Lazy type-alias-impl-trait take two
### user visible change 1: RPIT inference from recursive call sites
Lazy TAIT has an insta-stable change. The following snippet now compiles, because opaque types can now have their hidden type set from wherever the opaque type is mentioned.
```rust
fn bar(b: bool) -> impl std::fmt::Debug {
if b {
return 42
}
let x: u32 = bar(false); // this errors on stable
99
}
```
The return type of `bar` stays opaque, you can't do `bar(false) + 42`, you need to actually mention the hidden type.
### user visible change 2: divergence between RPIT and TAIT in return statements
Note that `return` statements and the trailing return expression are special with RPIT (but not TAIT). So
```rust
#![feature(type_alias_impl_trait)]
type Foo = impl std::fmt::Debug;
fn foo(b: bool) -> Foo {
if b {
return vec![42];
}
std::iter::empty().collect() //~ ERROR `Foo` cannot be built from an iterator
}
fn bar(b: bool) -> impl std::fmt::Debug {
if b {
return vec![42]
}
std::iter::empty().collect() // Works, magic (accidentally stabilized, not intended)
}
```
But when we are working with the return value of a recursive call, the behavior of RPIT and TAIT is the same:
```rust
type Foo = impl std::fmt::Debug;
fn foo(b: bool) -> Foo {
if b {
return vec![];
}
let mut x = foo(false);
x = std::iter::empty().collect(); //~ ERROR `Foo` cannot be built from an iterator
vec![]
}
fn bar(b: bool) -> impl std::fmt::Debug {
if b {
return vec![];
}
let mut x = bar(false);
x = std::iter::empty().collect(); //~ ERROR `impl Debug` cannot be built from an iterator
vec![]
}
```
### user visible change 3: TAIT does not merge types across branches
In contrast to RPIT, TAIT does not merge types across branches, so the following does not compile.
```rust
type Foo = impl std::fmt::Debug;
fn foo(b: bool) -> Foo {
if b {
vec![42_i32]
} else {
std::iter::empty().collect()
//~^ ERROR `Foo` cannot be built from an iterator over elements of type `_`
}
}
```
It is easy to support, but we should make an explicit decision to include the additional complexity in the implementation (it's not much, see a721052457cf513487fb4266e3ade65c29b272d2 which needs to be reverted to enable this).
### PR formalities
previous attempt: #92007
This PR also includes #92306 and #93783, as they were reverted along with #92007 in #93893fixes#93411fixes#88236fixes#89312fixes#87340fixes#86800fixes#86719fixes#84073fixes#83919fixes#82139fixes#77987fixes#74282fixes#67830fixes#62742fixes#54895
This allows to compute the `BodyOwnerKind` from `DefKind` only, and
removes a direct dependency of some MIR queries onto HIR.
As a side effect, it also simplifies metadata, since we don't need 4
flavours of `EntryKind::*Static` any more.
Remove `Session::one_time_diagnostic`
This is untracked mutable state, which modified the behaviour of queries.
It was used for 2 things: some full-blown errors, but mostly for lint declaration notes ("the lint level is defined here" notes).
It is replaced by the diagnostic deduplication infra which already exists in the diagnostic emitter.
A new diagnostic level `OnceNote` is introduced specifically for lint notes, to deduplicate subdiagnostics.
As a drive-by, diagnostic emission takes a `&mut` to allow dropping the `SubDiagnostic`s.
Remove redundant code from copy-suggestions
Follow up to #94375, just remove some code that is not necessary anymore. This may make the perf of such suggestions a little bit worse, but I don't think this is significant.
r? `@estebank`
There are a few places were we have to construct it, though, and a few
places that are more invasive to change. To do this, we create a
constructor with a long obvious name.
This commit makes `AdtDef` use `Interned`. Much the commit is tedious
changes to introduce getter functions. The interesting changes are in
`compiler/rustc_middle/src/ty/adt.rs`.
add `#[rustc_pass_by_value]` to more types
the only interesting changes here should be to `TransitiveRelation`, but I believe to be highly unlikely that we will ever use a non `Copy` type with this type.
Adt copy suggestions
Previously we've only suggested adding `Copy` bounds when the type being moved/copied is a type parameter (generic). With this PR we also suggest adding bounds when a type
- Can be copy
- All predicates that need to be satisfied for that are based on type params
i.e. we will suggest `T: Copy` for `Option<T>`, but won't suggest anything for `Option<String>`.
An example:
```rust
fn duplicate<T>(t: Option<T>) -> (Option<T>, Option<T>) {
(t, t)
}
```
New error (current compiler doesn't provide `help`:):
```text
error[E0382]: use of moved value: `t`
--> t.rs:2:9
|
1 | fn duplicate<T>(t: Option<T>) -> (Option<T>, Option<T>) {
| - move occurs because `t` has type `Option<T>`, which does not implement the `Copy` trait
2 | (t, t)
| - ^ value used here after move
| |
| value moved here
|
help: consider restricting type parameter `T`
|
1 | fn duplicate<T: Copy>(t: Option<T>) -> (Option<T>, Option<T>) {
| ++++++
```
Fixes#93623
r? ``````````@estebank``````````
``````````@rustbot`````````` label +A-diagnostics +A-suggestion-diagnostics +C-enhancement
----
I'm not at all sure if this is the right implementation for this kind of suggestion, but it seems to work :')
Previously we've only suggested adding `Copy` bounds when the type being
moved/copied is a type parameter (generic). With this commit we also
suggest adding bounds when a type
- Can be copy
- All predicates that need to be satisfied for that are based on type
params
i.e. we will suggest `T: Copy` for `Option<T>`, but won't suggest
anything for `Option<String>`.
Future work: it would be nice to also suggest adding `.clone()` calls
Populate liveness facts when calling `get_body_with_borrowck_facts` without `-Z polonius`
For a new feature of [Flowistry](https://github.com/willcrichton/flowistry), a static-analysis tool, we need to obtain a `mir::Body`'s liveness facts using `get_body_with_borrowck_facts` (added in #86977). We'd like to do this without passing `-Z polonius` as a compiler arg to avoid borrow checking the entire crate.
