 (Update rustc's information on Android's sanitizers)
- #102036 (Remove use of `io::ErrorKind::Other` in std)
- #102037 (Make cycle errors recoverable)
- #102069 (Skip `Equate` relation in `handle_opaque_type`)
- #102076 (rustc_transmute: fix big-endian discriminants)
- #102107 (Add missing space between notable trait tooltip and where clause)
- #102119 (Fix a typo “pararmeter” in error message)
- #102131 (Added which number is computed in compute_float.)
Failed merges:
r? `@ghost`
`@rustbot` modify labels: rollup
Make cycle errors recoverable
In particular, this allows rustdoc to recover from cycle errors when normalizing associated types for documentation.
In the past, ```@jackh726``` has said we need to be careful about overflow errors: https://github.com/rust-lang/rust/pull/91430#issuecomment-983997013
> Off the top of my head, we definitely should be careful about treating overflow errors the same as
"not implemented for some reason" errors. Otherwise, you could end up with behavior that is
different depending on recursion depth. But, that might be context-dependent.
But cycle errors should be safe to unconditionally report; they don't depend on the recursion depth, they will always be an error whenever they're encountered.
Helps with https://github.com/rust-lang/rust/issues/81091.
r? ```@lcnr``` cc ```@matthewjasper```
Normalize opaques w/ bound vars
First, we reenable normalization of opaque types with escaping late bound regions to fix rust-lang/miri#2433. This essentially reverts #89285.
Second, we mitigate the perf regression found in #88862 by simplifying the way that we relate (sub and eq) GeneratorWitness types.
This relies on the fact that we construct these GeneratorWitness types somewhat particularly (with all free regions found in the witness types replaced with late bound regions) -- but those bound regions really should be treated as existential regions, not universal ones. Those two facts leads me to believe that we do not need to use the full `higher_ranked_sub` machinery to relate two generator witnesses. I'm pretty confident that this is correct, but I'm glad to discuss this further.
Move and rename `SessionDiagnostic` & `SessionSubdiagnostic` traits and macros
After PR #101434, we want to:
- [x] Move `SessionDiagnostic` to `rustc_errors`.
- [x] Add `emit_` methods that accept `impl SessionDiagnostic` to `Handler`.
- [x] _(optional)_ Rename trait `SessionDiagnostic` to `DiagnosticHandler`.
- [x] _(optional)_ Rename macro `SessionDiagnostic` to `DiagnosticHandler`.
- [x] Update Rustc Dev Guide and Docs to reflect these changes. https://github.com/rust-lang/rustc-dev-guide/pull/1460
Now I am having build issues getting the compiler to build when trying to rename the macro.
<details>
<summary>See diagnostics errors and context when building.</summary>
```
error: diagnostics should only be created in `SessionDiagnostic`/`AddSubdiagnostic` impls
--> compiler/rustc_attr/src/session_diagnostics.rs:13:10
|
13 | #[derive(DiagnosticHandler)]
| ^^^^^^^^^^^^^^^^^ in this derive macro expansion
|
::: /Users/jhonny/.cargo/registry/src/github.com-1ecc6299db9ec823/synstructure-0.12.6/src/macros.rs:94:9
|
94 | / pub fn $derives(
95 | | i: $crate::macros::TokenStream
96 | | ) -> $crate::macros::TokenStream {
| |________________________________________- in this expansion of `#[derive(DiagnosticHandler)]`
|
note: the lint level is defined here
--> compiler/rustc_attr/src/lib.rs:10:9
|
10 | #![deny(rustc::diagnostic_outside_of_impl)]
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
```
And also this one:
```
error: diagnostics should only be created in `SessionDiagnostic`/`AddSubdiagnostic` impls
--> compiler/rustc_attr/src/session_diagnostics.rs:213:32
|
213 | let mut diag = handler.struct_span_err_with_code(
| ^^^^^^^^^^^^^^^^^^^^^^^^^
```
> **Note**
> Can't find where this message is coming from, because you can see in [this experimental branch](https://github.com/JhonnyBillM/rust/tree/experimental/trying-to-rename-session-diagnostic-macro) that I updated all errors and diags to say:
> error: diagnostics should only be created in **`DiagnosticHandler`**/`AddSubdiagnostic` impls
> and not:
> error: diagnostics should only be created in **`SessionDiagnostic`**/`AddSubdiagnostic` impls
</details>
I tried building the compiler in different ways (playing with the stages etc), but nothing worked.
