diff options
author | Daniel Baumann <daniel.baumann@progress-linux.org> | 2024-04-17 12:06:31 +0000 |
---|---|---|
committer | Daniel Baumann <daniel.baumann@progress-linux.org> | 2024-04-17 12:06:31 +0000 |
commit | 2ff14448863ac1a1dd9533461708e29aae170c2d (patch) | |
tree | 85b9fea2bbfe3f06473cfa381eed11f273b57c5c /library/std | |
parent | Adding debian version 1.64.0+dfsg1-1. (diff) | |
download | rustc-2ff14448863ac1a1dd9533461708e29aae170c2d.tar.xz rustc-2ff14448863ac1a1dd9533461708e29aae170c2d.zip |
Adding debian version 1.65.0+dfsg1-2.debian/1.65.0+dfsg1-2
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'library/std')
145 files changed, 4005 insertions, 2023 deletions
diff --git a/library/std/Cargo.toml b/library/std/Cargo.toml index 229e546e0..324ecc804 100644 --- a/library/std/Cargo.toml +++ b/library/std/Cargo.toml @@ -39,10 +39,10 @@ rand = "0.7" dlmalloc = { version = "0.2.3", features = ['rustc-dep-of-std'] } [target.x86_64-fortanix-unknown-sgx.dependencies] -fortanix-sgx-abi = { version = "0.3.2", features = ['rustc-dep-of-std'] } +fortanix-sgx-abi = { version = "0.5.0", features = ['rustc-dep-of-std'] } [target.'cfg(target_os = "hermit")'.dependencies] -hermit-abi = { version = "0.2.0", features = ['rustc-dep-of-std'] } +hermit-abi = { version = "0.2.6", features = ['rustc-dep-of-std'] } [target.wasm32-wasi.dependencies] wasi = { version = "0.11.0", features = ['rustc-dep-of-std'], default-features = false } diff --git a/library/std/src/alloc.rs b/library/std/src/alloc.rs index a05e0db3a..61c1ff578 100644 --- a/library/std/src/alloc.rs +++ b/library/std/src/alloc.rs @@ -68,7 +68,10 @@ pub use alloc_crate::alloc::*; /// The default memory allocator provided by the operating system. /// /// This is based on `malloc` on Unix platforms and `HeapAlloc` on Windows, -/// plus related functions. +/// plus related functions. However, it is not valid to mix use of the backing +/// system allocator with `System`, as this implementation may include extra +/// work, such as to serve alignment requests greater than the alignment +/// provided directly by the backing system allocator. /// /// This type implements the `GlobalAlloc` trait and Rust programs by default /// work as if they had this definition: diff --git a/library/std/src/backtrace.rs b/library/std/src/backtrace.rs index 05e9b2eb6..5cf6ec817 100644 --- a/library/std/src/backtrace.rs +++ b/library/std/src/backtrace.rs @@ -9,12 +9,6 @@ //! implementing `std::error::Error`) to get a causal chain of where an error //! was generated. //! -//! > **Note**: this module is unstable and is designed in [RFC 2504], and you -//! > can learn more about its status in the [tracking issue]. -//! -//! [RFC 2504]: https://github.com/rust-lang/rfcs/blob/master/text/2504-fix-error.md -//! [tracking issue]: https://github.com/rust-lang/rust/issues/53487 -//! //! ## Accuracy //! //! Backtraces are attempted to be as accurate as possible, but no guarantees @@ -64,7 +58,7 @@ //! `RUST_LIB_BACKTRACE` or `RUST_BACKTRACE` at runtime might not actually change //! how backtraces are captured. -#![unstable(feature = "backtrace", issue = "53487")] +#![stable(feature = "backtrace", since = "1.65.0")] #[cfg(test)] mod tests; @@ -110,6 +104,7 @@ use crate::vec::Vec; /// previous point in time. In some instances the `Backtrace` type may /// internally be empty due to configuration. For more information see /// `Backtrace::capture`. +#[stable(feature = "backtrace", since = "1.65.0")] #[must_use] pub struct Backtrace { inner: Inner, @@ -117,17 +112,21 @@ pub struct Backtrace { /// The current status of a backtrace, indicating whether it was captured or /// whether it is empty for some other reason. +#[stable(feature = "backtrace", since = "1.65.0")] #[non_exhaustive] #[derive(Debug, PartialEq, Eq)] pub enum BacktraceStatus { /// Capturing a backtrace is not supported, likely because it's not /// implemented for the current platform. + #[stable(feature = "backtrace", since = "1.65.0")] Unsupported, /// Capturing a backtrace has been disabled through either the /// `RUST_LIB_BACKTRACE` or `RUST_BACKTRACE` environment variables. + #[stable(feature = "backtrace", since = "1.65.0")] Disabled, /// A backtrace has been captured and the `Backtrace` should print /// reasonable information when rendered. + #[stable(feature = "backtrace", since = "1.65.0")] Captured, } @@ -174,6 +173,7 @@ enum BytesOrWide { Wide(Vec<u16>), } +#[stable(feature = "backtrace", since = "1.65.0")] impl fmt::Debug for Backtrace { fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result { let capture = match &self.inner { @@ -200,6 +200,7 @@ impl fmt::Debug for Backtrace { } } +#[unstable(feature = "backtrace_frames", issue = "79676")] impl fmt::Debug for BacktraceFrame { fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result { let mut dbg = fmt.debug_list(); @@ -288,6 +289,7 @@ impl Backtrace { /// /// To forcibly capture a backtrace regardless of environment variables, use /// the `Backtrace::force_capture` function. + #[stable(feature = "backtrace", since = "1.65.0")] #[inline(never)] // want to make sure there's a frame here to remove pub fn capture() -> Backtrace { if !Backtrace::enabled() { @@ -306,6 +308,7 @@ impl Backtrace { /// Note that capturing a backtrace can be an expensive operation on some /// platforms, so this should be used with caution in performance-sensitive /// parts of code. + #[stable(feature = "backtrace", since = "1.65.0")] #[inline(never)] // want to make sure there's a frame here to remove pub fn force_capture() -> Backtrace { Backtrace::create(Backtrace::force_capture as usize) @@ -313,6 +316,8 @@ impl Backtrace { /// Forcibly captures a disabled backtrace, regardless of environment /// variable configuration. + #[stable(feature = "backtrace", since = "1.65.0")] + #[rustc_const_stable(feature = "backtrace", since = "1.65.0")] pub const fn disabled() -> Backtrace { Backtrace { inner: Inner::Disabled } } @@ -356,6 +361,7 @@ impl Backtrace { /// Returns the status of this backtrace, indicating whether this backtrace /// request was unsupported, disabled, or a stack trace was actually /// captured. + #[stable(feature = "backtrace", since = "1.65.0")] #[must_use] pub fn