Merge crossbeam-channel into `std::sync::mpsc`
This PR imports the [`crossbeam-channel`](https://github.com/crossbeam-rs/crossbeam/tree/master/crossbeam-channel#crossbeam-channel) crate into the standard library as a private module, `sync::mpmc`. `sync::mpsc` is now implemented as a thin wrapper around `sync::mpmc`. The primary purpose of this PR is to resolve https://github.com/rust-lang/rust/issues/39364. The public API intentionally remains the same.
The reason https://github.com/rust-lang/rust/issues/39364 has not been fixed in over 5 years is that the current channel is *incredibly* complex. It was written many years ago and has sat mostly untouched since. `crossbeam-channel` has become the most popular alternative on crates.io, amassing over 30 million downloads. While crossbeam's channel is also complex, like all fast concurrent data structures, it avoids some of the major issues with the current implementation around dynamic flavor upgrades. The new implementation decides on the datastructure to be used when the channel is created, and the channel retains that structure until it is dropped.
Replacing `sync::mpsc` with a simpler, less performant implementation has been discussed as an alternative. However, Rust touts itself as enabling *fearless concurrency*, and having the standard library feature a subpar implementation of a core concurrency primitive doesn't feel right. The argument is that slower is better than broken, but this PR shows that we can do better.
As mentioned before, the primary purpose of this PR is to fix https://github.com/rust-lang/rust/issues/39364, and so the public API intentionally remains the same. *After* that problem is fixed, the fact that `sync::mpmc` now exists makes it easier to fix the primary limitation of `mpsc`, the fact that it only supports a single consumer. spmc and mpmc are two other common concurrency patterns, and this change enables a path to deprecating `mpsc` and exposing a general `sync::channel` module that supports multiple consumers. It also implements other useful methods such as `send_timeout`. That said, exposing MPMC and other new functionality is mostly out of scope for this PR, and it would be helpful if discussion stays on topic :)
For what it's worth, the new implementation has also been shown to be more performant in [some basic benchmarks](https://github.com/crossbeam-rs/crossbeam/tree/master/crossbeam-channel/benchmarks#results).
cc `@taiki-e`
r? rust-lang/libs
Add the `#[derive_const]` attribute
Closes#102371. This is a minimal patchset for the attribute to work. There are no restrictions on what traits this attribute applies to.
r? `````@oli-obk`````
Remove lock wrappers in `sys_common`
This moves the lazy allocation to `sys` (SGX and UNIX). While this leads to a bit more verbosity, it will simplify future improvements by making room in `sys_common` for platform-independent implementations.
This also removes the condvar check on SGX as it is not necessary for soundness and will be removed anyway once mutex has been made movable.
For simplicity's sake, `libunwind` also uses lazy allocation now on SGX. This will require an update to the C definitions before merging this (CC `@raoulstrackx).`
r? `@m-ou-se`
Make `Hash`, `Hasher` and `BuildHasher` `#[const_trait]` and make `Sip` const `Hasher`
This PR enables using Hashes in const context.
r? ``@fee1-dead``
The new implementation doesn't use weak lang items and instead changes
`#[alloc_error_handler]` to an attribute macro just like
`#[global_allocator]`.
The attribute will generate the `__rg_oom` function which is called by
the compiler-generated `__rust_alloc_error_handler`. If no `__rg_oom`
function is defined in any crate then the compiler shim will call
`__rdl_oom` in the alloc crate which will simply panic.
This also fixes link errors with `-C link-dead-code` with
`default_alloc_error_handler`: `__rg_oom` was previously defined in the
alloc crate and would attempt to reference the `oom` lang item, even if
it didn't exist. This worked as long as `__rg_oom` was excluded from
linking since it was not called.
This is a prerequisite for the stabilization of
`default_alloc_error_handler` (#102318).
Truncate thread names on Linux and Apple targets
These targets have system limits on the thread names, 16 and 64 bytes
respectively, and `pthread_setname_np` returns an error if the name is
longer. However, we're not in a context that can propagate errors when
we call this, and we used to implicitly truncate on Linux with `prctl`,
so now we manually truncate these names ahead of time.
r? ``````@thomcc``````
Fix grammar in docs for std::io::Read
Two independent clauses were incorrectly joined by a bare comma. The simplest fix would be to switch to a semicolon, but I think it's slightly better to keep the comma and use the coordinating conjunction "so".