## Question
**Do we need to build or do something different when renaming a macro and identifiers?**
For context, see experimental commit f2193a98b4 where the macro and symbols are renamed, but it doesn't compile.
FIX - ambiguous Diagnostic link in docs
UPDATE - rename diagnostic_items to IntoDiagnostic and AddToDiagnostic
[Gardening] FIX - formatting via `x fmt`
FIX - rebase conflicts. NOTE: Confirm wheather or not we want to handle TargetDataLayoutErrorsWrapper this way
DELETE - unneeded allow attributes in Handler method
FIX - broken test
FIX - Rebase conflict
UPDATE - rename residual _SessionDiagnostic and fix LintDiag link
In particular, this allows rustdoc to recover from cycle errors when normalizing associated types for documentation.
In the past, `@jackh726` has said we need to be careful about overflow errors:
> Off the top of my head, we definitely should be careful about treating overflow errors the same as
"not implemented for some reason" errors. Otherwise, you could end up with behavior that is
different depending on recursion depth. But, that might be context-dependent.
But cycle errors should be safe to unconditionally report; they don't depend on the recursion depth, they will always be an error whenever they're encountered.
On later stages, the feature is already stable.
Result of running:
rg -l "feature.let_else" compiler/ src/librustdoc/ library/ | xargs sed -s -i "s#\\[feature.let_else#\\[cfg_attr\\(bootstrap, feature\\(let_else\\)#"
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).
const_generics: correctly deal with bound variables
removes the hack in `resolve` which was needed because we evaluated constants without caring about their bound variables.
Each commit should be fairly self-contained, even if they build on each other
r? `@jackh726`
Migrate another part of rustc_infer to session diagnostic
Probably will migrate another file before marking this one as ready-to-merge.
`@rustbot` label +A-translation
r? rust-lang/diagnostics
cc https://github.com/rust-lang/rust/issues/100717
Update `SessionDiagnostic::into_diagnostic` to take `Handler` instead of `ParseSess`
Suggested by the team in [this Zulip Topic](https://rust-lang.zulipchat.com/#narrow/stream/336883-i18n/topic/.23100717.20SessionDiagnostic.20on.20Handler).
`Handler` already has almost all the capabilities of `ParseSess` when it comes to diagnostic emission, in this migration we only needed to add the ability to access `source_map` from the emitter in order to get a `Snippet` and the `start_point`. Not sure if adding these two methods [`span_to_snippet_from_emitter` and `span_start_point_from_emitter`] is the best way to address this gap.
P.S. If this goes in the right direction, then we probably may want to move `SessionDiagnostic` to `rustc_errors` and rename it to `DiagnosticHandler` or something similar.
r? `@davidtwco`
r? `@compiler-errors`
Suggested by the team in this Zulip Topic https://rust-lang.zulipchat.com/#narrow/stream/336883-i18n/topic/.23100717.20SessionDiagnostic.20on.20Handler
Handler already has almost all the capabilities of ParseSess when it comes to diagnostic emission, in this migration we only needed to add the ability to access source_map from the emitter in order to get a Snippet and the start_point. Not sure if this is the best way to address this gap
This PR will fix some typos detected by [typos].
I only picked the ones I was sure were spelling errors to fix, mostly in
the comments.
[typos]: https://github.com/crate-ci/typos
Attempt to normalize `FnDef` signature in `InferCtxt::cmp`
Stashes a normalization callback in `InferCtxt` so that the signature we get from `tcx.fn_sig(..).subst(..)` in `InferCtxt::cmp` can be properly normalized, since we cannot expect for it to have normalized types since it comes straight from astconv.
This is kind of a hack, but I will say that `@jyn514` found the fact that we present unnormalized types to be very confusing in real life code, and I agree with that feeling. Though altogether I am still a bit unsure about whether this PR is worth the effort, so I'm open to alternatives and/or just closing it outright.
On the other hand, this isn't a ridiculously heavy implementation anyways -- it's less than a hundred lines of changes, and half of that is just miscellaneous cleanup.
