std: use futex in `Once`
Now that we have efficient locks, let's optimize the rest of `sync` as well. This PR adds a futex-based implementation for `Once`, which drastically simplifies the implementation compared to the generic version, which is provided as fallback for platforms without futex (Windows only supports them on newer versions, so it uses the fallback for now).
Instead of storing a linked list of waiters, the new implementation adds another state (`QUEUED`), which is set when there are waiting threads. These now use `futex_wait` on that state and are woken by the running thread when it finishes and notices the `QUEUED` state, thereby avoiding unnecessary calls to `futex_wake_all`.
Rollup of 6 pull requests
Successful merges:
- #102300 (Use a macro to not have to copy-paste `ConstFnMutClosure::new(&mut fold, NeverShortCircuit::wrap_mut_2_imp)).0` everywhere)
- #102475 (unsafe keyword: trait examples and unsafe_op_in_unsafe_fn update)
- #102760 (Avoid repeated re-initialization of the BufReader buffer)
- #102764 (Check `WhereClauseReferencesSelf` after all other object safety checks)
- #102779 (Fix `type_of` ICE)
- #102780 (run Miri CI when std::sys changes)
Failed merges:
r? `@ghost`
`@rustbot` modify labels: rollup
Check `WhereClauseReferencesSelf` after all other object safety checks
This fixes the ICE because it causes us to detect another *non-lint* `MethodViolationCode` first, instead of breaking on `WhereClauseReferencesSelf`.
We could also approach this issue by instead returning a vector of *all* of the `MethodViolationCode`s, and just reporting the first one we see, but treating it as a hard error if we return both `WhereClauseReferencesSelf` and some other violation code -- let me know if this is desired.
Fixes#102762
Avoid repeated re-initialization of the BufReader buffer
Fixes https://github.com/rust-lang/rust/issues/102727
We accidentally removed this in https://github.com/rust-lang/rust/pull/98748. It looks so redundant. But it isn't.
The default `Read::read_buf` will defensively initialize the whole buffer, if any of it is indicated to be uninitialized. In uses where reads from the wrapped `Read` impl completely fill the `BufReader`, `initialized` and `filled` are the same, and this extra member isn't required. But in the reported issue, the `BufReader` wraps a `Read` impl which will _never_ fill the whole buffer. So the default `Read::read_buf` implementation repeatedly re-initializes the extra space in the buffer.
This adds back the extra `initialized` member, which ensures that the default `Read::read_buf` only zero-initialized the buffer once, and I've tried to add a comment which explains this whole situation.
unsafe keyword: trait examples and unsafe_op_in_unsafe_fn update
Having a safe `fn` in an `unsafe trait` vs an `unsafe fn` in a safe `trait` are pretty different situations, but the distinction is subtle and can confuse even seasoned Rust developers. So let's have explicit examples of both. I also removed the existing `unsafe trait` example since it was rather strange.
Also the `unsafe_op_in_unsafe_fn` lint can help disentangle the two sides of `unsafe`, so update the docs to account for that.
Use a macro to not have to copy-paste `ConstFnMutClosure::new(&mut fold, NeverShortCircuit::wrap_mut_2_imp)).0` everywhere
Also use that macro to replace a bunch of places that had custom closure-wrappers.
+35 -114 sounds good to me.
Remove `TypeckResults` from `InferCtxt`
`InferCtxt` currently has `in_progress_typeck_results` which is only used for some diagnostics during typeck. It adds a lifetime which propagates through a lot of code. This PR moves that field into a new helper struct `TypeErrCtxt`.
let-else: test else block with non-never uninhabited type
let else currently does not allow uninhabited types for the `else` block that aren't `!`. One can maybe think about relaxing this in the future, but if it is done, it should be an explicit choice and not an unexpected side effect of e.g. a refactor. Thus, I'm extending a test that will fail if the behaviour changes.
Disable compressed debug sections on i586-gnu
Compressed debug is enabled by default for gas (assembly) on Linux/x86
targets, and we started building our own in #102530, but that made our
`compiler_builtins` incompatible with binutils < 2.32. Add an explicit
option to disable that in our crosstool-ng config. Fixes#102703.
rustdoc: remove unused CSS `.docblock a:not(.srclink)`
This selector was added in c7312fbae4, because the list of impl items could be nested below `docblock`.
c7312fbae4/src/librustdoc/html/render.rs (L3841-L3845)
Now that rustdoc toggles have been switched to `<details>`, there shouldn't be any need to put things inside docblock containers just to give them disclosure toggles.
rustdoc: remove unused CSS `.content .item-list`
When these rules were added in 4fd061c426 (yeah, that's the very first commit of rustdoc_ng), `.item-list` was a `<ul>`, and this would override the default style for that tag.
In c1b1d6804b, it was changed to use a `<div>` tag, so these rules are both no-ops.
Compressed debug is enabled by default for gas (assembly) on Linux/x86
targets, and we started building our own in #102530, but that made our
`compiler_builtins` incompatible with binutils < 2.32. Add an explicit
option to disable that in our crosstool-ng config. Fixes#102703.