status(&self) -> BacktraceStatus { match self.inner { @@ -375,6 +381,7 @@ impl<'a> Backtrace { } } +#[stable(feature = "backtrace", since = "1.65.0")] impl fmt::Display for Backtrace { fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result { let capture = match &self.inner { diff --git a/library/std/src/collections/hash/map.rs b/library/std/src/collections/hash/map.rs index db811343f..9845d1faf 100644 --- a/library/std/src/collections/hash/map.rs +++ b/library/std/src/collections/hash/map.rs @@ -9,6 +9,8 @@ use crate::borrow::Borrow; use crate::cell::Cell; use crate::collections::TryReserveError; use crate::collections::TryReserveErrorKind; +#[cfg(not(bootstrap))] +use crate::error::Error; use crate::fmt::{self, Debug}; #[allow(deprecated)] use crate::hash::{BuildHasher, Hash, Hasher, SipHasher13}; @@ -2158,6 +2160,15 @@ impl<'a, K: Debug, V: Debug> fmt::Display for OccupiedError<'a, K, V> { } } +#[cfg(not(bootstrap))] +#[unstable(feature = "map_try_insert", issue = "82766")] +impl<'a, K: fmt::Debug, V: fmt::Debug> Error for OccupiedError<'a, K, V> { + #[allow(deprecated)] + fn description(&self) -> &str { + "key already exists" + } +} + #[stable(feature = "rust1", since = "1.0.0")] impl<'a, K, V, S> IntoIterator for &'a HashMap<K, V, S> { type Item = (&'a K, &'a V); diff --git a/library/std/src/collections/hash/map/tests.rs b/library/std/src/collections/hash/map/tests.rs index 7ebc41588..cb3032719 100644 --- a/library/std/src/collections/hash/map/tests.rs +++ b/library/std/src/collections/hash/map/tests.rs @@ -268,10 +268,13 @@ fn test_lots_of_insertions() { // Try this a few times to make sure we never screw up the hashmap's // internal state. - for _ in 0..10 { + let loops = if cfg!(miri) { 2 } else { 10 }; + for _ in 0..loops { assert!(m.is_empty()); - for i in 1..1001 { + let count = if cfg!(miri) { 101 } else { 1001 }; + + for i in 1..count { assert!(m.insert(i, i).is_none()); for j in 1..=i { @@ -279,42 +282,42 @@ fn test_lots_of_insertions() { assert_eq!(r, Some(&j)); } - for j in i + 1..1001 { + for j in i + 1..count { let r = m.get(&j); assert_eq!(r, None); } } - for i in 1001..2001 { + for i in count..(2 * count) { assert!(!m.contains_key(&i)); } // remove forwards - for i in 1..1001 { + for i in 1..count { assert!(m.remove(&i).is_some()); for j in 1..=i { assert!(!m.contains_key(&j)); } - for j in i + 1..1001 { + for j in i + 1..count { assert!(m.contains_key(&j)); } } - for i in 1..1001 { + for i in 1..count { assert!(!m.contains_key(&i)); } - for i in 1..1001 { + for i in 1..count { assert!(m.insert(i, i).is_none()); } // remove backwards - for i in (1..1001).rev() { + for i in (1..count).rev() { assert!(m.remove(&i).is_some()); - for j in i..1001 { + for j in i..count { assert!(!m.contains_key(&j)); } @@ -817,6 +820,7 @@ fn test_retain() { } #[test] +#[cfg_attr(miri, ignore)] // Miri does not support signalling OOM #[cfg_attr(target_os = "android", ignore)] // Android used in CI has a broken dlmalloc fn test_try_reserve() { let mut empty_bytes: HashMap<u8, u8> = HashMap::new(); diff --git a/library/std/src/collections/hash/set.rs b/library/std/src/collections/hash/set.rs index abff82788..5b6a415fa 100644 --- a/library/std/src/collections/hash/set.rs +++ b/library/std/src/collections/hash/set.rs @@ -239,7 +239,7 @@ impl<T, S> HashSet<T, S> { /// /// If the returned iterator is dropped before being fully consumed, it /// drops the remaining elements. The returned iterator keeps a mutable - /// borrow on the vector to optimize its implementation. + /// borrow on the set to optimize its implementation. /// /// # Examples /// diff --git a/library/std/src/error.rs b/library/std/src/error.rs index 722df119d..e45059595 100644 --- a/library/std/src/error.rs +++ b/library/std/src/error.rs @@ -1,175 +1,51 @@ -//! Interfaces for working with Errors. -//! -//! # Error Handling In Rust -//! -//! The Rust language provides two complementary systems for constructing / -//! representing, reporting, propagating, reacting to, and discarding errors. -//! These responsibilities are collectively known as "error handling." The -//! components of the first system, the panic runtime and interfaces, are most -//! commonly used to represent bugs that have been detected in your program. The -//! components of the second system, `Result`, the error traits, and user -//! defined types, are used to represent anticipated runtime failure modes of -//! your program. -//! -//! ## The Panic Interfaces -//! -//! The following are the primary interfaces of the panic system and the -//! responsibilities they cover: -//! -//! * [`panic!`] and [`panic_any`] (Constructing, Propagated automatically) -//! * [`PanicInfo`] (Reporting) -//! * [`set_hook`], [`take_hook`], and [`#[panic_handler]`][panic-handler] (Reporting) -//! * [`catch_unwind`] and [`resume_unwind`] (Discarding, Propagating) -//! -//! The following are the primary interfaces of the error system and the -//! responsibilities they cover: -//! -//! * [`Result`] (Propagating, Reacting) -//! * The [`Error`] trait (Reporting) -//! * User defined types (Constructing / Representing) -//! * [`match`] and [`downcast`] (Reacting) -//! * The question mark operator ([`?`]) (Propagating) -//! * The partially stable [`Try`] traits (Propagating, Constructing) -//! * [`Termination`] (Reporting) -//! -//! ## Converting Errors into Panics -//! -//! The panic and error systems are not entirely distinct. Often times errors -//! that are anticipated runtime failures in an API might instead represent bugs -//! to a caller. For these situations the standard library provides APIs for -//! constructing panics with an `Error` as it's source. -//! -//! * [`Result::unwrap`] -//! * [`Result::expect`] -//! -//! These functions are equivalent, they either return the inner value if the -//! `Result` is `Ok` or panic if the `Result` is `Err` printing the inner error -//! as the source. The only difference between them is that with `expect` you -//! provide a panic error message to be printed alongside the source, whereas -//! `unwrap` has a default message indicating only that you unwraped an `Err`. -//! -//! Of the two, `expect` is generally preferred since its `msg` field allows you -//! to convey your