This is stacked onto #100471 which is basically unrelated, and it can be rebased off of that when that lands or if needed.
---
The code:
```rust
trait Foo { type Bar; }
impl<T> Foo for T {
type Bar = i32;
}
fn foo<T>(_: <T as Foo>::Bar) {}
fn needs_i32_ref_fn(f: fn(&'static i32)) {}
fn main() {
needs_i32_ref_fn(foo::<()>);
}
```
Before:
```
= note: expected fn pointer `fn(&'static i32)`
found fn item `fn(<() as Foo>::Bar) {foo::<()>}`
```
After:
```
= note: expected fn pointer `fn(&'static i32)`
found fn item `fn(i32) {foo::<()>}`
```
Do not leak type variables from opaque type relation
The "root cause" is that we call `InferCtxt::resolve_vars_if_possible` (3d9dd681f5) on the types we get back in `TypeError::Sorts` since I added a call to it in `InferCtxt::same_type_modulo_infer`. However if this `TypeError` comes from a `InferCtxt::commit_if_ok`, then it may reference type variables that do not exist anymore, which is problematic.
We avoid this by substituting the `TypeError` with the types we had before being generalized while handling opaques.
This is kinda gross, and I feel like we can get the same issue from other places where we generalize type/const inference variables. Maybe not? I don't know.
Fixes#99914Fixes#99970Fixes#100463
Rollup of 9 pull requests
Successful merges:
- #95376 (Add `vec::Drain{,Filter}::keep_rest`)
- #100092 (Fall back when relating two opaques by substs in MIR typeck)
- #101019 (Suggest returning closure as `impl Fn`)
- #101022 (Erase late bound regions before comparing types in `suggest_dereferences`)
- #101101 (interpret: make read-pointer-as-bytes a CTFE-only error with extra information)
- #101123 (Remove `register_attr` feature)
- #101175 (Don't --bless in pre-push hook)
- #101176 (rustdoc: remove unused CSS selectors for `.table-display`)
- #101180 (Add another MaybeUninit array test with const)
Failed merges:
r? `@ghost`
`@rustbot` modify labels: rollup
Fall back when relating two opaques by substs in MIR typeck
This is certainly _one_ way to fix#100075. Not really confident it's the _best_ way to do it, though.
The root cause of this issue is that during MIR type-check, we end up trying to equate an opaque against the same opaque def-id but with different substs. Because of the way that we replace RPITs during (HIR) typeck with an inference variable, we don't end up emitting a type-checking error, so the delayed MIR bug causes an ICE.
See the `src/test/ui/impl-trait/issue-100075-2.rs` test below to make that clear -- in that example, we try to equate `{impl Sized} substs=[T]` and `{impl Sized} substs=[Option<T>]`, which causes an ICE. This new logic will instead cause us to infer `{impl Sized} substs=[Option<T>]` as the hidden type for `{impl Sized} substs=[T]`, which causes a proper error to be emitted later on when we check that an opaque isn't recursive.
I'm open to closing this in favor of something else. Ideally we'd fix this in typeck, but the thing we do to ensure backwards compatibility with weird RPIT cases makes that difficult. Also open to discussing this further.
Revert let_chains stabilization
This is the revert against master, the beta revert was already done in #100538.
Bumps the stage0 compiler which already has it reverted.
Rollup of 7 pull requests
Successful merges:
- #100898 (Do not report too many expr field candidates)
- #101056 (Add the syntax of references to their documentation summary.)
- #101106 (Rustdoc-Json: Retain Stripped Modules when they are imported, not when they have items)
- #101131 (CTFE: exposing pointers and calling extern fn is just impossible)
- #101141 (Simplify `get_trait_ref` fn used for `virtual_function_elimination`)
- #101146 (Various changes to logging of borrowck-related code)
- #101156 (Remove `Sync` requirement from lint pass objects)
Failed merges:
r? `@ghost`
`@rustbot` modify labels: rollup
Remove separate indexing of early-bound regions
~Based on https://github.com/rust-lang/rust/pull/99728.~
This PR copies some modifications from https://github.com/rust-lang/rust/pull/97839 around object lifetime defaults.
