Rollup of 7 pull requests
Successful merges:
- #139419 (Error on recursive opaque ty in HIR typeck)
- #141236 (Resolved issue with mismatched types triggering ICE in certain scenarios)
- #141253 (Warning added when dependency crate has async drop types, and the feature is disabled)
- #141269 (rustc-dev-guide subtree update)
- #141275 (`gather_locals`: only visit guard pattern guards when checking the guard)
- #141279 (`lower_to_hir` cleanups)
- #141285 (Add tick to `RePlaceholder` debug output)
r? `@ghost`
`@rustbot` modify labels: rollup
`gather_locals`: only visit guard pattern guards when checking the guard
When checking a pattern with guards in it, `GatherLocalsVisitor` will visit both the pattern (when type-checking the let, arm, or param containing it) and local declarations in the guard expression (when checking the guard itself). This keeps it from visiting the guard when visiting the pattern, since otherwise it would gather locals from the guard twice, which would lead to a delayed bug: "evaluated expression more than once".
Tracking issue for guard patterns: #129967
Warning added when dependency crate has async drop types, and the feature is disabled
In continue of https://github.com/rust-lang/rust/pull/141031.
When dependency crate has non-empty `adt_async_destructor` table in metadata, and `async_drop` feature is disabled for local crate, warning will be emitted.
Test `dependency-dropped` has two revisions - with and without feature enabled. With feature enabled, async drop for dropee is executed ("Async drop" printed). Without the feature enabled, sync drop is executed ("Sync drop" printed) and warning is emitted.
Warning example:
```
warning: found async drop types in dependecy `async_drop_dep`, but async_drop feature is disabled for `dependency_dropped`
--> $DIR/dependency-dropped.rs:7:1
|
LL | #![cfg_attr(with_feature, feature(async_drop))]
| ^
|
= help: if async drop type will be dropped in a crate without `feature(async_drop)`, sync Drop will be used
```
Resolved issue with mismatched types triggering ICE in certain scenarios
## Background
The function `annotate_mut_binding_to_immutable_binding` called in `emit_coerce_suggestions` performs a type comparison between the `expected` and `found` types from `ExpectedFound` in the `TypeError`. This can fail if the `found` type contains a region variable that's been rolled back.
## What is being changed?
This updates `annotate_mut_binding_to_immutable_binding` to use `expr_ty` and `expected` from the parent function instead of the types from the `TypeError`. This sidesteps the issue of using `found` from `TypeError` which may leak lingering inference region variables.
This does change the diagnostic behavior to _only_ support cases where the expected outermost type is `&T`, but that seems to be the intended functionality.
Also fixed the example in the `annotate_mut_binding_to_immutable_binding` rustdocs.
r? rust-lang/types
Fixes#140823
Error on recursive opaque ty in HIR typeck
"Non-trivially recursive" opaques are opaques whose hidden types are inferred to be equal to something other than themselves. For example, if we have a TAIT like `type TAIT = impl Sized`, if we infer the hidden type to be `TAIT := (TAIT,)`, that would be a non-trivial recursive definition. We don't want to support opaques that are non-trivially recursive, since they will (almost!! -- see caveat below) always result in borrowck errors, and are generally a pain to deal with.
On the contrary, trivially recursive opaques may occur today because the old solver overagerly uses `replace_opaque_types_with_inference_vars`. This infer var can then later be constrained to be equal to the opaque itself. These cases will not necessarily result in borrow-checker errors, since other uses of the opaque may properly constrain the opaque. If there are no other uses we may instead fall back to `()` today.
The only weird case that we have to unfortunately deal with was discovered in https://github.com/rust-lang/rust/issues/139406:
```rust
#![allow(unconditional_recursion)]
fn what1<T>(x: T) -> impl Sized {
what1(x)
}
fn what2<T>(x: T) -> impl Sized {
what2(what2(x))
}
fn print_return_type<T, U>(_: impl Fn(T) -> U) {
println!("{}", std::any::type_name::<U>())
}
fn main() {
print_return_type(what1::<i32>); // ()
print_return_type(what2::<i32>); // i32
}
```
> HIR typeck eagerly replaces the return type with an infer var, ending up with `RPIT<T> = RPIT<RPIT<T>>` in the storage. While we return this in the `TypeckResults`, it's never actually used anywhere.