intent and assumptions which makes tracking down the source -//! of a panic easier. `unwrap` on the other hand can still be a good fit in -//! situations where you can trivially show that a piece of code will never -//! panic, such as `"127.0.0.1".parse::<std::net::IpAddr>().unwrap()` or early -//! prototyping. -//! -//! # Common Message Styles -//! -//! There are two common styles for how people word `expect` messages. Using -//! the message to present information to users encountering a panic -//! ("expect as error message") or using the message to present information -//! to developers debugging the panic ("expect as precondition"). -//! -//! In the former case the expect message is used to describe the error that -//! has occurred which is considered a bug. Consider the following example: -//! -//! ```should_panic -//! // Read environment variable, panic if it is not present -//! let path = std::env::var("IMPORTANT_PATH").unwrap(); -//! ``` -//! -//! In the "expect as error message" style we would use expect to describe -//! that the environment variable was not set when it should have been: -//! -//! ```should_panic -//! let path = std::env::var("IMPORTANT_PATH") -//! .expect("env variable `IMPORTANT_PATH` is not set"); -//! ``` -//! -//! In the "expect as precondition" style, we would instead describe the -//! reason we _expect_ the `Result` should be `Ok`. With this style we would -//! prefer to write: -//! -//! ```should_panic -//! let path = std::env::var("IMPORTANT_PATH") -//! .expect("env variable `IMPORTANT_PATH` should be set by `wrapper_script.sh`"); -//! ``` -//! -//! The "expect as error message" style does not work as well with the -//! default output of the std panic hooks, and often ends up repeating -//! information that is already communicated by the source error being -//! unwrapped: -//! -//! ```text -//! thread 'main' panicked at 'env variable `IMPORTANT_PATH` is not set: NotPresent', src/main.rs:4:6 -//! ``` -//! -//! In this example we end up mentioning that an env variable is not set, -//! followed by our source message that says the env is not present, the -//! only additional information we're communicating is the name of the -//! environment variable being checked. -//! -//! The "expect as precondition" style instead focuses on source code -//! readability, making it easier to understand what must have gone wrong in -//! situations where panics are being used to represent bugs exclusively. -//! Also, by framing our expect in terms of what "SHOULD" have happened to -//! prevent the source error, we end up introducing new information that is -//! independent from our source error. -//! -//! ```text -//! thread 'main' panicked at 'env variable `IMPORTANT_PATH` should be set by `wrapper_script.sh`: NotPresent', src/main.rs:4:6 -//! ``` -//! -//! In this example we are communicating not only the name of the -//! environment variable that should have been set, but also an explanation -//! for why it should have been set, and we let the source error display as -//! a clear contradiction to our expectation. -//! -//! **Hint**: If you're having trouble remembering how to phrase -//! expect-as-precondition style error messages remember to focus on the word -//! "should" as in "env variable should be set by blah" or "the given binary -//! should be available and executable by the current user". -//! -//! [`panic_any`]: crate::panic::panic_any -//! [`PanicInfo`]: crate::panic::PanicInfo -//! [`catch_unwind`]: crate::panic::catch_unwind -//! [`resume_unwind`]: crate::panic::resume_unwind -//! [`downcast`]: crate::error::Error -//! [`Termination`]: crate::process::Termination -//! [`Try`]: crate::ops::Try -//! [panic hook]: crate::panic::set_hook -//! [`set_hook`]: crate::panic::set_hook -//! [`take_hook`]: crate::panic::take_hook -//! [panic-handler]: <https://doc.rust-lang.org/nomicon/panic-handler.html> -//! [`match`]: ../../std/keyword.match.html -//! [`?`]: ../../std/result/index.html#the-question-mark-operator- - +#![doc = include_str!("../../core/src/error.md")] #![stable(feature = "rust1", since = "1.0.0")] -// A note about crates and the facade: -// -// Originally, the `Error` trait was defined in libcore, and the impls -// were scattered about. However, coherence objected to this -// arrangement, because to create the blanket impls for `Box` required -// knowing that `&str: !Error`, and we have no means to deal with that -// sort of conflict just now. Therefore, for the time being, we have -// moved the `Error` trait into libstd. As we evolve a sol'n to the -// coherence challenge (e.g., specialization, neg impls, etc) we can -// reconsider what crate these items belong in. - #[cfg(test)] mod tests; +#[cfg(bootstrap)] use core::array; +#[cfg(bootstrap)] use core::convert::Infallible; +#[cfg(bootstrap)] use crate::alloc::{AllocError, LayoutError}; -use crate::any::{Demand, Provider, TypeId}; +#[cfg(bootstrap)] +use crate::any::Demand; +#[cfg(bootstrap)] +use crate::any::{Provider, TypeId}; use crate::backtrace::Backtrace; +#[cfg(bootstrap)] use crate::borrow::Cow; +#[cfg(bootstrap)] use crate::cell; +#[cfg(bootstrap)] use crate::char; -use crate::fmt::{self, Debug, Display, Write}; +#[cfg(bootstrap)] +use crate::fmt::Debug; +#[cfg(bootstrap)] +use crate::fmt::Display; +use crate::fmt::{self, Write}; +#[cfg(bootstrap)] use crate::io; +#[cfg(bootstrap)] use crate::mem::transmute; +#[cfg(bootstrap)] use crate::num; +#[cfg(bootstrap)] use crate::str; +#[cfg(bootstrap)] use crate::string; +#[cfg(bootstrap)] use crate::sync::Arc; +#[cfg(bootstrap)] use crate::time; +#[cfg(not(bootstrap))] +#[stable(feature = "rust1", since = "1.0.0")] +pub use core::error::Error; + /// `Error` is a trait representing the basic expectations for error values, /// i.e., values of type `E` in [`Result<T, E>`]. /// @@ -182,14 +58,15 @@ use crate::time; /// assert_eq!