Support for doing this was added in #88983, but the Polonius input facts used for liveness analysis are empty. This happens because the liveness input facts are populated in `liveness::generate` depending only on the value of `AllFacts::enabled` (which is toggled via compiler args).
This PR propagates the [`use_polonius`](8b09ba6a5d/compiler/rustc_borrowck/src/nll.rs (L168)) flag to `liveness::generate` to support populating liveness facts without requiring the `-Z polonius` flag.
This fix is somewhat patchy - if it'd be better to add more widely-accessible state (like `AllFacts::enabled`) I'd be open to ideas!
rustc_errors: let `DiagnosticBuilder::emit` return a "guarantee of emission".
That is, `DiagnosticBuilder` is now generic over the return type of `.emit()`, so we'll now have:
* `DiagnosticBuilder<ErrorReported>` for error (incl. fatal/bug) diagnostics
* can only be created via a `const L: Level`-generic constructor, that limits allowed variants via a `where` clause, so not even `rustc_errors` can accidentally bypass this limitation
* asserts `diagnostic.is_error()` on emission, just in case the construction restriction was bypassed (e.g. by replacing the whole `Diagnostic` inside `DiagnosticBuilder`)
* `.emit()` returns `ErrorReported`, as a "proof" token that `.emit()` was called
(though note that this isn't a real guarantee until after completing the work on
#69426)
* `DiagnosticBuilder<()>` for everything else (warnings, notes, etc.)
* can also be obtained from other `DiagnosticBuilder`s by calling `.forget_guarantee()`
This PR is a companion to other ongoing work, namely:
* #69426
and it's ongoing implementation:
#93222
the API changes in this PR are needed to get statically-checked "only errors produce `ErrorReported` from `.emit()`", but doesn't itself provide any really strong guarantees without those other `ErrorReported` changes
* #93244
would make the choices of API changes (esp. naming) in this PR fit better overall
In order to be able to let `.emit()` return anything trustable, several changes had to be made:
* `Diagnostic`'s `level` field is now private to `rustc_errors`, to disallow arbitrary "downgrade"s from "some kind of error" to "warning" (or anything else that doesn't cause compilation to fail)
* it's still possible to replace the whole `Diagnostic` inside the `DiagnosticBuilder`, sadly, that's harder to fix, but it's unlikely enough that we can paper over it with asserts on `.emit()`
* `.cancel()` now consumes `DiagnosticBuilder`, preventing `.emit()` calls on a cancelled diagnostic
* it's also now done internally, through `DiagnosticBuilder`-private state, instead of having a `Level::Cancelled` variant that can be read (or worse, written) by the user
* this removes a hazard of calling `.cancel()` on an error then continuing to attach details to it, and even expect to be able to `.emit()` it
* warnings were switched to *only* `can_emit_warnings` on emission (instead of pre-cancelling early)
* `struct_dummy` was removed (as it relied on a pre-`Cancelled` `Diagnostic`)
* since `.emit()` doesn't consume the `DiagnosticBuilder` <sub>(I tried and gave up, it's much more work than this PR)</sub>,
we have to make `.emit()` idempotent wrt the guarantees it returns
* thankfully, `err.emit(); err.emit();` can return `ErrorReported` both times, as the second `.emit()` call has no side-effects *only* because the first one did do the appropriate emission
* `&mut Diagnostic` is now used in a lot of function signatures, which used to take `&mut DiagnosticBuilder` (in the interest of not having to make those functions generic)
* the APIs were already mostly identical, allowing for low-effort porting to this new setup
* only some of the suggestion methods needed some rework, to have the extra `DiagnosticBuilder` functionality on the `Diagnostic` methods themselves (that change is also present in #93259)
* `.emit()`/`.cancel()` aren't available, but IMO calling them from an "error decorator/annotator" function isn't a good practice, and can lead to strange behavior (from the caller's perspective)
* `.downgrade_to_delayed_bug()` was added, letting you convert any `.is_error()` diagnostic into a `delay_span_bug` one (which works because in both cases the guarantees available are the same)
This PR should ideally be reviewed commit-by-commit, since there is a lot of fallout in each.
r? `@estebank` cc `@Manishearth` `@nikomatsakis` `@mark-i-m`
Always format to internal String in FmtPrinter
This avoids monomorphizing for different parameters, decreasing generic code
instantiated downstream from rustc_middle -- locally seeing 7% unoptimized LLVM IR
line wins on rustc_borrowck, for example.
We likely can't/shouldn't get rid of the Result-ness on most functions, though some
further cleanup avoiding fmt::Error where we now know it won't occur may be possible,
though somewhat painful -- fmt::Write is a pretty annoying API to work with in practice
when you're trying to use it infallibly.
Introduce `ChunkedBitSet` and use it for some dataflow analyses.
This reduces peak memory usage significantly for some programs with very
large functions.
r? `@ghost`
This reduces peak memory usage significantly for some programs with very
large functions, such as:
- `keccak`, `unicode_normalization`, and `match-stress-enum`, from
the `rustc-perf` benchmark suite;
- `http-0.2.6` from crates.io.
The new type is used in the analyses where the bitsets can get huge
(e.g. 10s of thousands of bits): `MaybeInitializedPlaces`,
`MaybeUninitializedPlaces`, and `EverInitializedPlaces`.
Some refactoring was required in `rustc_mir_dataflow`. All existing
analysis domains are either `BitSet` or a trivial wrapper around
`BitSet`, and access in a few places is done via `Borrow<BitSet>` or
`BorrowMut<BitSet>`. Now that some of these domains are `ClusterBitSet`,
that no longer works. So this commit replaces the `Borrow`/`BorrowMut`
usage with a new trait `BitSetExt` containing the needed bitset
operations. The impls just forward these to the underlying bitset type.
This required fiddling with trait bounds in a few places.