>
> MIR building then results in the following statement
> ```rust
> let _0: impl RPIT<T> /* the return place */ = build<RPIT<T>>(_some_local);
> ```
> Unlike HIR typeck MIR typeck now directly equates `RPIT<T>` with `RPIT<RPIT<T>>`. This does not try to define `RPIT` but instead relates its generic arguments b9856b6e40/compiler/rustc_infer/src/infer/relate/type_relating.rs (L185-L190)
>
> This means we relate `T` with `RPIT<T>`, which results in a defining use `RPIT<T> = T`
I think it's pretty obvious that this is not desirable behavior, and according to the crater run there were no regressions, so let's break this so that we don't have any inference hazards in the new solver.
In the future `what2` may end up compiling again by also falling back to `()`. However, that is not yet guaranteed and the transition to this state is made significantly harder by not temporarily breaking it on the way. It is also concerning to change the inferred hidden type like this without any notification to the user, even if likely not an issue in this concrete case.
update llvm-tools logic for `dist` and `install` steps
First commit aligns `build_steps::compile` and `build_steps::dist` logics for copying llvm-tools, and the second commit adds the correct `should_run` condition for `LlvmTools` step as the previous one was clearly incorrect.
Fixes#140913
When checking a pattern with guards in it, `GatherLocalsVisitor` will
visit both the pattern (when type-checking the let, arm, or param
containing it) and the guard expression (when checking the guard
itself). This keeps it from visiting the guard when visiting the
pattern, since otherwise it would gather locals from the guard twice,
which would lead to a delayed bug: "evaluated expression more than
once".
make std::intrinsics functions actually be intrinsics
Most of the functions in `std::intrinsics` are actually intrinsics, but some are not: for historical reasons, `std::intrinsics::{copy,copy_nonoverlapping,write_bytes}` are accessible on stable, and the versions in `std::ptr` are just re-exports. These functions are not intrinsics, but wrappers around the intrinsic, because they add extra debug assertions.
This PR makes the functions in `std::intrinsics` actually be intrinsics.
- The advantage is that we can now use it in tests that need to directly call the intrinsic, thus removing a footgun for compiler development. We also remove the extended user-facing doc comments of these functions out of a file that should be largely internal documentation.
- The downside is that if users are using those functions directly, they will not get the debug assertions any more. Note however that those users are already ignoring a deprecation warning, so I think this is fine. Furthermore, if someone imports the `intrinsic` name of this function and turns that into a function pointer, that will no longer work, since only the wrapper functions can be turned into a function pointer. I would be rather surprised if anyone did this, though... and again, they must have already ignored a deprecation warning. Still, seems worth a crater run, if there's general agreement that we want to go ahead with this change.
(`intrinsics::drop_in_place` also remains not-an-intrinsic, which bugs me, but oh well, not much we can do about it; we can't remove it from the module as the path is accidentally-stable.)
Cc `@rust-lang/libs-api` `@saethlin`
Properly remove Noratrieb from review rotation
I've put myself on vacation a while ago, but really I just want to remove myself from the queue because I couldn't get around to reviewing all the PRs, I'm still here and available :3.