(err.to_string(), "invalid digit found in string"); /// ``` /// -/// Errors may provide cause chain information. [`Error::source()`] is generally +/// Errors may provide cause information. [`Error::source()`] is generally /// used when errors cross "abstraction boundaries". If one module must report /// an error that is caused by an error from a lower-level module, it can allow /// accessing that error via [`Error::source()`]. This makes it possible for the /// high-level module to provide its own errors while also revealing some of the -/// implementation for debugging via `source` chains. +/// implementation for debugging. #[stable(feature = "rust1", since = "1.0.0")] #[cfg_attr(not(test), rustc_diagnostic_item = "Error")] +#[cfg(bootstrap)] pub trait Error: Debug + Display { /// The lower-level source of this error, if any. /// @@ -333,8 +210,8 @@ pub trait Error: Debug + Display { /// } /// /// impl std::error::Error for Error { - /// fn provide<'a>(&'a self, req: &mut Demand<'a>) { - /// req + /// fn provide<'a>(&'a self, demand: &mut Demand<'a>) { + /// demand /// .provide_ref::<MyBacktrace>(&self.backtrace) /// .provide_ref::<dyn std::error::Error + 'static>(&self.source); /// } @@ -352,13 +229,14 @@ pub trait Error: Debug + Display { /// ``` #[unstable(feature = "error_generic_member_access", issue = "99301")] #[allow(unused_variables)] - fn provide<'a>(&'a self, req: &mut Demand<'a>) {} + fn provide<'a>(&'a self, demand: &mut Demand<'a>) {} } +#[cfg(bootstrap)] #[unstable(feature = "error_generic_member_access", issue = "99301")] impl<'b> Provider for dyn Error + 'b { - fn provide<'a>(&'a self, req: &mut Demand<'a>) { - self.provide(req) + fn provide<'a>(&'a self, demand: &mut Demand<'a>) { + self.provide(demand) } } @@ -370,6 +248,7 @@ mod private { pub struct Internal; } +#[cfg(bootstrap)] #[stable(feature = "rust1", since = "1.0.0")] impl<'a, E: Error + 'a> From<E> for Box<dyn Error + 'a> { /// Converts a type of [`Error`] into a box of dyn [`Error`]. @@ -402,6 +281,7 @@ impl<'a, E: Error + 'a> From<E> for Box<dyn Error + 'a> { } } +#[cfg(bootstrap)] #[stable(feature = "rust1", since = "1.0.0")] impl<'a, E: Error + Send + Sync + 'a> From<E> for Box<dyn Error + Send + Sync + 'a> { /// Converts a type of [`Error`] + [`Send`] + [`Sync`] into a box of @@ -440,6 +320,7 @@ impl<'a, E: Error + Send + Sync + 'a> From<E> for Box<dyn Error + Send + Sync + } } +#[cfg(bootstrap)] #[stable(feature = "rust1", since = "1.0.0")] impl From<String> for Box<dyn Error + Send + Sync> { /// Converts a [`String`] into a box of dyn [`Error`] + [`Send`] + [`Sync`]. @@ -483,6 +364,7 @@ impl From<String> for Box<dyn Error + Send + Sync> { } } +#[cfg(bootstrap)] #[stable(feature = "string_box_error", since = "1.6.0")] impl From<String> for Box<dyn Error> { /// Converts a [`String`] into a box of dyn [`Error`]. @@ -504,6 +386,7 @@ impl From<String> for Box<dyn Error> { } } +#[cfg(bootstrap)] #[stable(feature = "rust1", since = "1.0.0")] impl<'a> From<&str> for Box<dyn Error + Send + Sync + 'a> { /// Converts a [`str`] into a box of dyn [`Error`] + [`Send`] + [`Sync`]. @@ -527,6 +410,7 @@ impl<'a> From<&str> for Box<dyn Error + Send + Sync + 'a> { } } +#[cfg(bootstrap)] #[stable(feature = "string_box_error", since = "1.6.0")] impl From<&str> for Box<dyn Error> { /// Converts a [`str`] into a box of dyn [`Error`]. @@ -548,6 +432,7 @@ impl From<&str> for Box<dyn Error> { } } +#[cfg(bootstrap)] #[stable(feature = "cow_box_error", since = "1.22.0")] impl<'a, 'b> From<Cow<'b, str>> for Box<dyn Error + Send + Sync + 'a> { /// Converts a [`Cow`] into a box of dyn [`Error`] + [`Send`] + [`Sync`]. @@ -569,6 +454,7 @@ impl<'a, 'b> From<Cow<'b, str>> for Box<dyn Error + Send + Sync + 'a> { } } +#[cfg(bootstrap)] #[stable(feature = "cow_box_error", since = "1.22.0")] impl<'a> From<Cow<'a, str>> for Box<dyn Error> { /// Converts a [`Cow`] into a box of dyn [`Error`]. @@ -589,9 +475,11 @@ impl<'a> From<Cow<'a, str>> for Box<dyn Error> { } } +#[cfg(bootstrap)] #[unstable(feature = "never_type", issue = "35121")] impl Error for ! {} +#[cfg(bootstrap)] #[unstable( feature = "allocator_api", reason = "the precise API and guarantees it provides may be tweaked.", @@ -599,9 +487,11 @@ impl Error for ! {} )] impl Error for AllocError {} +#[cfg(bootstrap)] #[stable(feature = "alloc_layout", since = "1.28.0")] impl Error for LayoutError {} +#[cfg(bootstrap)] #[stable(feature = "rust1", since = "1.0.0")] impl Error for str::ParseBoolError { #[allow(deprecated)] @@ -610,6 +500,7 @@ impl Error for str::ParseBoolError { } } +#[cfg(bootstrap)] #[stable(feature = "rust1", since = "1.0.0")] impl Error for str::Utf8Error { #[allow(deprecated)] @@ -618,6 +509,7 @@ impl Error for str::Utf8Error { } } +#[cfg(bootstrap)] #[stable(feature = "rust1", since = "1.0.0")] impl Error for num::ParseIntError { #[allow(deprecated)] @@ -626,6 +518,7 @@ impl Error for num::ParseIntError { } } +#[cfg(bootstrap)] #[stable(feature = "try_from", since = "1.34.0")] impl Error for num::TryFromIntError { #[allow(deprecated)] @@ -634,6 +527,7 @@ impl Error for num::TryFromIntError { } } +#[cfg(bootstrap)] #[stable(feature = "try_from", since = "1.34.0")] impl Error for array::TryFromSliceError { #[allow(deprecated)] @@ -642,6 +536,7 @@ impl Error for array::TryFromSliceError { } } +#[cfg(bootstrap)] #[stable(feature = "rust1", since = "1.0.0")] impl Error for num::ParseFloatError { #[allow(deprecated)] @@ -650,6 +545,7 @@ impl Error for num::ParseFloatError { } } +#[cfg(bootstrap)] #[stable(feature = "rust1", since = "1.0.0")] impl Error for string::FromUtf8Error { #[allow(deprecated)] @@ -658,6 +554,7 @@ impl Error for string::FromUtf8Error { } } +#[cfg(bootstrap)] #[stable(feature = "rust1", since = "1.0.0")] impl Error for string::FromUtf16Error { #[allow(deprecated)] @@ -666,6 +563,7 @@ impl Error for string::FromUtf16Error { } } +#[cfg(bootstrap)] #[stable(feature = "str_parse_error2", since = "1.8.0")] impl Error for Infallible { fn description(&self) -> &str { @@ -673,6 +571,7 @@ impl Error for Infallible { } } +#[cfg(bootstrap)] #[stable(feature = "decode_utf16", since = "1.9.0")] impl Error for char::DecodeUtf16Error { #[allow(deprecated)] @@ -681,9 +580,11 @@ impl Error for char::DecodeUtf16Error { } } +#[cfg(bootstrap)] #[stable(feature = "u8_from_char", since = "1.59.0")] impl Error for char::TryFromCharError {} +#[cfg(bootstrap)] #[unstable(feature = "map_try_insert", issue = "82766")] impl<'a, K: Debug + Ord, V: Debug> Error for crate::collections::btree_map::OccupiedError<'a, K, V> @@ -694,6 +595,7 @@ impl<'a, K: Debug + Ord, V: Debug> Error } } +#[cfg(bootstrap)] #[unstable(feature = "map_try_insert", issue = "82766")] impl<'a, K: Debug, V: Debug> Error for crate::collections::hash_map::OccupiedError<'a, K, V> { #[allow(deprecated)] @@ -702,6 +604,7 @@ impl<'a, K: Debug, V: Debug> Error for crate::collections::hash_map::OccupiedErr } } +#[cfg(bootstrap)] #[stable(feature = "box_error", since = "1.8.0")] impl<T: Error> Error for Box<T> { #[allow(deprecated, deprecated_in_future)] @@ -719,6 +622,7 @@ impl<T: Error> Error for Box<T> { } } +#[cfg(bootstrap)] #[unstable(feature = "thin_box", issue = "92791")] impl<T: ?Sized + crate::error::Error> crate::error::Error for crate::boxed::ThinBox<T> { fn source(&self) -> Option<&(dyn crate::error::Error + 'static)> { @@ -727,6 +631,7 @@ impl<T: ?Sized + crate::error::Error> crate::error::Error for crate::boxed::Thin } } +#[cfg(bootstrap)] #[stable(feature = "error_by_ref", since = "1.51.0")] impl<'a, T: Error + ?Sized> Error for &'a T { #[allow(deprecated, deprecated_in_future)] @@ -743,11 +648,12 @@ impl<'a, T: Error + ?Sized> Error for &'a T { Error::source(&**self) } - fn provide<'b>(&'b self, req: &mut Demand<'b>) { - Error::provide(&**self, req); + fn provide<'b>(&'b self, demand: &mut Demand<'b>) { + Error::provide(&**self, demand); } } +#[cfg(bootstrap)] #[stable(feature = "arc_error", since = "1.52.0")] impl<T: Error + ?Sized> Error for Arc<T> { #[allow(deprecated, deprecated_in_future)] @@ -764,11 +670,12 @@ impl<T: Error + ?Sized> Error for Arc<T> { Error::source(&**self) } - fn provide<'a>(&'a self, req: &mut Demand<'a>) { - Error::provide(&**self, req); + fn provide<'a>(&'a self, demand: &mut Demand<'a>) { + Error::provide(&**self, demand); } } +#[cfg(bootstrap)] #[stable(feature = "fmt_error", since = "1.11.0")] impl Error for fmt::Error { #[allow(deprecated)] @@ -777,6 +684,7 @@ impl Error for fmt::Error { } } +#[cfg(bootstrap)] #[stable(feature = "try_borrow", since = "1.13.0")] impl Error for cell::BorrowError { #[allow(deprecated)] @@ -785,6 +693,7 @@ impl Error for cell::BorrowError { } } +#[cfg(bootstrap)] #[stable(feature = "try_borrow", since = "1.13.0")] impl Error for cell::BorrowMutError { #[allow(deprecated)] @@ -793,6 +702,7 @@ impl Error for cell::BorrowMutError { } } +#[cfg(bootstrap)] #[stable(feature = "try_from", since = "1.34.0")] impl Error for char::CharTryFromError { #[allow(deprecated)] @@ -801,6 +711,7 @@ impl Error for char::CharTryFromError { } } +#[cfg(bootstrap)] #[stable(feature = "char_from_str", since = "1.20.0")] impl Error for char::ParseCharError { #[allow(deprecated)] @@ -809,12 +720,15 @@ impl Error for char::ParseCharError { } } +#[cfg(bootstrap)] #[stable(feature = "try_reserve", since = "1.57.0")] impl Error for alloc::collections::TryReserveError {} +#[cfg(bootstrap)] #[unstable(feature = "duration_checked_float", issue = "83400")] impl Error for time::FromFloatSecsError {} +#[cfg(bootstrap)] #[stable(feature = "rust1", since = "1.0.0")] impl Error for alloc::ffi::NulError { #[allow(deprecated)] @@ -823,6 +737,7 @@ impl Error for alloc::ffi::NulError { } } +#[cfg(bootstrap)] #[stable(feature = "rust1", since = "1.0.0")] impl From<alloc::ffi::NulError> for io::Error { /// Converts a [`alloc::ffi::NulError`] into a [`io::Error`]. @@ -831,6 +746,7 @@ impl From<alloc::ffi::NulError> for io::Error { } } +#[cfg(bootstrap)] #[stable(feature = "frombyteswithnulerror_impls", since = "1.17.0")] impl Error for core::ffi::FromBytesWithNulError { #[allow(deprecated)] @@ -839,12 +755,15 @@ impl Error for core::ffi::FromBytesWithNulError { } } +#[cfg(bootstrap)] #[unstable(feature = "cstr_from_bytes_until_nul", issue = "95027")] impl Error for core::ffi::FromBytesUntilNulError {} +#[cfg(bootstrap)] #[stable(feature = "cstring_from_vec_with_nul", since = "1.58.0")] impl Error for alloc::ffi::FromVecWithNulError {} +#[cfg(bootstrap)] #[stable(feature = "cstring_into", since = "1.7.0")] impl Error for alloc::ffi::IntoStringError { #[allow(deprecated)] @@ -857,6 +776,7 @@ impl Error for alloc::ffi::IntoStringError { } } +#[cfg(bootstrap)] impl<'a> dyn Error + 'a { /// Request a reference of type `T` as context about this error. #[unstable(feature = "error_generic_member_access", issue = "99301")] @@ -872,6 +792,7 @@ impl<'a> dyn Error + 'a { } // Copied from `any.rs`. +#[cfg(bootstrap)] impl dyn Error + 'static { /// Returns `true` if the inner type is the same as `T`. #[stable(feature = "error_downcast", since = "1.3.0")] @@ -912,6 +833,7 @@ impl dyn Error + 'static { } } +#[cfg(bootstrap)] impl dyn Error + 'static + Send { /// Forwards to the method defined on the type `dyn Error`. #[stable(feature = "error_downcast", since = "1.3.0")] @@ -947,6 +869,7 @@ impl dyn Error + 'static + Send { } } +#[cfg(bootstrap)] impl dyn Error + 'static + Send + Sync { /// Forwards to the method defined on the type `dyn Error`. #[stable(feature = "error_downcast", since = "1.3.0")] @@ -982,6 +905,7 @@ impl dyn Error + 'static + Send + Sync { } } +#[cfg(bootstrap)] impl dyn Error { #[inline] #[stable(feature = "error_downcast", since = "1.3.0")] @@ -1041,7 +965,7 @@ impl dyn Error { /// // let err : Box<Error> = b.into(); // or /// let err = &b as &(dyn Error); /// - /// let mut iter = err.chain(); + /// let mut iter = err.sources(); /// /// assert_eq!("B".to_string(), iter.next().unwrap().to_string()); /// assert_eq!