The commit also:
- Moves `static_assert_size` from `rustc_data_structures` to
`rustc_index` so it can be used in the latter; the former now
re-exports it so existing users are unaffected.
- Factors out some common "clear excess bits in the final word"
functionality in `bit_set.rs`.
- Uses `fill` in a few places instead of loops.
Adopt let else in more places
Continuation of #89933, #91018, #91481, #93046, #93590, #94011.
I have extended my clippy lint to also recognize tuple passing and match statements. The diff caused by fixing it is way above 1 thousand lines. Thus, I split it up into multiple pull requests to make reviewing easier. This is the biggest of these PRs and handles the changes outside of rustdoc, rustc_typeck, rustc_const_eval, rustc_trait_selection, which were handled in PRs #94139, #94142, #94143, #94144.
Specifically, rename the `Const` struct as `ConstS` and re-introduce `Const` as
this:
```
pub struct Const<'tcx>(&'tcx Interned<ConstS>);
```
This now matches `Ty` and `Predicate` more closely, including using
pointer-based `eq` and `hash`.
Notable changes:
- `mk_const` now takes a `ConstS`.
- `Const` was copy, despite being 48 bytes. Now `ConstS` is not, so need a
we need separate arena for it, because we can't use the `Dropless` one any
more.
- Many `&'tcx Const<'tcx>`/`&Const<'tcx>` to `Const<'tcx>` changes
- Many `ct.ty` to `ct.ty()` and `ct.val` to `ct.val()` changes.
- Lots of tedious sigil fiddling.
The variant names are exported, so we can use them directly (possibly
with a `ty::` qualifier). Lots of places already do this, this commit
just increases consistency.
Specifically, change `Region` from this:
```
pub type Region<'tcx> = &'tcx RegionKind;
```
to this:
```
pub struct Region<'tcx>(&'tcx Interned<RegionKind>);
```
This now matches `Ty` and `Predicate` more closely.
Things to note
- Regions have always been interned, but we haven't been using pointer-based
`Eq` and `Hash`. This is now happening.
- I chose to impl `Deref` for `Region` because it makes pattern matching a lot
nicer, and `Region` can be viewed as just a smart wrapper for `RegionKind`.
- Various methods are moved from `RegionKind` to `Region`.
- There is a lot of tedious sigil changes.
- A couple of types like `HighlightBuilder`, `RegionHighlightMode` now have a
`'tcx` lifetime because they hold a `Ty<'tcx>`, so they can call `mk_region`.
- A couple of test outputs change slightly, I'm not sure why, but the new
outputs are a little better.
Specifically, change `Ty` from this:
```
pub type Ty<'tcx> = &'tcx TyS<'tcx>;
```
to this
```
pub struct Ty<'tcx>(Interned<'tcx, TyS<'tcx>>);
```
There are two benefits to this.
- It's now a first class type, so we can define methods on it. This
means we can move a lot of methods away from `TyS`, leaving `TyS` as a
barely-used type, which is appropriate given that it's not meant to
be used directly.
- The uniqueness requirement is now explicit, via the `Interned` type.
E.g. the pointer-based `Eq` and `Hash` comes from `Interned`, rather
than via `TyS`, which wasn't obvious at all.
Much of this commit is boring churn. The interesting changes are in
these files:
- compiler/rustc_middle/src/arena.rs
- compiler/rustc_middle/src/mir/visit.rs
- compiler/rustc_middle/src/ty/context.rs
- compiler/rustc_middle/src/ty/mod.rs
Specifically:
- Most mentions of `TyS` are removed. It's very much a dumb struct now;
`Ty` has all the smarts.
- `TyS` now has `crate` visibility instead of `pub`.
- `TyS::make_for_test` is removed in favour of the static `BOOL_TY`,
which just works better with the new structure.
- The `Eq`/`Ord`/`Hash` impls are removed from `TyS`. `Interned`s impls
of `Eq`/`Hash` now suffice. `Ord` is now partly on `Interned`
(pointer-based, for the `Equal` case) and partly on `TyS`
(contents-based, for the other cases).
- There are many tedious sigil adjustments, i.e. adding or removing `*`
or `&`. They seem to be unavoidable.
Inherit lifetimes for async fn instead of duplicating them.
The current desugaring of `async fn foo<'a>(&usize) -> &u8` is equivalent to
```rust
fn foo<'a, '0>(&'0 usize) -> foo<'static, 'static>::Opaque<'a, '0, '_>;
type foo<'_a, '_0>::Opaque<'a, '0, '1> = impl Future<Output = &'1 u8>;
```
following the RPIT model.
Duplicating all the inherited lifetime parameters and setting the inherited version to `'static` makes lowering more complex and causes issues like #61949. This PR removes the duplication of inherited lifetimes to directly use
```rust
fn foo<'a, '0>(&'0 usize) -> foo<'a, '0>::Opaque<'_>;
type foo<'a, '0>::Opaque<'1> = impl Future<Output = &'1 u8>;
```
following the TAIT model.
Fixes https://github.com/rust-lang/rust/issues/61949
Currently, any higher-ranked region errors involving opaque types
fall back to a generic "higher-ranked subtype error" message when
run under NLL. This PR adds better error message handling for this
case, giving us the same kinds of error messages that we currently
get without NLL:
```
error: implementation of `MyTrait` is not general enough
--> $DIR/opaque-hrtb.rs:12:13
|
LL | fn foo() -> impl for<'a> MyTrait<&'a str> {
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ implementation of `MyTrait` is not general enough
|
= note: `impl MyTrait<&'2 str>` must implement `MyTrait<&'1 str>`, for any lifetime `'1`...
= note: ...but it actually implements `MyTrait<&'2 str>`, for some specific lifetime `'2`
error: aborting due to previous error
```
To accomplish this, several different refactoring needed to be made:
* We now have a dedicated `InstantiateOpaqueType` struct which
implements `TypeOp`. This is used to invoke `instantiate_opaque_types`
during MIR type checking.
* `TypeOp` is refactored to pass around a `MirBorrowckCtxt`, which is
needed to report opaque type region errors.