These modifications allow to stop counting generic parameters during lifetime resolution, and rely on the indexing given by `rustc_typeck::collect`.
InferCtxt tainted_by_errors_flag should be Option<ErrorGuaranteed>
Fixes#100321.
Use Cell<Option<ErrorGuaranteed>> to guarantee that we emit an error when that flag is set.
Make `same_type_modulo_infer` a proper `TypeRelation`
Specifically, this fixes#100690 because we no longer consider a `ReLateBound` and a `ReVar` to be equal. `ReVar` can only be equal to free regions or static.
Revert "Rollup merge of #97346 - JohnTitor:remove-back-compat-hacks, …
…r=oli-obk"
This reverts commit c703d11dcc, reversing
changes made to 64eb9ab869.
it didn't apply cleanly, so now it works the same for RPIT and for TAIT instead of just working for RPIT, but we should keep those in sync anyway. It also exposed a TAIT bug (see the feature gated test that now ICEs).
r? `@pnkfelix`
fixes#99536
remove `commit_unconditionally`
`commit_unconditionally` is a noop unless we somehow inspect the current state of our snapshot. The only thing which does that is the leak check which was only used in one place where `commit_if_ok` is probably at least as, or even more, correct.
r? rust-lang/types
make `PlaceholderConst` not store the type of the const
Currently the `Placeholder` variant on `ConstKind` is 28 bytes when with this PR its 8 bytes, i am not sure this is really useful at all rn since `Unevaluated` and `Value` variants are huge still but eventually it should be possible to get both down to 16 bytes 🤔. Mostly opening this to see if this change has any perf impact when done before it can make `ConstKind`/`ConstS` smaller
`codegen_fulfill_obligation` expect erased regions
it's a query, so by erasing regions before calling it, we get better caching.
This doesn't actually change anything as its already the status quo.
use `check_region_obligations_and_report_errors` to avoid ICEs
If we don't call `process_registered_region_obligations` before `resolve_regions_and_report_errors` then we'll ICE if we have any region obligations, and `check_region_obligations_and_report_errors` just does both of these for us in a nice convenient function.
Fixes#53475
r? types
Previously, `infcx.instantiate_canonical_*` maps the root universe in `canonical` into `ty::UniverseIndex::Root`, I think because it assumes it works with a fresh `infcx` but this is not true for the use cases in mir typeck. Now the root universe is mapped into `infcx.universe()`.
I catched this accidentally while reviewing the code. I'm not sure if this is the right fix or if it is really a bug!
Check that RPITs constrained by a recursive call in a closure are compatible
Fixes#99073
Adapts a similar visitor pattern to `find_opaque_ty_constraints` (that we use to check TAITs), but with some changes:
0. Only walk the "OnlyBody" children, instead of all items in the RPIT's defining scope
1. Only walk through the body's children if we found a constraining usage
2. Don't actually do any inference, just do a comparison and error if they're mismatched
----
r? `@oli-obk` -- you know all this impl-trait stuff best... is this the right approach? I can explain the underlying issue better if you'd like, in case that might reveal a better solution. Not sure if it's possible to gather up the closure's defining usages of the RPIT while borrowck'ing the outer function, that might be a better place to put this check...
Use full type name instead of just saying `impl Trait` in "captures lifetime" error
I think this is very useful, especially when there's >1 `impl Trait`, and it just means passing around a bit more info that we already have access to.
handle consts with param/infer in `const_eval_resolve` better
This PR addresses [this thread here](https://github.com/rust-lang/rust/pull/99449#discussion_r924141230). Was this the change you were looking for ``@lcnr?``
Interestingly, one test has begun to pass. Was that expected?
r? ``@lcnr``
Improve suggestions for returning binding
Fixes#99525
Also reworks the cause codes for match and if a bit, I think cleaning them up in a positive way.
We no longer need to call `could_remove_semicolon` in successful code, which might save a few cycles?
Remove the unused StableSet and StableMap types from rustc_data_structures.
The current implementation is not "stable" in the same sense that `HashStable` and `StableHasher` are stable, i.e. across compilation sessions. So, in my opinion, it's better to remove those types (which are basically unused anyway) than to give the wrong impression that these are safe for incr. comp.