("A".to_string(), iter.next().unwrap().to_string()); @@ -1050,8 +974,19 @@ impl dyn Error { /// ``` #[unstable(feature = "error_iter", issue = "58520")] #[inline] - pub fn chain(&self) -> Chain<'_> { - Chain { current: Some(self) } + pub fn sources(&self) -> Sources<'_> { + // You may think this method would be better in the Error trait, and you'd be right. + // Unfortunately that doesn't work, not because of the object safety rules but because we + // save a reference to self in Sources below as a trait object. If this method was + // declared in Error, then self would have the type &T where T is some concrete type which + // implements Error. We would need to coerce self to have type &dyn Error, but that requires + // that Self has a known size (i.e., Self: Sized). We can't put that bound on Error + // since that would forbid Error trait objects, and we can't put that bound on the method + // because that means the method can't be called on trait objects (we'd also need the + // 'static bound, but that isn't allowed because methods with bounds on Self other than + // Sized are not object-safe). Requiring an Unsize bound is not backwards compatible. + + Sources { current: Some(self) } } } @@ -1061,12 +996,14 @@ impl dyn Error { /// its sources, use `skip(1)`. #[unstable(feature = "error_iter", issue = "58520")] #[derive(Clone, Debug)] -pub struct Chain<'a> { +#[cfg(bootstrap)] +pub struct Sources<'a> { current: Option<&'a (dyn Error + 'static)>, } +#[cfg(bootstrap)] #[unstable(feature = "error_iter", issue = "58520")] -impl<'a> Iterator for Chain<'a> { +impl<'a> Iterator for Sources<'a> { type Item = &'a (dyn Error + 'static); fn next(&mut self) -> Option<Self::Item> { @@ -1076,6 +1013,7 @@ impl<'a> Iterator for Chain<'a> { } } +#[cfg(bootstrap)] impl dyn Error + Send { #[inline] #[stable(feature = "error_downcast", since = "1.3.0")] @@ -1089,6 +1027,7 @@ impl dyn Error + Send { } } +#[cfg(bootstrap)] impl dyn Error + Send + Sync { #[inline] #[stable(feature = "error_downcast", since = "1.3.0")] @@ -1104,8 +1043,8 @@ impl dyn Error + Send + Sync { /// An error reporter that prints an error and its sources. /// -/// Report also exposes configuration options for formatting the error chain, either entirely on a -/// single line, or in multi-line format with each cause in the error chain on a new line. +/// Report also exposes configuration options for formatting the error sources, either entirely on a +/// single line, or in multi-line format with each source on a new line. /// /// `Report` only requires that the wrapped error implement `Error`. It doesn't require that the /// wrapped error be `Send`, `Sync`, or `'static`. @@ -1246,7 +1185,7 @@ impl dyn Error + Send + Sync { /// # Err(SuperError { source: SuperErrorSideKick }) /// # } /// -/// fn main() -> Result<(), Report> { +/// fn main() -> Result<(), Report<SuperError>> { /// get_super_error()?; /// Ok(()) /// } @@ -1293,7 +1232,7 @@ impl dyn Error + Send + Sync { /// # Err(SuperError { source: SuperErrorSideKick }) /// # } /// -/// fn main() -> Result<(), Report> { +/// fn main() -> Result<(), Report<SuperError>> { /// get_super_error() /// .map_err(Report::from) /// .map_err(|r| r.pretty(true).show_backtrace(true))?; @@ -1450,11 +1389,10 @@ impl<E> Report<E> { /// /// **Note**: Report will search for the first `Backtrace` it can find starting from the /// outermost error. In this example it will display the backtrace from the second error in the - /// chain, `SuperErrorSideKick`. + /// sources, `SuperErrorSideKick`. /// /// ```rust /// #![feature(error_reporter)] - /// #![feature(backtrace)] /// #![feature(provide_any)] /// #![feature(error_generic_member_access)] /// # use std::error::Error; @@ -1489,9 +1427,8 @@ impl<E> Report<E> { /// } /// /// impl Error for SuperErrorSideKick { - /// fn provide<'a>(&'a self, req: &mut Demand<'a>) { - /// req - /// .provide_ref::<Backtrace>(&self.backtrace); + /// fn provide<'a>(&'a self, demand: &mut Demand<'a>) { + /// demand.provide_ref::<Backtrace>(&self.backtrace); /// } /// } /// @@ -1548,7 +1485,7 @@ where let backtrace = backtrace.or_else(|| { self.error .source() - .map(|source| source.chain().find_map(|source| source.request_ref())) + .map(|source| source.sources().find_map(|source| source.request_ref())) .flatten() }); backtrace @@ -1559,7 +1496,7 @@ where fn fmt_singleline(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { write!(f, "{}", self.error)?; - let sources = self.error.source().into_iter().flat_map(<dyn Error>::chain); + let sources = self.error.source().into_iter().flat_map(<dyn Error>::sources); for cause in sources { write!(f, ": {cause}")?; @@ -1580,73 +1517,7 @@ where let multiple = cause.source().is_some(); - for (ind, error) in cause.chain().enumerate() { - writeln!(f)?; - let mut indented = Indented { inner: f }; - if multiple { - write!(indented, "{ind: >4}: {error}")?; - } else { - write!(indented, " {error}")?; - } - } - } - - if self.show_backtrace { - let backtrace = self.backtrace(); - - if let Some(backtrace) = backtrace { - let backtrace = backtrace.to_string(); - - f.write_str("\n\nStack backtrace:\n")?; - f.write_str(backtrace.trim_end())?; - } - } - - Ok(()) - } -} - -impl Report<Box<dyn Error>> { - fn backtrace(&self) -> Option<&Backtrace> { - // have to grab the backtrace on the first error directly since that error may not be - // 'static - let backtrace = self.error.request_ref(); - let backtrace = backtrace.or_else(|| { - self.error - .source() - .map(|source| source.chain().find_map(|source| source.request_ref())) - .flatten() - }); - backtrace - } - - /// Format the report as a single line. - #[unstable(feature = "error_reporter", issue = "90172")] - fn fmt_singleline(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { - write!(f, "{}", self.error)?; - - let sources = self.error.source().into_iter().flat_map(<dyn Error>::chain); - - for cause in sources { - write!(f, ": {cause}")?; - } - - Ok(()) - } - - /// Format the report as multiple lines, with each error cause on its own line. - #[unstable(feature = "error_reporter", issue = "90172")] - fn fmt_multiline(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { - let error = &self.error; - - write!(f, "{error}")?; - - if let Some(cause) = error.source() { - write!(f, "\n\nCaused by:")?; - - let multiple = cause.source().is_some(); - - for (ind, error) in cause.chain().enumerate() { + for (ind, error) in cause.sources().enumerate() { writeln!