* We no longer assume that all `TypeOp`s correspond to canonicalized
queries. This allows us to properly handle opaque type instantiation
(which does not occur in a query) as a `TypeOp`.
A new `ErrorInfo` associated type is used to determine what
additional information is used during higher-ranked region error
handling.
* The body of `try_extract_error_from_fulfill_cx`
has been moved out to a new function `try_extract_error_from_region_constraints`.
This allows us to re-use the same error reporting code between
canonicalized queries (which can extract region constraints directly
from a fresh `InferCtxt`) and opaque type handling (which needs to take
region constraints from the pre-existing `InferCtxt` that we use
throughout MIR borrow checking).
Lazy type-alias-impl-trait
Previously opaque types were processed by
1. replacing all mentions of them with inference variables
2. memorizing these inference variables in a side-table
3. at the end of typeck, resolve the inference variables in the side table and use the resolved type as the hidden type of the opaque type
This worked okayish for `impl Trait` in return position, but required lots of roundabout type inference hacks and processing.
This PR instead stops this process of replacing opaque types with inference variables, and just keeps the opaque types around.
Whenever an opaque type `O` is compared with another type `T`, we make the comparison succeed and record `T` as the hidden type. If `O` is compared to `U` while there is a recorded hidden type for it, we grab the recorded type (`T`) and compare that against `U`. This makes implementing
* https://github.com/rust-lang/rfcs/pull/2515
much simpler (previous attempts on the inference based scheme were very prone to ICEs and general misbehaviour that was not explainable except by random implementation defined oddities).
r? `@nikomatsakis`
fixes#93411fixes#88236
[borrowck] Fix help on mutating &self in async fns
Previously, when rustc was provided an async function that tried to
mutate through a shared reference to an implicit self (as shown in the
ui test), rustc would suggest modifying the parameter signature
to `&mut` + the fully qualified name of the ty (in the case of the repro
`S`). If a user modified their code to match the suggestion, the
compiler would not accept it.
This commit modifies the suggestion so that when rustc is provided the
ui test that is also attached in this commit, it suggests (correctly)
`&mut self`. We try to be careful about distinguishing between implicit
and explicit self annotations, since the latter seem to be handled
correctly already.
This is my first PR here so I'm pretty sure I probably missed something/could use better terminology. I also didn't try to make the match exhaustive since implicit self is the only real special case that I need to handle (that I'm aware of), and I'm pretty sure there's a cleaner way to do this so any advice would be greatly appreciated! (I'm also not terribly confident about how I wrote the ui tests)
here is your cc as requested `@compiler-errors`
This is an attempt to fix#93093
by using an opaque type obligation to bubble up comparisons between opaque types and other types
Also uses proper obligation causes so that the body id works, because out of some reason nll uses body ids for logic instead of just diagnostics.
Store a `Symbol` instead of an `Ident` in `AssocItem`
This is the same idea as #92533, but for `AssocItem` instead
of `VariantDef`/`FieldDef`.
With this change, we no longer have any uses of
`#[stable_hasher(project(...))]`
Remove ordering traits from `OutlivesConstraint`
In two cases where this ordering was used, I've replaced the sorting to use a key that does not rely on `DefId` being `Ord`. This is part of #90317. If I understand correctly, whether this is correct depends on whether the `RegionVid`s are tracked during incremental compilation. But I might be mistaken in this approach. cc `@cjgillot`
Previously, when rustc was provided an async function that tried to
mutate through a shared reference to an implicit self (as shown in the
ui test), rustc would suggest modifying the parameter signature
to `&mut` + the fully qualified name of the ty (in the case of the repro
`S`). If a user modified their code to match the suggestion, the
compiler would not accept it.
This commit modifies the suggestion so that when rustc is provided the
ui test that is also attached in this commit, it suggests (correctly)
`&mut self`. We try to be careful about distinguishing between implicit
and explicit self annotations, since the latter seem to be handled
correctly already.
Fixesrust-lang/rust#93093
Ensure that early-bound function lifetimes are always 'local'
During borrowchecking, we treat any free (early-bound) regions on
the 'defining type' as `RegionClassification::External`. According
to the doc comments, we should only have 'external' regions when
checking a closure/generator.
However, a plain function can also have some if its regions
be considered 'early bound' - this occurs when the region is
constrained by an argument, appears in a `where` clause, or
in an opaque type. This was causing us to incorrectly mark these
regions as 'external', which caused some diagnostic code
to act as if we were referring to a 'parent' region from inside
a closure.
This PR marks all instantiated region variables as 'local'
when we're borrow-checking something other than a
closure/generator/inline-const.
Update some rustc dependencies to deduplicate them
This PR updates `rand` and `itertools` in rustc (not the whole workspace) in order to deduplicate them (and hopefully slightly improve compile times).
~~Currently, `object` is still duplicated, but https://github.com/rust-lang/thorin/pull/15 and updating `thorin` in the future will remove the use of version 0.27.~~ Update: Thorin 0.2 has now been released, and this PR updates `rustc_codegen_ssa` to use it and deduplicate the `object` crate.
There's a final tiny rustc dependency, `cfg-if`, which will be left: as both versions 0.1.x and 1.0 looked to be heavily depended on, they will require a few cascading updates to be removed.
This is the same idea as #92533, but for `AssocItem` instead
of `VariantDef`/`FieldDef`.
With this change, we no longer have any uses of
`#[stable_hasher(project(...))]`
In two cases where this ordering was used, I've replaced the sorting
to use a key that does not include DefId. I'm not sure this is correct
in terms of our goals from #90317, or otherwise.
ProjectionPredicate should be able to handle both associated types and consts so this adds the
first step of that. It mainly just pipes types all the way down, not entirely sure how to handle
consts, but hopefully that'll come with time.
Remove deprecated LLVM-style inline assembly
The `llvm_asm!` was deprecated back in #87590 1.56.0, with intention to remove
it once `asm!` was stabilized, which already happened in #91728 1.59.0. Now it
is time to remove `llvm_asm!` to avoid continued maintenance cost.