Update libstd's libc to 0.2.135 (to make `libstd` no longer pull in `libiconv.dylib` on Darwin)
This is to pull in https://github.com/rust-lang/libc/pull/2944.
It's related to https://github.com/rust-lang/rust/pull/102766, in that they both remove unused dylibs from libstd on Darwin platforms. As a result, I'm marking this as relnotes since everybody agreed it was good to add it to the other as well. (The note should be about no longer linking against libiconv -- the libc update is irrelevant).
Might as well have the same reviewer too.
r? `@Mark-Simulacrum`
Don't link to `libresolv` in libstd on Darwin
Currently we link `libresolv` into every Rust program on apple targets despite never using it (as of https://github.com/rust-lang/rust/pull/44965). I had thought we needed this for `getaddrinfo` or something, but we do not / cannot safely use it.
I'd like to fix this for `libiconv` too (the other library we pull in. that's harder since it's coming in through `libc`, which is https://github.com/rust-lang/libc/pull/2944)).
---
This may warrant release notes. I'm not sure but I've added the flag regardless -- It's a change to the list of dylibs every Rust program pulls in, so it's worth mentioning.
It's pretty unlikely anybody was relying on this being pulled in, and `std` does not guarantee that it will link (and thus transitively provide access to) any particular system library -- anybody relying on that behavior would already be broken when dynamically linking std. That is, there's an outside chance something will fail to link on macOS and iOS because it was accidentally relying on our unnecessary dependency.
(If that *does* happen, that project could be easily fixed by linking libresolv explicitly on those platforms, probably via `#[link(name = "resolv")] extern {}`,` -Crustc-link-lib=resolv`, `println!("cargo:rustc-link-lib=resolv")`, or one of several places in `.config/cargo.toml`)
---
I'm also going to preemptively add the nomination for discussing this in the libs meeting. Basically: Do we care about programs that assume we will bring libraries in that we do not use. `libresolv` and `libiconv` on macOS/iOS are in this camp (`libresolv` because we used to use it, and `libiconv` because the `libc` crate was unintentionally(?) pulling it in to every Rust program).
I'd like to remove them both, but this may cause link issues programs that are relying on `std` to depend on them transitively. (Relying on std for this does not work in all build configurations, so this seems very fragile, and like a use case we should not support).
More generally, IMO we should not guarantee the specific set of system-provided libraries we use (beyond what is implied by an OS version requirement), which means we'd be free to remove this cruft.
Stabilize `duration_checked_float`
## Stabilization Report
This stabilization report is for a stabilization of `duration_checked_float`, tracking issue: https://github.com/rust-lang/rust/issues/83400.
### Implementation History
- https://github.com/rust-lang/rust/pull/82179
- https://github.com/rust-lang/rust/pull/90247
- https://github.com/rust-lang/rust/pull/96051
- Changed error type to `FromFloatSecsError` in https://github.com/rust-lang/rust/pull/90247
- https://github.com/rust-lang/rust/pull/96051 changes the rounding mode to round-to-nearest instead of truncate.
## API Summary
This stabilization report proposes the following API to be stabilized in `core`, along with their re-exports in `std`:
```rust
// core::time
impl Duration {
pub const fn try_from_secs_f32(secs: f32) -> Result<Duration, TryFromFloatSecsError>;
pub const fn try_from_secs_f64(secs: f64) -> Result<Duration, TryFromFloatSecsError>;
}
#[derive(Debug, Clone, PartialEq, Eq)]
pub struct TryFromFloatSecsError { ... }
impl core::fmt::Display for TryFromFloatSecsError { ... }
impl core::error::Error for TryFromFloatSecsError { ... }
```
These functions are made const unstable under `duration_consts_float`, tracking issue #72440.
There is an open question in the tracking issue around what the error type should be called which I was hoping to resolve in the context of an FCP.
In this stabilization PR, I have altered the name of the error type to `TryFromFloatSecsError`. In my opinion, the error type shares the name of the method (adjusted to accommodate both types of floats), which is consistent with other error types in `core`, `alloc` and `std` like `TryReserveError` and `TryFromIntError`.