(f)?; let mut indented = Indented { inner: f }; if multiple { @@ -1683,17 +1554,6 @@ where } #[unstable(feature = "error_reporter", issue = "90172")] -impl<'a, E> From<E> for Report<Box<dyn Error + 'a>> -where - E: Error + 'a, -{ - fn from(error: E) -> Self { - let error = box error; - Report { error, show_backtrace: false, pretty: false } - } -} - -#[unstable(feature = "error_reporter", issue = "90172")] impl<E> fmt::Display for Report<E> where E: Error, @@ -1703,13 +1563,6 @@ where } } -#[unstable(feature = "error_reporter", issue = "90172")] -impl fmt::Display for Report<Box<dyn Error>> { - fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { - if self.pretty { self.fmt_multiline(f) } else { self.fmt_singleline(f) } - } -} - // This type intentionally outputs the same format for `Display` and `Debug`for // situations where you unwrap a `Report` or return it from main. #[unstable(feature = "error_reporter", issue = "90172")] diff --git a/library/std/src/f32.rs b/library/std/src/f32.rs index 933b52b4d..3dd5b1250 100644 --- a/library/std/src/f32.rs +++ b/library/std/src/f32.rs @@ -1,4 +1,4 @@ -//! Constants specific to the `f32` single-precision floating point type. +//! Constants for the `f32` single-precision floating point type. //! //! *[See also the `f32` primitive type](primitive@f32).* //! diff --git a/library/std/src/f32/tests.rs b/library/std/src/f32/tests.rs index 69fa203ff..4ec16c84a 100644 --- a/library/std/src/f32/tests.rs +++ b/library/std/src/f32/tests.rs @@ -299,6 +299,84 @@ fn test_is_sign_negative() { assert!((-f32::NAN).is_sign_negative()); } +#[allow(unused_macros)] +macro_rules! assert_f32_biteq { + ($left : expr, $right : expr) => { + let l: &f32 = &$left; + let r: &f32 = &$right; + let lb = l.to_bits(); + let rb = r.to_bits(); + assert_eq!(lb, rb, "float {} ({:#x}) is not equal to {} ({:#x})", *l, lb, *r, rb); + }; +} + +// Ignore test on x87 floating point, these platforms do not guarantee NaN +// payloads are preserved and flush denormals to zero, failing the tests. +#[cfg(not(target_arch = "x86"))] +#[test] +fn test_next_up() { + let tiny = f32::from_bits(1); + let tiny_up = f32::from_bits(2); + let max_down = f32::from_bits(0x7f7f_fffe); + let largest_subnormal = f32::from_bits(0x007f_ffff); + let smallest_normal = f32::from_bits(0x0080_0000); + assert_f32_biteq!(f32::NEG_INFINITY.next_up(), f32::MIN); + assert_f32_biteq!(f32::MIN.next_up(), -max_down); + assert_f32_biteq!((-1.0 - f32::EPSILON).next_up(), -1.0); + assert_f32_biteq!((-smallest_normal).next_up(), -largest_subnormal); + assert_f32_biteq!((-tiny_up).next_up(), -tiny); + assert_f32_biteq!((-tiny).next_up(), -0.0f32); + assert_f32_biteq!((-0.0f32).next_up(), tiny); + assert_f32_biteq!(0.0f32.next_up(), tiny); + assert_f32_biteq!(tiny.next_up(), tiny_up); + assert_f32_biteq!(largest_subnormal.next_up(), smallest_normal); + assert_f32_biteq!(1.0f32.next_up(), 1.0 + f32::EPSILON); + assert_f32_biteq!(f32::MAX.next_up(), f32::INFINITY); + assert_f32_biteq!(f32::INFINITY.next_up(), f32::INFINITY); + + // Check that NaNs roundtrip. + let nan0 = f32::NAN; + let nan1 = f32::from_bits(f32::NAN.to_bits() ^ 0x002a_aaaa); + let nan2 = f32::from_bits(f32::NAN.to_bits() ^ 0x0055_5555); + assert_f32_biteq!(nan0.next_up(), nan0); + assert_f32_biteq!(nan1.next_up(), nan1); + assert_f32_biteq!(nan2.next_up(), nan2); +} + +// Ignore test on x87 floating point, these platforms do not guarantee NaN +// payloads are preserved and flush denormals to zero, failing the tests. +#[cfg(not(target_arch = "x86"))] +#[test] +fn test_next_down() { + let tiny = f32::from_bits(1); + let tiny_up = f32::from_bits(2); + let max_down = f32::from_bits(0x7f7f_fffe); + let largest_subnormal = f32::from_bits(0x007f_ffff); + let smallest_normal = f32::from_bits(0x0080_0000); + assert_f32_biteq!(f32::NEG_INFINITY.next_down(), f32::NEG_INFINITY); + assert_f32_biteq!(f32::MIN.next_down(), f32::NEG_INFINITY); + assert_f32_biteq!((-max_down).next_down(), f32::MIN); + assert_f32_biteq!((-1.0f32).next_down(), -1.0 - f32::EPSILON); + assert_f32_biteq!((-largest_subnormal).next_down(), -smallest_normal); + assert_f32_biteq!((-tiny).next_down(), -tiny_up); + assert_f32_biteq!((-0.0f32).next_down(), -tiny); + assert_f32_biteq!((0.0f32).next_down(), -tiny); + assert_f32_biteq!(tiny.next_down(), 0.0f32); + assert_f32_biteq!(tiny_up.next_down(), tiny); + assert_f32_biteq!(smallest_normal.next_down(), largest_subnormal); + assert_f32_biteq!((1.0 + f32::EPSILON).next_down(), 1.0f32); + assert_f32_biteq!(f32::MAX.next_down(), max_down); + assert_f32_biteq!(f32::INFINITY.next_down(), f32::MAX); + + // Check that NaNs roundtrip. + let nan0 = f32::NAN; + let nan1 = f32::from_bits(f32::NAN.to_bits() ^ 0x002a_aaaa); + let nan2 = f32::from_bits(f32::NAN.to_bits() ^ 0x0055_5555); + assert_f32_biteq!(nan0.next_down(), nan0); + assert_f32_biteq!(nan1.next_down(), nan1); + assert_f32_biteq!(nan2.next_down(), nan2); +} + #[test] fn test_mul_add() { let nan: f32 = f32::NAN; diff --git a/library/std/src/f64.rs b/library/std/src/f64.rs index a9aa84f70..31351a879 100644 --- a/library/std/src/f64.rs +++ b/library/std/src/f64.rs @@ -1,4 +1,4 @@ -//! Constants specific to the `f64` double-precision floating point type. +//! Constants for the `f64` double-precision floating point type. //! //! *[See also the `f64` primitive type](primitive@f64).* //! diff --git a/library/std/src/f64/tests.rs b/library/std/src/f64/tests.rs index 5c163cfe9..12baa68f4 100644 --- a/library/std/src/f64/tests.rs +++ b/library/std/src/f64/tests.rs @@ -289,6 +289,82 @@ fn test_is_sign_negative() { assert!((-f64::NAN).is_sign_negative()); } +#[allow(unused_macros)] +macro_rules! assert_f64_biteq { + ($left : expr, $right : expr) => { + let l: &f64 = &$left; + let r: &f64 = &$right; + let lb = l.to_bits(); + let rb = r.to_bits(); + assert_eq!(lb, rb, "float {} ({:#x}) is not equal to {} ({:#x})", *l, lb, *r, rb); + }; +} + +// Ignore test on x87 floating point, these platforms do not guarantee NaN +// payloads are preserved and flush denormals to zero, failing the tests. +#[cfg(not(target_arch = "x86"))] +#[test] +fn test_next_up() { + let tiny = f64::from_bits(1); + let tiny_up = f64::from_bits(2); + let max_down = f64::from_bits(0x7fef_ffff_ffff_fffe); + let largest_subnormal = f64::from_bits(0x000f_ffff_ffff_ffff); + let smallest_normal = f64::from_bits(0x0010_0000_0000_0000); + assert_f64_biteq!(f64::NEG_INFINITY.next_up(), f64::MIN); + assert_f64_biteq!(f64::MIN.next_up(), -max_down); + assert_f64_biteq!((-1.0 - f64::EPSILON).next_up(), -1.0); + assert_f64_biteq!((-smallest_normal).next_up(), -largest_subnormal); + assert_f64_biteq!((-tiny_up).next_up(), -tiny); + assert_f64_biteq!((-tiny).next_up(), -0.0f64); + assert_f64_biteq!((-0.0f64).next_up(), tiny); + assert_f64_biteq!(0.0f64.next_up(), tiny); + assert_f64_biteq!