I plan to provide new "stable" collection types soon that can be used to replace `FxHashMap` and `FxHashSet` in query results (see [draft](69d03ac7a7)). It's unsound that `HashMap` and `HashSet` implement `HashStable` (see https://github.com/rust-lang/rust/issues/98890 for a recent P-critical bug caused by this) -- so we should make some progress there.
Formalize defining_use_anchor
This tackles issue #57961
Introduces new enum called `DefiningAnchor` that replaces `Option<LocalDefId>` of `defining_use_anchor`. Now every use of it is explicit and exhaustively matched, catching errors like one in the linked issue. This is not a perfect fix but it's a step in the right direction.
r? `@oli-obk`
Do not allow typeck children items to constrain outer RPITs
Fixes#99073 in a simpler and more conservative way than #99079. Simply raise a mismatched types error if we try to constrain an RPIT in an item that isn't the RPIT's parent.
r? `@oli-obk`
Remove some usages of `guess_head_span`
No need to pass things through `guess_head_span` if they already point to the head span.
Only major change is that we point to the head span of `enum`s on some errors now, which I prefer.
r? `@cjgillot`
Move abstract const to middle
Moves AbstractConst (and all associated methods) to rustc middle for use in `rustc_infer`.
This allows for const resolution in infer to use abstract consts to walk consts and check if
they are resolvable.
This attempts to resolve the issue where `Foo<{ concrete const }, generic T>` is incorrectly marked as conflicting, and is independent from the other issue where nested abstract consts must be resolved.
r? `@lcnr`
Implement `for<>` lifetime binder for closures
This PR implements RFC 3216 ([TI](https://github.com/rust-lang/rust/issues/97362)) and allows code like the following:
```rust
let _f = for<'a, 'b> |a: &'a A, b: &'b B| -> &'b C { b.c(a) };
// ^^^^^^^^^^^--- new!
```
cc ``@Aaron1011`` ``@cjgillot``
don't use `commit_if_ok` during `higher_ranked_sub`
This snapshot doesn't really do anything useful for us, especially once we deal with placeholder outlive bounds during trait solving.
I guess that currently the idea is that `higher_ranked_sub` could cause a later `leak_check` to fail even if the combine operation isn't actually relevant. But really, using combine outside of snapshot and ignoring its result is wrong anyways, as it can constrain inference variables.
r? rust-lang/types
don't succeed `evaluate_obligation` query if new opaque types were registered
fixes#98608fixes#98604
The root cause of all this is that in type flag computation we entirely ignore nongeneric things like struct fields and the signature of function items. So if a flag had to be set for a struct if it is set for a field, that will only happen if the field is generic, as only the generic parameters are checked.
I now believe we cannot use type flags to handle opaque types. They seem like the wrong tool for this.
Instead, this PR replaces the previous logic by adding a new variant of `EvaluatedToOk`: `EvaluatedToOkModuloOpaqueTypes`, which says that there were some opaque types that got hidden types bound, but that binding may not have been legal (because we don't know if the opaque type was in its defining scope or not).
Make TAIT behave exactly like RPIT
fixes https://github.com/rust-lang/rust/issues/96552
This makes type-alias-impl-trait behave like return-position-impl-trait. Unfortunately it also causes some cases to stop compiling due to "needing type annotations" and makes panicking cause fallback for the hidden type to `()`.
All of these are addressable, but we should probably address them for RPIT and TAIT together
r? ``@lcnr``
Avoid some `&str` to `String` conversions with `MultiSpan::push_span_label`
This patch removes some`&str` to `String` conversions with `MultiSpan::push_span_label`.
Clean up arg mismatch diagnostic, generalize tuple wrap suggestion
This is based on top of #97542, so just look at the last commit which contains the relevant changes.
1. Remove `final_arg_types` which was one of the last places we were using raw (`usize`) indices instead of typed indices in the arg mismatch suggestion code.
2. Improve the tuple wrap suggestion, now we suggest things like `call(a, b, c, d)` -> `call(a, (b, c), d)` 😺
3. Folded in fix#98645
Reverse folder hierarchy
#91318 introduced a trait for infallible folders distinct from the fallible version. For some reason (completely unfathomable to me now that I look at it with fresh eyes), the infallible trait was a supertrait of the fallible one: that is, all fallible folders were required to also be infallible. Moreover the `Error` associated type was defined on the infallible trait! It's so absurd that it has me questioning whether I was entirely sane.