Closes#70173.
Closes#92794.
Closes#87612.
Closes#82065.
cc `@rust-lang/wg-inline-asm`
r? `@Amanieu`
Add `#[track_caller]` to `mirbug`
When a "'no errors encountered even though `delay_span_bug` issued" error results from the `mirbug` function, the file location information points to the `mirbug` function itself, rather than its caller. This doesn't make sense, since the caller is the real source of the bug. Adding `#[track_caller]` will produce diagnostics that are more useful to anyone fixing the ICE.
Closure capture cleanup & refactor
Follow up of #89648
Each commit is self-contained and the rationale/changes are documented in the commit message, so it's advisable to review commit by commit.
The code is significantly cleaner (at least IMO), but that could have some perf implication, so I'd suggest a perf run.
r? `@wesleywiser`
cc `@arora-aman`
The field is also renamed from `ident` to `name. In most cases,
we don't actually need the `Span`. A new `ident` method is added
to `VariantDef` and `FieldDef`, which constructs the full `Ident`
using `tcx.def_ident_span()`. This method is used in the cases
where we actually need an `Ident`.
This makes incremental compilation properly track changes
to the `Span`, without all of the invalidations caused by storing
a `Span` directly via an `Ident`.
Region info is completely unnecessary for upvar capture kind computation
and is only needed to create the final upvar tuple ty. Doing so makes
creation of UpvarCapture very cheap and expose further cleanup opportunity.
Remove `NullOp::Box`
Follow up of #89030 and MCP rust-lang/compiler-team#460.
~1 month later nothing seems to be broken, apart from a small regression that #89332 (1aac85bb716c09304b313d69d30d74fe7e8e1a8e) shows could be regained by remvoing the diverging path, so it shall be safe to continue and remove `NullOp::Box` completely.
r? `@jonas-schievink`
`@rustbot` label T-compiler
During borrowchecking, we treat any free (early-bound) regions on
the 'defining type' as `RegionClassification::External`. According
to the doc comments, we should only have 'external' regions when
checking a closure/generator.
However, a plain function can also have some if its regions
be considered 'early bound' - this occurs when the region is
constrained by an argument, appears in a `where` clause, or
in an opaque type. This was causing us to incorrectly mark these
regions as 'external', which caused some diagnostic code
to act as if we were referring to a 'parent' region from inside
a closure.
This PR marks all instantiated region variables as 'local'
when we're borrow-checking something other than a
closure/generator/inline-const.
Region inference contains several bitsets which are filled with large intervals
representing liveness. These can cause excessive memory usage, and are
relatively slow when growing to large sizes compared to the IntervalSet.
Instead of special-casing mutable pointers/references, we
now support general generic types (currently, we handle
`ty::Ref`, `ty::RawPtr`, and `ty::Adt`)
When a `ty::Adt` is involved, we show an additional note
explaining which of the type's generic parameters is
invariant (e.g. the `T` in `Cell<T>`). Currently, we don't
explain *why* a particular generic parameter ends up becoming
invariant. In the general case, this could require printing
a long 'backtrace' of types, so doing this would be
more suitable for a follow-up PR.
We still only handle the case where our variance switches
to `ty::Invariant`.
The `AggregateKind` enum ends up in the final mir `Body`. Currently,
any changes to `AdtDef` (regardless of how significant they are)
will legitimately cause the overall result of `optimized_mir` to change,
invalidating any codegen re-use involving that mir.
This will get worse once we start hashing the `Span` inside `FieldDef`
(which is itself contained in `AdtDef`).
To try to reduce these kinds of invalidations, this commit changes
`AggregateKind::Adt` to store just the `DefId`, instead of the full
`AdtDef`. This allows the result of `optimized_mir` to be unchanged
if the `AdtDef` changes in a way that doesn't actually affect any
of the MIR we build.
Remove `SymbolStr`
This was originally proposed in https://github.com/rust-lang/rust/pull/74554#discussion_r466203544. As well as removing the icky `SymbolStr` type, it allows the removal of a lot of `&` and `*` occurrences.
Best reviewed one commit at a time.
r? `@oli-obk`
Point at capture points for non-`'static` reference crossing a `yield` point
```
error[E0759]: `self` has an anonymous lifetime `'_` but it needs to satisfy a `'static` lifetime requirement
--> $DIR/issue-72312.rs:10:24
|
LL | pub async fn start(&self) {
| ^^^^^ this data with an anonymous lifetime `'_`...
...
LL | require_static(async move {
| -------------- ...is required to live as long as `'static` here...
LL | &self;
| ----- ...and is captured here
|
note: `'static` lifetime requirement introduced by this trait bound
--> $DIR/issue-72312.rs:2:22
|
LL | fn require_static<T: 'static>(val: T) -> T {
| ^^^^^^^
error: aborting due to previous error
For more information about this error, try `rustc --explain E0759`.
```
Fix#72312.
In Rust, nesting method calls with both require `&mut` access to `self`
produces a borrow-check error:
error[E0499]: cannot borrow `*self` as mutable more than once at a time
--> src/lib.rs:7:14
|
7 | self.foo(self.bar());
| ---------^^^^^^^^^^-
| | | |
| | | second mutable borrow occurs here
| | first borrow later used by call
| first mutable borrow occurs here
That's because Rust has a left-to-right evaluation order, and the method
receiver is passed first. Thus, the argument to the method cannot then
mutate `self`.
There's an easy solution to this error: just extract a local variable
for the inner argument:
let tmp = self.bar();
self.foo(tmp);
However, the error doesn't give any suggestion of how to solve the
problem. As a result, new users may assume that it's impossible to
express their code correctly and get stuck.
This commit adds a (non-structured) suggestion to extract a local
variable for the inner argument to solve the error. The suggestion uses
heuristics that eliminate most false positives, though there are a few
false negatives (cases where the suggestion should be emitted but is
not). Those other cases can be implemented in a future change.