## Experience Report
Code such as this is ready to be converted to a checked API to ensure it is panic free:
```rust
impl Time {
pub fn checked_add_f64(&self, seconds: f64) -> Result<Self, TimeError> {
// Fail safely during `f64` conversion to duration
if seconds.is_nan() || seconds.is_infinite() {
return Err(TzOutOfRangeError::new().into());
}
if seconds.is_sign_positive() {
self.checked_add(Duration::from_secs_f64(seconds))
} else {
self.checked_sub(Duration::from_secs_f64(-seconds))
}
}
}
```
See: https://github.com/artichoke/artichoke/issues/2194.
`@rustbot` label +T-libs-api -T-libs
cc `@mbartlett21`
kmc-solid: Handle errors returned by `SOLID_FS_ReadDir`
Fixes the issue where the `std::fs::ReadDir` implementaton of the [`*-kmc-solid_*`](https://doc.rust-lang.org/nightly/rustc/platform-support/kmc-solid.html) Tier 3 targets silently suppressed errors returned by the underlying `SOLID_FS_ReadDir` system function. The new implementation correctly handles all cases:
- `SOLID_ERR_NOTFOUND` indicates the end of directory stream.
- `SOLID_ERR_OK` + non-empty `d_name` indicates success.
- Some old filesystem drivers may return `SOLID_ERR_OK` + empty `d_name` to indicate the end of directory stream.
- Any other negative values (per ITRON convention) represent an error.
Document surprising and dangerous fs::Permissions behaviour on Unix
This documents the very surprising behaviour that `set_readonly(false)` will make a file *world writable* on Unix. I would go so far as to say that this function should be deprecated on Unix, or maybe even entirely. But documenting the bad behaviour is a good first step.
Fixes#74895
Eliminate 280-byte memset from ReadDir iterator
This guy:
1536ab1b38/library/std/src/sys/unix/fs.rs (L589)
It turns out `libc::dirent64` is quite big—https://docs.rs/libc/0.2.135/libc/struct.dirent64.html. In #103135 this memset accounted for 0.9% of the runtime of iterating a big directory.
Almost none of the big zeroed value is ever used. We memcpy a tiny prefix (19 bytes) into it, and then read just 9 bytes (`d_ino` and `d_type`) back out. We can read exactly those 9 bytes we need directly from the original entry_ptr instead.
## History
This code got added in #93459 and tweaked in #94272 and #94750.
Prior to #93459, there was no memset but a full 280 bytes were being copied from the entry_ptr.
<table><tr><td>copy 280 bytes</td></tr></table>
This was not legal because not all of those bytes might be initialized, or even allocated, depending on the length of the directory entry's name, leading to a segfault. That PR fixed the segfault by creating a new zeroed dirent64 and copying just the guaranteed initialized prefix into it.
<table><tr><td>memset 280 bytes</td><td>copy 19 bytes</td></tr></table>
However this was still buggy because it used `addr_of!((*entry_ptr).d_name)`, which is considered UB by Miri in the case that the full extent of entry_ptr is not in bounds of the same allocation. (Arguably this shouldn't be a requirement, but here we are.)
The UB got fixed by #94272 by replacing `addr_of` with some pointer manipulation based on `offset_from`, but still fundamentally the same operation.
<table><tr><td>memset 280 bytes</td><td>copy 19 bytes</td></tr></table>
Then #94750 noticed that only 9 of those 19 bytes were even being used, so we could pick out only those 9 to put in the ReadDir value.
<table><tr><td>memset 280 bytes</td><td>copy 19 bytes</td><td>copy 9 bytes</td></tr></table>
After my PR we just grab the 9 needed bytes directly from entry_ptr.
<table><tr><td>copy 9 bytes</td></tr></table>
The resulting code is more complex but I believe still worthwhile to land for the following reason. This is an extremely straightforward thing to accomplish in C and clearly libc assumes that; literally just `entry_ptr->d_name`. The extra work in comparison to accomplish it in Rust is not an example of any actual safety being provided by Rust. I believe it's useful to have uncovered that and think about what could be done in the standard library or language to support this obvious operation better.
## References
- https://man7.org/linux/man-pages/man3/readdir.3.html