This trait reverses the hierarchy, so that the fallible trait is a supertrait of the infallible one: all infallible folders are required to also be fallible (which is a trivial blanket implementation). This of course makes much more sense! It also enables the `Error` associated type to sit on the fallible trait, where it sensibly belongs.
There is one downside however: folders expose a `tcx` accessor method. Since the blanket fallible implementation for infallible folders only has access to a generic `F: TypeFolder`, we need that trait to expose such an accessor to which we can delegate. Alternatively it's possible to extract that accessor into a separate `HasTcx` trait (or similar) that would then be a supertrait of both the fallible and infallible folder traits: this would ensure that there's only one unambiguous `tcx` method, at the cost of a little additional boilerplate. If desired, I can submit that as a separate PR.
r? ````@jackh726````
fix universes in the NLL type tests
In the NLL code, we were not accommodating universes in the
`type_test` logic.
Fixes#98095.
r? `@compiler-errors`
This breaks some tests, however, so the purpose of this branch is more explanatory and perhaps to do a crater run.
Point at return expression for RPIT-related error
Certainly this needs some diagnostic refining, but I wanted to show that it was possible first and foremost. Not sure if this is the right approach. Open to feedback.
Fixes#80583
lub: don't bail out due to empty binders
allows for the following to compile. The equivalent code using `struct Wrapper<'upper>(fn(&'upper ());` already compiles on stable.
```rust
let _: fn(&'upper ()) = match v {
true => lt_in_fn::<'a>(),
false => lt_in_fn::<'b>(),
};
```
see https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=7034a677190110941223cafac6632f70 for a complete example
r? ```@rust-lang/types```
#91318 introduced a trait for infallible folders distinct from the fallible version. For some reason (completely unfathomable to me now that I look at it with fresh eyes), the infallible trait was a supertrait of the fallible one: that is, all fallible folders were required to also be infallible. Moreover the `Error` associated type was defined on the infallible trait! It's so absurd that it has me questioning whether I was entirely sane.
This trait reverses the hierarchy, so that the fallible trait is a supertrait of the infallible one: all infallible folders are required to also be fallible (which is a trivial blanket implementation). This of course makes much more sense! It also enables the `Error` associated type to sit on the fallible trait, where it sensibly belongs.
There is one downside however: folders expose a `tcx` accessor method. Since the blanket fallible implementation for infallible folders only has access to a generic `F: TypeFolder`, we need that trait to expose such an accessor to which we can delegate. Alternatively it's possible to extract that accessor into a separate `HasTcx` trait (or similar) that would then be a supertrait of both the fallible and infallible folder traits: this would ensure that there's only one unambiguous `tcx` method, at the cost of a little additional boilerplate. If desired, I can submit that as a separate PR.
r? @jackh726
The code now accepts `Binder<OutlivesPredicate>`
instead of just `OutlivesPredicate` and thus exercises
the new, generalized `IfEqBound` codepaths. Note though
that we never *produce* Binder<OutlivesPredicate>, so we
are only testing a subset of those codepaths that excludes
actual higher-ranked outlives bounds.
Make `ExprKind::Closure` a struct variant.
Simple refactor since we both need it to introduce additional fields in `ExprKind::Closure`.
r? ``@Aaron1011``
Rename the `ConstS::val` field as `kind`.
And likewise for the `Const::val` method.
Because its type is called `ConstKind`. Also `val` is a confusing name
because `ConstKind` is an enum with seven variants, one of which is
called `Value`. Also, this gives consistency with `TyS` and `PredicateS`
which have `kind` fields.
The commit also renames a few `Const` variables from `val` to `c`, to
avoid confusion with the `ConstKind::Value` variant.
r? `@BoxyUwU`
Remove RegionckMode in favor of calling new skip_region_resolution
Simple cleanup. We can skip a bunch of stuff for places where NLL does the region checking, so skip earlier.
r? rust-lang/types
And likewise for the `Const::val` method.
Because its type is called `ConstKind`. Also `val` is a confusing name
because `ConstKind` is an enum with seven variants, one of which is
called `Value`. Also, this gives consistency with `TyS` and `PredicateS`
which have `kind` fields.