This selector was added in c7312fbae4,
because the list of impl items could be nested below `docblock`.
c7312fbae4/src/librustdoc/html/render.rs (L3841-L3845)
Now that rustdoc toggles have been switched to `<details>`, there shouldn't
be any need to put things inside docblock containers just to give them
disclosure toggles.
When these rules were added in 4fd061c426
(yeah, that's the very first commit of rustdoc_ng), `.item-list` was a
`<ul>`, and this would override the default style for that tag.
In c1b1d6804b, it was changed to use a
`<div>` tag, so these rules are both no-ops.
lint::unsafe_removed_from_name: fix false positive result when allowed
changelog: [`unsafe_removed_from_name`] Fix allowing on imports produces a false positive on `useless_attribute`.
Fixes: #9197
Signed-off-by: Andy-Python-Programmer <andypythonappdeveloper@gmail.com>
Remove `-Ztime`
Because it has a lot of overlap with `-Ztime-passes` but is generally less useful. Plus some related cleanups.
Best reviewed one commit at a time.
r? `@davidtwco`
Fix overconstrained Send impls in btree internals
Fixes https://github.com/dtolnay/async-trait/issues/215.
Minimal repro:
```rust
use std::collections::btree_map::Iter;
fn require_send<T: Send>(_: T) {}
fn main() {
require_send(async {
let _iter = None::<Iter<(), &()>>;
async {}.await;
});
}
```
```console
error: higher-ranked lifetime error
--> src/main.rs:6:5
|
6 | / require_send(async {
7 | | let _iter = None::<Iter<(), &()>>;
8 | | async {}.await;
9 | | });
| |______^
|
= note: could not prove `impl Future<Output = ()>: Send`
```
Not-quite-so-minimal repro:
```rust
use std::collections::BTreeMap;
use std::future::Future;
fn spawn<T: Future + Send>(_: T) {}
async fn f() {
let map = BTreeMap::<u32, Box<dyn Send + Sync>>::new();
for _ in &map {
async {}.await;
}
}
fn main() {
spawn(f());
}
```
```console
error: higher-ranked lifetime error
--> src/main.rs:14:5
|
14 | spawn(f());
| ^^^^^^^^^^
|
= note: could not prove `impl Future<Output = ()>: Send`
```
I am not familiar with the btree internals, but it seems clear to me that the `async fn f` above should return a Send future. Using HashMap instead of BTreeMap in that code makes it already return a Send future.
The _"higher-ranked lifetime error"_ message may be a regression in Rust 1.63. Using older compilers the error message was more detailed:
```console
error: implementation of `Send` is not general enough
--> src/main.rs:14:5
|
14 | spawn(f());
| ^^^^^ implementation of `Send` is not general enough
|
= note: `Send` would have to be implemented for the type `alloc::collections::btree::node::NodeRef<alloc::collections::btree::node::marker::Immut<'0>, u32, Box<(dyn Send + Sync + '1)>, alloc::collections::btree::node::marker::LeafOrInternal>`, for any two lifetimes `'0` and `'1`...
= note: ...but `Send` is actually implemented for the type `alloc::collections::btree::node::NodeRef<alloc::collections::btree::node::marker::Immut<'2>, u32, Box<dyn Send + Sync>, alloc::collections::btree::node::marker::LeafOrInternal>`, for some specific lifetime `'2`
error: implementation of `Send` is not general enough
--> src/main.rs:14:5
|
14 | spawn(f());
| ^^^^^ implementation of `Send` is not general enough
|
= note: `Send` would have to be implemented for the type `alloc::collections::btree::node::NodeRef<alloc::collections::btree::node::marker::Immut<'0>, u32, Box<(dyn Send + Sync + '1)>, alloc::collections::btree::node::marker::Leaf>`, for any two lifetimes `'0` and `'1`...
= note: ...but `Send` is actually implemented for the type `alloc::collections::btree::node::NodeRef<alloc::collections::btree::node::marker::Immut<'2>, u32, Box<dyn Send + Sync>, alloc::collections::btree::node::marker::Leaf>`, for some specific lifetime `'2`
```
make `compare_const_impl` a query and use it in `instance.rs`
Fixes#88365
the bug in #88365 was caused by some `instance.rs` code using the `PartialEq` impl on `Ty` to check that the type of the associated const in an impl is the same as the type of the associated const in the trait definition. This was wrong for two reasons:
- the check typeck does is that the impl type is a subtype of the trait definition's type (see `mismatched_impl_ty_2.rs` which [was ICEing](https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=f6d60ebe6745011f0d52ab2bc712025d) before this PR on stable)
- it assumes that if two types are equal then the `PartialEq` impl will reflect that which isnt true for higher ranked types or type level constants when `feature(generic_const_exprs)` is enabled (see `mismatched_impl_ty_3.rs` for higher ranked types which was [ICEing on stable](https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=d7af131a655ed515b035624626c62c71))
r? `@lcnr`
Add a temporary workaround for multiline formart arg inlining
per suggestion in
https://github.com/rust-lang/rust/pull/102729#discussion_r988990080
workaround for an internal crash when handling multi-line format argument inlining.
changelog: none
(no point for changelog because it is still a new lint being introduced)