Update books
## rust-lang/book
4 commits in d33916341d480caede1d0ae57cbeae23aab23e88..230c68bc1e08f5f3228384a28cc228c81dfbd10d
2025-05-19 14:25:14 UTC to 2025-05-08 21:28:56 UTC
- Chapter 6 from tech review (rust-lang/book#4370)
- Chapter 5 from tech review (rust-lang/book#4359)
- Chapter 4 from tech review (rust-lang/book#4358)
- Chapter 3 from tech review (rust-lang/book#4353)
## rust-lang/reference
12 commits in 387392674d74656f7cb437c05a96f0c52ea8e601..acd0231ebc74849f6a8907b5e646ce86721aad76
2025-05-19 15:41:22 UTC to 2025-05-06 21:36:01 UTC
- Add doc for avx512 target features (rust-lang/reference#1778)
- Parse grammar without regexes (rust-lang/reference#1827)
- Parse optionals and repeats without regexes (rust-lang/reference#1826)
- Fix grammar for `RangePatternBound` regarding literals (rust-lang/reference#1825)
- Fix grammar for `LiteralPattern` regarding `-` (rust-lang/reference#1824)
- Doc: Add the LoongArch stabilized target features (rust-lang/reference#1707)
- Fix naked em-dash (rust-lang/reference#1820)
- Add missing attribute for statement macros (rust-lang/reference#1819)
- Make linked rules are clicked, highlight the color (rust-lang/reference#1817)
- Use the reference grammar for inline assembly (rust-lang/reference#1807)
- Fix typo in introduction (rust-lang/reference#1810)
- Add an example admonition (rust-lang/reference#1812)
## rust-lang/rust-by-example
2 commits in 8a8918c698534547fa8a1a693cb3e7277f0bfb2f..c9d151f9147c4808c77f0375ba3fa5d54443cb9e
2025-05-13 17:49:05 UTC to 2025-05-13 17:48:43 UTC
- fix(docs): standardize on `no_run` attribute for documentation examples (rust-lang/rust-by-example#1929)
- Fix typo in Japanese translation (rust-lang/rust-by-example#1928)
trim cache module in utils bootstrap
We don't use other variants of Interner in bootstrap, so this PR streamlines the bootstrap cache utils module.
r? `@onur-ozkan`
Replace `try_reserve_exact` with `try_with_capacity` in `std::fs::read`
This change restores the previous behavior prior to #117925. That PR was made to handle OOM errors that turn into a panic with `Vec::with_capacity`. `try_reserve_exact` was used for that since there was no `try_with_capacity` method at the time. It was added later in #120504. I think it'd a better fit here.
I've put myself on vacation a while ago, but really I just want to remove myself from the queue because I couldn't get around to reviewing all the PRs, I'm still here and available :3.
Rollup of 6 pull requests
Successful merges:
- #131200 (Handle `rustc_query_system` cases of `rustc::potential_query_instability` lint)
- #141244 (windows: document that we rely on an undocumented property of GetUserProfileDirectoryW)
- #141247 (skip compiler tools sanity checks on certain commands)
- #141248 (fix data race in ReentrantLock fallback for targets without 64bit atomics)
- #141249 (introduce common macro for `MutVisitor` and `Visitor` to dedup code)
- #141253 (Warning added when dependency crate has async drop types, and the feature is disabled)
r? `@ghost`
`@rustbot` modify labels: rollup
Warning added when dependency crate has async drop types, and the feature is disabled
In continue of https://github.com/rust-lang/rust/pull/141031.
When dependency crate has non-empty `adt_async_destructor` table in metadata, and `async_drop` feature is disabled for local crate, warning will be emitted.
Test `dependency-dropped` has two revisions - with and without feature enabled. With feature enabled, async drop for dropee is executed ("Async drop" printed). Without the feature enabled, sync drop is executed ("Sync drop" printed) and warning is emitted.
Warning example:
```
warning: found async drop types in dependecy `async_drop_dep`, but async_drop feature is disabled for `dependency_dropped`
--> $DIR/dependency-dropped.rs:7:1
|
LL | #![cfg_attr(with_feature, feature(async_drop))]
| ^
|
= help: if async drop type will be dropped in a crate without `feature(async_drop)`, sync Drop will be used
```
introduce common macro for `MutVisitor` and `Visitor` to dedup code
helps with #127615.
I can do everything in one go but I figured it might be worth it to open a PR first for vibeck.
r? oli-obk
fix data race in ReentrantLock fallback for targets without 64bit atomics
See [Zulip](https://rust-lang.zulipchat.com/#narrow/channel/269128-miri/topic/reentrant.20lock.20failure.20on.20musl) for details: the address used to identify a thread might get lazily allocated inside `tls_addr()`, so if we call that *after* doing the `tls_addr.load()` it is too late to establish synchronization with prior threads that used the same address -- the `load()` thus races with the `store()` by that prior thread, and might hence see outdated values, and then the entire logic breaks down.
r? `@joboet`