The commit also renames a few `Const` variables from `val` to `c`, to
avoid confusion with the `ConstKind::Value` variant.
Mention `infer::Trace` methods on `infer::At` methods' docs
I missed that you could do `infcx.at(...).trace(...).eq(a, b)` when `a` and `b` dont implement `ToTrace` but does implement `Relate` these docs would have helped see that 😅
This commit makes type folding more like the way chalk does it.
Currently, `TypeFoldable` has `fold_with` and `super_fold_with` methods.
- `fold_with` is the standard entry point, and defaults to calling
`super_fold_with`.
- `super_fold_with` does the actual work of traversing a type.
- For a few types of interest (`Ty`, `Region`, etc.) `fold_with` instead
calls into a `TypeFolder`, which can then call back into
`super_fold_with`.
With the new approach, `TypeFoldable` has `fold_with` and
`TypeSuperFoldable` has `super_fold_with`.
- `fold_with` is still the standard entry point, *and* it does the
actual work of traversing a type, for all types except types of
interest.
- `super_fold_with` is only implemented for the types of interest.
Benefits of the new model.
- I find it easier to understand. The distinction between types of
interest and other types is clearer, and `super_fold_with` doesn't
exist for most types.
- With the current model is easy to get confused and implement a
`super_fold_with` method that should be left defaulted. (Some of the
precursor commits fixed such cases.)
- With the current model it's easy to call `super_fold_with` within
`TypeFolder` impls where `fold_with` should be called. The new
approach makes this mistake impossible, and this commit fixes a number
of such cases.
- It's potentially faster, because it avoids the `fold_with` ->
`super_fold_with` call in all cases except types of interest. A lot of
the time the compile would inline those away, but not necessarily
always.
Remove migrate borrowck mode
Closes#58781Closes#43234
# Stabilization proposal
This PR proposes the stabilization of `#![feature(nll)]` and the removal of `-Z borrowck`. Current borrow checking behavior of item bodies is currently done by first infering regions *lexically* and reporting any errors during HIR type checking. If there *are* any errors, then MIR borrowck (NLL) never occurs. If there *aren't* any errors, then MIR borrowck happens and any errors there would be reported. This PR removes the lexical region check of item bodies entirely and only uses MIR borrowck. Because MIR borrowck could never *not* be run for a compiled program, this should not break any programs. It does, however, change diagnostics significantly and allows a slightly larger set of programs to compile.
Tracking issue: #43234
RFC: https://github.com/rust-lang/rfcs/blob/master/text/2094-nll.md
Version: 1.63 (2022-06-30 => beta, 2022-08-11 => stable).
## Motivation
Over time, the Rust borrow checker has become "smarter" and thus allowed more programs to compile. There have been three different implementations: AST borrowck, MIR borrowck, and polonius (well, in progress). Additionally, there is the "lexical region resolver", which (roughly) solves the constraints generated through HIR typeck. It is not a full borrow checker, but does emit some errors.
The AST borrowck was the original implementation of the borrow checker and was part of the initially stabilized Rust 1.0. In mid 2017, work began to implement the current MIR borrow checker and that effort ompleted by the end of 2017, for the most part. During 2018, efforts were made to migrate away from the AST borrow checker to the MIR borrow checker - eventually culminating into "migrate" mode - where HIR typeck with lexical region resolving following by MIR borrow checking - being active by default in the 2018 edition.
In early 2019, migrate mode was turned on by default in the 2015 edition as well, but with MIR borrowck errors emitted as warnings. By late 2019, these warnings were upgraded to full errors. This was followed by the complete removal of the AST borrow checker.
In the period since, various errors emitted by the MIR borrow checker have been improved to the point that they are mostly the same or better than those emitted by the lexical region resolver.
While there do remain some degradations in errors (tracked under the [NLL-diagnostics tag](https://github.com/rust-lang/rust/issues?q=is%3Aopen+is%3Aissue+label%3ANLL-diagnostics), those are sufficiently small and rare enough that increased flexibility of MIR borrow check-only is now a worthwhile tradeoff.