```
error[E0759]: `self` has an anonymous lifetime `'_` but it needs to satisfy a `'static` lifetime requirement
--> $DIR/issue-72312.rs:10:24
|
LL | pub async fn start(&self) {
| ^^^^^ this data with an anonymous lifetime `'_`...
...
LL | require_static(async move {
| -------------- ...is required to live as long as `'static` here...
LL | &self;
| ----- ...and is captured here
|
note: `'static` lifetime requirement introduced by this trait bound
--> $DIR/issue-72312.rs:2:22
|
LL | fn require_static<T: 'static>(val: T) -> T {
| ^^^^^^^
error: aborting due to previous error
For more information about this error, try `rustc --explain E0759`.
```
Fix#72312.
Fix ICE when `yield`ing in function returning `impl Trait`
Change an assert to a `delay_span_bug` and remove an unwrap, that should fix it.
Fixes#91477
Cleanup: Eliminate ConstnessAnd
This is almost a behaviour-free change and purely a refactoring. "almost" because we appear to be using the wrong ParamEnv somewhere already, and this is now exposed by failing a test using the unstable `~const` feature.
We most definitely need to review all `without_const` and at some point should probably get rid of many of them by using `TraitPredicate` instead of `TraitRef`.
This is a continuation of https://github.com/rust-lang/rust/pull/90274.
r? `@oli-obk`
cc `@spastorino` `@ecstatic-morse`
Optimize live point computation
This refactors the live-point computation to lower per-MIR-instruction costs by operating on a largely per-block level. This doesn't fundamentally change the number of operations necessary, but it greatly improves the practical performance by aggregating bit manipulation into ranges rather than single-bit; this scales much better with larger blocks.
On the benchmark provided in #90445, with 100,000 array elements, walltime for a check build is improved from 143 seconds to 15.
I consider the tiny losses here acceptable given the many small wins on real world benchmarks and large wins on stress tests. The new code scales much better, but on some subset of inputs the slightly higher constant overheads decrease performance somewhat. Overall though, this is expected to be a big win for pathological cases (as illustrated by the test case motivating this work) and largely not material for non-pathological cases. I consider the new code somewhat easier to follow, too.
Type inference for inline consts
Fixes#78132Fixes#78174Fixes#81857Fixes#89964
Perform type checking/inference of inline consts in the same context as the outer def, similar to what is currently done to closure.
Doing so would require `closure_base_def_id` of the inline const to return the outer def, and since `closure_base_def_id` can be called on non-local crate (and thus have no HIR available), a new `DefKind` is created for inline consts.
The type of the generated anon const can capture lifetime of outer def, so we couldn't just use the typeck result as the type of the inline const's def. Closure has a similar issue, and it uses extra type params `CK, CS, U` to capture closure kind, input/output signature and upvars. I use a similar approach for inline consts, letting it have an extra type param `R`, and then `typeof(InlineConst<[paremt generics], R>)` would just be `R`. In borrowck region requirements are also propagated to the outer MIR body just like it's currently done for closure.
With this PR, inline consts in expression position are quitely usable now; however the usage in pattern position is still incomplete -- since those does not remain in the MIR borrowck couldn't verify the lifetime there. I have left an ignored test as a FIXME.
Some disucssions can be found on [this Zulip thread](https://rust-lang.zulipchat.com/#narrow/stream/260443-project-const-generics/topic/inline.20consts.20typeck).
cc `````@spastorino````` `````@lcnr`````
r? `````@nikomatsakis`````
`````@rustbot````` label A-inference F-inline_const T-compiler
This is just replicating the previous algorithm, but taking advantage of the
bitset structures to optimize into tighter and better optimized loops.
Particularly advantageous on enormous MIR blocks, which are relatively rare in
practice.
Revert "Add rustc lint, warning when iterating over hashmaps"
Fixes perf regressions introduced in https://github.com/rust-lang/rust/pull/90235 by temporarily reverting the relevant PR.
Add BorrowSet to public api
This PR adds `BorrowSet` to the public api so that verification tools can obtain the activation and reservation points of two phase borrows without having to redo calculations themselves (and thus potentially differently from rustc).
Turns out we already can obtain `MoveData` thanks to the public `HasMoveData` trait, so constructing a `BorrowSet` should not provide much of an issue. However, I can't speak to the soundness of this approach, is it safe to take an under-approximation of `MoveData`?
r? `@nikomatsakis`
Implement coherence checks for negative trait impls
The main purpose of this PR is to be able to [move Error trait to core](https://github.com/rust-lang/project-error-handling/issues/3).
This feature is necessary to handle the following from impl on box.
```rust
impl From<&str> for Box<dyn Error> { ... }
```
Without having negative traits affect coherence moving the error trait into `core` and moving that `From` impl to `alloc` will cause the from impl to no longer compiler because of a potential future incompatibility. The compiler indicates that `&str` _could_ introduce an `Error` impl in the future, and thus prevents the `From` impl in `alloc` that would cause overlap with `From<E: Error> for Box<dyn Error>`. Adding `impl !Error for &str {}` with the negative trait coherence feature will disable this error by encoding a stability guarantee that `&str` will never implement `Error`, making the `From` impl compile.
We would have this in `alloc`:
```rust
impl From<&str> for Box<dyn Error> {} // A
impl<E> From<E> for Box<dyn Error> where E: Error {} // B
```
and this in `core`:
```rust
trait Error {}
impl !Error for &str {}
```
r? `@nikomatsakis`
This PR was built on top of `@yaahc` PR #85764.
Language team proposal: to https://github.com/rust-lang/lang-team/issues/96
Don't mark for loop iter expression as desugared
We typically don't mark spans of lowered things as desugared. This helps Clippy rightly discern when code is (not) from expansion. This was discovered by ``@flip1995`` at https://github.com/rust-lang/rust-clippy/pull/7789#issuecomment-939289501.
Adopt let_else across the compiler
This performs a substitution of code following the pattern:
```
let <id> = if let <pat> = ... { identity } else { ... : ! };
```
To simplify it to:
```
let <pat> = ... { identity } else { ... : ! };
```
By adopting the `let_else` feature (cc #87335).