## What is stabilized
As said previously, this does not fundamentally change the landscape of accepted programs. However, there are a [few](https://github.com/rust-lang/rust/issues?q=is%3Aopen+is%3Aissue+label%3ANLL-fixed-by-NLL) cases where programs can compile under `feature(nll)`, but not otherwise.
There are two notable patterns that are "fixed" by this stabilization. First, the `scoped_threads` feature, which is a continutation of a pre-1.0 API, can sometimes emit a [weird lifetime error](https://github.com/rust-lang/rust/issues/95527) without NLL. Second, actually seen in the standard library. In the `Extend` impl for `HashMap`, there is an implied bound of `K: 'a` that is available with NLL on but not without - this is utilized in the impl.
As mentioned before, there are a large number of diagnostic differences. Most of them are better, but some are worse. None are serious or happen often enough to need to block this PR. The biggest change is the loss of error code for a number of lifetime errors in favor of more general "lifetime may not live long enough" error. While this may *seem* bad, the former error codes were just attempts to somewhat-arbitrarily bin together lifetime errors of the same type; however, on paper, they end up being roughly the same with roughly the same kinds of solutions.
## What isn't stabilized
This PR does not completely remove the lexical region resolver. In the future, it may be possible to remove that (while still keeping HIR typeck) or to remove it together with HIR typeck.
## Tests
Many test outputs get updated by this PR. However, there are number of tests specifically geared towards NLL under `src/test/ui/nll`
## History
* On 2017-07-14, [tracking issue opened](https://github.com/rust-lang/rust/issues/43234)
* On 2017-07-20, [initial empty MIR pass added](https://github.com/rust-lang/rust/pull/43271)
* On 2017-08-29, [RFC opened](https://github.com/rust-lang/rfcs/pull/2094)
* On 2017-11-16, [Integrate MIR type-checker with NLL](https://github.com/rust-lang/rust/pull/45825)
* On 2017-12-20, [NLL feature complete](https://github.com/rust-lang/rust/pull/46862)
* On 2018-07-07, [Don't run AST borrowck on mir mode](https://github.com/rust-lang/rust/pull/52083)
* On 2018-07-27, [Add migrate mode](https://github.com/rust-lang/rust/pull/52681)
* On 2019-04-22, [Enable migrate mode on 2015 edition](https://github.com/rust-lang/rust/pull/59114)
* On 2019-08-26, [Don't downgrade errors on 2015 edition](https://github.com/rust-lang/rust/pull/64221)
* On 2019-08-27, [Remove AST borrowck](https://github.com/rust-lang/rust/pull/64790)
rewrite error handling for unresolved inference vars
Pretty much completely rewrites `fn emit_inference_failure_err`.
This new setup should hopefully be easier to extend and is already a lot better when looking for generic arguments.
Because this is a rewrite there are still some parts which are lacking, these are tracked in #94483 and will be fixed in later PRs.
r? `@estebank` `@petrochenkov`
Diagnose anonymous lifetimes errors more uniformly between async and regular fns
Async fns and regular fns are desugared differently. For the former, we create a generic parameter at HIR level. For the latter, we just create an anonymous region for typeck.
I plan to migrate regular fns to the async fn desugaring.
Before that, this PR attempts to merge the diagnostics for both cases.
r? ```@estebank```
Finish bumping stage0
It looks like the last time had left some remaining cfg's -- which made me think
that the stage0 bump was actually successful. This brings us to a released 1.62
beta though.
This now brings us to cfg-clean, with the exception of check-cfg-features in bootstrap;
I'd prefer to leave that for a separate PR at this time since it's likely to be more tricky.
cc https://github.com/rust-lang/rust/pull/97147#issuecomment-1132845061
r? `@pietroalbini`
It looks like the last time had left some remaining cfg's -- which made me think
that the stage0 bump was actually successful. This brings us to a released 1.62
beta though.
Lifetime variance fixes for rustc
#97287 migrates rustc to a `Ty` type that is invariant over its lifetime `'tcx`, so I need to fix a bunch of places that assume that `Ty<'a>` and `Ty<'b>` can be unified by shortening both to some common lifetime.
This is doable, since many lifetimes are already `'tcx`, so all this PR does is be a bit more explicit that elided lifetimes are actually `'tcx`.
Split out from #97287 so the compiler team can review independently.