The PR also updates the syn crate because the currently used version of the crate doesn't support `let_else` syntax yet.
Note: Generally I'm the person who *removes* usages of unstable features from the compiler, not adds more usages of them, but in this instance I think it hopefully helps the feature get stabilized sooner and in a better state. I have written a [comment](https://github.com/rust-lang/rust/issues/87335#issuecomment-944846205) on the tracking issue about my experience and what I feel could be improved before stabilization of `let_else`.
Remove redundant member-constraint check
impl trait will, for each lifetime in the hidden type, register a "member constraint" that says the lifetime must be equal or outlive one of the lifetimes of the impl trait. These member constraints will be solved by borrowck
But, as you can see in the big red block of removed code, there was an ad-hoc check for member constraints happening at the site where they get registered. This check had some minor effects on diagnostics, but will fall down on its feet with my big type alias impl trait refactor. So we removed it and I pulled the removal out into a (hopefully) reviewable PR that works on master directly.
This performs a substitution of code following the pattern:
let <id> = if let <pat> = ... { identity } else { ... : ! };
To simplify it to:
let <pat> = ... { identity } else { ... : ! };
By adopting the let_else feature.
Add check that live_region is live in sanitize_promoted
This pull request fixes#88434 by adding a check in `sanitize_promoted` to ensure that only regions which are actually live are added to the `liveness_constraints` of the `BorrowCheckContext`.
To implement this change, I needed to add a method to `LivenessValues` which gets the elements contained by a region:
/// Returns an iterator of all the elements contained by the region `r`
crate fn get_elements(&self, row: N) -> impl Iterator<Item = Location> + '_
Then, inside `sanitize_promoted`, we check whether the iterator returned by this method is non-empty to ensure that the region is actually live at at least one location before adding that region to the `liveness_constraints` of the `BorrowCheckContext`.
This is my first pull request to the Rust repo, so any feedback on how I can improve this pull request or if there is a better way to fix this issue would be very appreciated.
Re-use TypeChecker instead of passing around some of its fields
In the future (for lazy TAIT) we will need more of its fields, but even ignoring that, this change seems reasonable on its own to me.
Fixes#67007
Currently, a 'borrowed data escapes' error does not mention
the specific lifetime involved (except indirectly through a suggestion
about adding a lifetime bound). We now explain the specific lifetime
relationship that failed to hold, which improves otherwise vague
error messages.
Don't suggest replacing region with 'static in NLL
Fixes#73159
This is similar to #69350 - if the user didn't initially
write out a 'static lifetime, adding 'static in response to
a lifetime error is usually the wrong thing to do.
Stabilize `const_panic`
Closes#51999
FCP completed in #89006
```@rustbot``` label +A-const-eval +A-const-fn +T-lang
cc ```@oli-obk``` for review (not `r?`'ing as not on lang team)
Remove some feature gates
The first commit removes various feature gates that are unused. The second commit replaces some `Fn` implementations with `Iterator` implementations, which is much cleaner IMO. The third commit replaces an unboxed_closures feature gate with min_specialization. For some reason the unboxed_closures feature gate suppresses the min_specialization feature gate from triggering on an `TrustedStep` impl. The last comment just turns a regular comment into a doc comment as drive by cleanup. I can move it to a separate PR if preferred.
Fixes#73159
This is similar to #69350 - if the user didn't initially
write out a 'static lifetime, adding 'static in response to
a lifetime error is usually the wrong thing to do.
Pick one possible lifetime in case there are multiple choices
In case a lifetime variable is created, but doesn't have an obvious lifetime in the list of named lifetimes that it should be inferred to, just pick the first one for the diagnostic.
This happens e.g. in
```rust
fn foo<'a, 'b>(a: Struct<'a>, b: Struct<'b>) -> impl Trait<'a, 'b> {
if bar() { a } else { b }
}
```
where we get a lifetime variable that combines the lifetimes of `a` and `b` creating a lifetime that is the intersection of both. Right now the type system cannot express this and thus we get an error, but that error also can't express this.
I can also create an entirely new diagnostic that mentions all involved lifetimes, so it would actually mention `'a` and `'b` instead of just `'b`.
Avoid spurious "previous iteration of loop" errors
Only follow backwards edges during `get_moved_indexes` if the move path is definitely initialized at loop entry. Otherwise, the error occurred prior to the loop, so we ignore the backwards edges to avoid generating misleading "value moved here, in previous iteration of loop" errors.
This patch also slightly improves the analysis of inits, including `NonPanicPathOnly` initializations (which are ignored by `drop_flag_effects::for_location_inits`). This is required for the definite initialization analysis, but may also help find certain skipped reinits in rare cases.
Patch passes all non-ignored src/test/ui testcases.
Fixes#72649.
This PR has several interconnected pieces:
1. In some of the NLL region error code, we now pass
around an `ObligationCause`, instead of just a plain `Span`.
This gets forwarded into `fulfill_cx.register_predicate_obligation`
during error reporting.
2. The general InferCtxt error reporting code is extended to
handle `ObligationCauseCode::BindingObligation`
3. A new enum variant `ConstraintCategory::Predicate` is added.
We try to avoid using this as the 'best blame constraint' - instead,
we use it to enhance the `ObligationCause` of the `BlameConstraint`
that we do end up choosing.
As a result, several NLL error messages now contain the same
"the lifetime requirement is introduced here" message as non-NLL
errors.
Having an `ObligationCause` available will likely prove useful
for future improvements to NLL error messages.
Introduce `Rvalue::ShallowInitBox`
Polished version of #88700.
Implements MCP rust-lang/compiler-team#460, and should allow #43596 to go forward.
In short, creating an empty box is split from a nullary-op `NullOp::Box` into two steps, first a call to `exchange_malloc`, then a `Rvalue::ShallowInitBox` which transmutes `*mut u8` to a shallow-initialized `Box<T>`. This allows the `exchange_malloc` call to unwind. Details can be found in the MCP.
`NullOp::Box` is not yet removed, purely to make reverting easier in case anything goes wrong as the result of this PR. If revert is needed a reversion of "Use Rvalue::ShallowInitBox for box expression" commit followed by a test bless should be sufficient.
Experiments in #88700 showed a very slight compile-time perf regression due to (supposedly) slightly more time spent in LLVM. We could omit unwind edge generation (in non-`oom=panic` case) in box expression MIR construction to restore perf; but I don't think it's necessary since runtime perf isn't affected and perf difference is rather small.
Be explicit about using Binder::dummy
This is somewhat of a late followup to the binder refactor PR. It removes `ToPredicate` and `ToPolyTraitImpls` that hide the use of `Binder::dummy`. While this does make code a bit more verbose, it allows us be more careful about where we create binders.
Another alternative here might be to add a new trait `ToBinder` or something with a `dummy()` fn. Which could still allow grepping but allows doing something like `trait_ref.dummy()` (but I also wonder if longer-term, it would be better to be even more explicit with a `bind_with_vars(ty::List::empty())` *but* that's not clear yet.
r? ``@nikomatsakis``
Add `ConstraintCategory::Usage` for handling aggregate construction
In some cases, we emit borrowcheck diagnostics pointing
at a particular field expression in a struct expression
(e.g. `MyStruct { field: my_expr }`). However, this
behavior currently relies on us choosing the
`ConstraintCategory::Boring` with the 'correct' span.
When adding additional variants to `ConstraintCategory`,
(or changing existing usages away from `ConstraintCategory::Boring`),
the current behavior can easily get broken, since a non-boring
constraint will get chosen over a boring one.
To make the diagnostic output less fragile, this commit
adds a `ConstraintCategory::Usage` variant. We use this variant
for the temporary assignments created for each field of
an aggregate we are constructing.
Using this new variant, we can emit a message mentioning
"this usage", emphasizing the fact that the error message
is related to the specific use site (in the struct expression).
This is preparation for additional work on improving NLL error messages
(see #57374)
Use explicit log level in tracing instrument macro
Specify a log level in tracing instrument macro explicitly.
Additionally reduce the used log level from a default info level to a
debug level (all of those appear to be developer oriented logs, so there
should be no need to include them in release builds).
Point to closure when emitting 'cannot move out' for captured variable
Attempts to fix#87456. The error message now points to the capturing closure, but I was not able to explain _why_ the closure implements `Fn` or `FnMut` (`TypeckResults::closure_kind_origins` did not contain anything for the closure in question).
cc `@Aaron1011`
Point at argument instead of call for their obligations
When an obligation is introduced by a specific `fn` argument, point at
the argument instead of the `fn` call if the obligation fails to be
fulfilled.
Move the information about pointing at the call argument expression in
an unmet obligation span from the `FulfillmentError` to a new
`ObligationCauseCode`.
When giving an error about an obligation introduced by a function call
that an argument doesn't fulfill, and that argument is a block, add a
span_label pointing at the innermost tail expression.
Current output:
```
error[E0425]: cannot find value `x` in this scope
--> f10.rs:4:14
|
4 | Some(x * 2)
| ^ not found in this scope
error[E0277]: expected a `FnOnce<({integer},)>` closure, found `Option<_>`
--> f10.rs:2:31
|
2 | let p = Some(45).and_then({
| ______________________--------_^
| | |
| | required by a bound introduced by this call
3 | | |x| println!("doubling {}", x);
4 | | Some(x * 2)
| | -----------
5 | | });
| |_____^ expected an `FnOnce<({integer},)>` closure, found `Option<_>`
|
= help: the trait `FnOnce<({integer},)>` is not implemented for `Option<_>`
```
Previous output:
```
error[E0425]: cannot find value `x` in this scope
--> f10.rs:4:14
|
4 | Some(x * 2)
| ^ not found in this scope
error[E0277]: expected a `FnOnce<({integer},)>` closure, found `Option<_>`
--> f10.rs:2:22
|
2 | let p = Some(45).and_then({
| ^^^^^^^^ expected an `FnOnce<({integer},)>` closure, found `Option<_>`
|
= help: the trait `FnOnce<({integer},)>` is not implemented for `Option<_>`
```
Partially address #27300. Will require rebasing on top of #88546.
In some cases, we emit borrowcheck diagnostics pointing
at a particular field expression in a struct expression
(e.g. `MyStruct { field: my_expr }`). However, this
behavior currently relies on us choosing the
`ConstraintCategory::Boring` with the 'correct' span.
When adding additional variants to `ConstraintCategory`,
(or changing existing usages away from `ConstraintCategory::Boring`),
the current behavior can easily get broken, since a non-boring
constraint will get chosen over a boring one.
To make the diagnostic output less fragile, this commit
adds a `ConstraintCategory::Usage` variant. We use this variant
for the temporary assignments created for each field of
an aggregate we are constructing.
Using this new variant, we can emit a message mentioning
"this usage", emphasizing the fact that the error message
is related to the specific use site (in the struct expression).
This is preparation for additional work on improving NLL error messages
(see #57374)
Move the information about pointing at the call argument expression in
an unmet obligation span from the `FulfillmentError` to a new
`ObligationCauseCode`.
Specify a log level in tracing instrument macro explicitly.
Additionally reduce the used log level from a default info level to a
debug level (all of those appear to be developer oriented logs, so there
should be no need to include them in release builds).
Only follow backwards edges during get_moved_indexes if the move path is
definitely initialized at loop entry. Otherwise, the error occurred prior to the
loop, so we ignore the backwards edges to avoid generating misleading "value
moved here, in previous iteration of loop" errors.
This patch also slightly improves the analysis of inits, including
NonPanicPathOnly initializations (which are ignored by
drop_flag_effects::for_location_inits). This is required for the definite
initialization analysis, but may also help find certain skipped reinits in rare
cases.
Patch passes all non-ignored src/test/ui testcases.