(tiny.next_up(), tiny_up); + assert_f64_biteq!(largest_subnormal.next_up(), smallest_normal); + assert_f64_biteq!(1.0f64.next_up(), 1.0 + f64::EPSILON); + assert_f64_biteq!(f64::MAX.next_up(), f64::INFINITY); + assert_f64_biteq!(f64::INFINITY.next_up(), f64::INFINITY); + + let nan0 = f64::NAN; + let nan1 = f64::from_bits(f64::NAN.to_bits() ^ 0x000a_aaaa_aaaa_aaaa); + let nan2 = f64::from_bits(f64::NAN.to_bits() ^ 0x0005_5555_5555_5555); + assert_f64_biteq!(nan0.next_up(), nan0); + assert_f64_biteq!(nan1.next_up(), nan1); + assert_f64_biteq!(nan2.next_up(), nan2); +} + +// Ignore test on x87 floating point, these platforms do not guarantee NaN +// payloads are preserved and flush denormals to zero, failing the tests. +#[cfg(not(target_arch = "x86"))] +#[test] +fn test_next_down() { + let tiny = f64::from_bits(1); + let tiny_up = f64::from_bits(2); + let max_down = f64::from_bits(0x7fef_ffff_ffff_fffe); + let largest_subnormal = f64::from_bits(0x000f_ffff_ffff_ffff); + let smallest_normal = f64::from_bits(0x0010_0000_0000_0000); + assert_f64_biteq!(f64::NEG_INFINITY.next_down(), f64::NEG_INFINITY); + assert_f64_biteq!(f64::MIN.next_down(), f64::NEG_INFINITY); + assert_f64_biteq!((-max_down).next_down(), f64::MIN); + assert_f64_biteq!((-1.0f64).next_down(), -1.0 - f64::EPSILON); + assert_f64_biteq!((-largest_subnormal).next_down(), -smallest_normal); + assert_f64_biteq!((-tiny).next_down(), -tiny_up); + assert_f64_biteq!((-0.0f64).next_down(), -tiny); + assert_f64_biteq!((0.0f64).next_down(), -tiny); + assert_f64_biteq!(tiny.next_down(), 0.0f64); + assert_f64_biteq!(tiny_up.next_down(), tiny); + assert_f64_biteq!(smallest_normal.next_down(), largest_subnormal); + assert_f64_biteq!((1.0 + f64::EPSILON).next_down(), 1.0f64); + assert_f64_biteq!(f64::MAX.next_down(), max_down); + assert_f64_biteq!(f64::INFINITY.next_down(), f64::MAX); + + let nan0 = f64::NAN; + let nan1 = f64::from_bits(f64::NAN.to_bits() ^ 0x000a_aaaa_aaaa_aaaa); + let nan2 = f64::from_bits(f64::NAN.to_bits() ^ 0x0005_5555_5555_5555); + assert_f64_biteq!(nan0.next_down(), nan0); + assert_f64_biteq!(nan1.next_down(), nan1); + assert_f64_biteq!(nan2.next_down(), nan2); +} + #[test] fn test_mul_add() { let nan: f64 = f64::NAN; diff --git a/library/std/src/ffi/os_str.rs b/library/std/src/ffi/os_str.rs index a0a5c003d..80ed34157 100644 --- a/library/std/src/ffi/os_str.rs +++ b/library/std/src/ffi/os_str.rs @@ -290,7 +290,8 @@ impl OsString { /// in the given `OsString`. The string may reserve more space to speculatively avoid /// frequent reallocations. After calling `try_reserve`, capacity will be /// greater than or equal to `self.len() + additional` if it returns `Ok(())`. - /// Does nothing if capacity is already sufficient. + /// Does nothing if capacity is already sufficient. This method preserves + /// the contents even if an error occurs. /// /// See the main `OsString` documentation information about encoding and capacity units. /// diff --git a/library/std/src/fs.rs b/library/std/src/fs.rs index c8e131b6e..c6c78dc39 100644 --- a/library/std/src/fs.rs +++ b/library/std/src/fs.rs @@ -13,13 +13,13 @@ mod tests; use crate::ffi::OsString; use crate::fmt; -use crate::io::{self, IoSlice, IoSliceMut, Read, ReadBuf, Seek, SeekFrom, Write}; +use crate::io::{self, BorrowedCursor, IoSlice, IoSliceMut, Read, Seek, SeekFrom, Write}; use crate::path::{Path, PathBuf}; use crate::sys::fs as fs_imp; use crate::sys_common::{AsInner, AsInnerMut, FromInner, IntoInner}; use crate::time::SystemTime; -/// A reference to an open file on the filesystem. +/// An object providing access to an open file on the filesystem. /// /// An instance of a `File` can be read and/or written depending on what options /// it was opened with. Files also implement [`Seek`] to alter the logical cursor @@ -377,6 +377,35 @@ impl File { OpenOptions::new().write(true).create(true).truncate(true).open(path.as_ref()) } + /// Creates a new file in read-write mode; error if the file exists. + /// + /// This function will create a file if it does not exist, or return an error if it does. This + /// way, if the call succeeds, the file returned is guaranteed to be new. + /// + /// This option is useful because it is atomic. Otherwise between checking whether a file + /// exists and creating a new one, the file may have been created by another process (a TOCTOU + /// race condition / attack). + /// + /// This can also be written using + /// `File::options().read(true).write(true).create_new(true).open(...)`. + /// + /// # Examples + /// + /// ```no_run + /// #![feature(file_create_new)] + /// + /// use std::fs::File; + /// + /// fn main() -> std::io::Result<()> { + /// let mut f = File::create_new("foo.txt")?; + /// Ok(()) + /// } + /// ``` + #[unstable(feature = "file_create_new", issue = "none")] + pub fn create_new<P: AsRef<Path>>(path: P) -> io::Result<File> { + OpenOptions::new().read(true).write(true).create_new(true).open(path.as_ref()) + } + /// Returns a new OpenOptions object. /// /// This function returns a new OpenOptions object that you can use to @@ -703,8 +732,8 @@ impl Read for File { self.inner.read_vectored(bufs) } - fn read_buf(&mut self, buf: &mut ReadBuf<'_>) -> io::Result<()> { - self.inner.read_buf(buf) + fn read_buf(&mut self, cursor: BorrowedCursor<'_>) -> io::Result<()> { + self.inner.read_buf(cursor) } #[inline] @@ -755,8 +784,8 @@ impl Read for &File { self.inner.read(buf) } - fn read_buf(&mut self, buf: &mut ReadBuf<'_>) -> io::Result<()> { - self.inner.read_buf(buf) + fn read_buf(&mut self, cursor: BorrowedCursor<'_>) -> io::Result<()> { + self.inner.read_buf(cursor) } fn read_vectored(&mut self, bufs: &mut [IoSliceMut<'_>]) -> io::Result<usize> { diff --git a/library/std/src/io/buffered/bufreader.rs b/library/std/src/io/buffered/bufreader.rs index f7fbaa9c2..4f339a18a 100644 --- a/library/std/src/io/buffered/bufreader.rs +++ b/library/std/src/io/buffered/bufreader.rs @@ -2,7 +2,7 @@ mod buffer; use crate::fmt; use crate::io::{ - self, BufRead, IoSliceMut, Read, ReadBuf, Seek, SeekFrom, SizeHint, DEFAULT_BUF_SIZE, + self, BorrowedCursor, BufRead, IoSliceMut, Read, Seek, SeekFrom, SizeHint, DEFAULT_BUF_SIZE, }; use buffer::Buffer; @@ -224,6 +224,14 @@ impl<R> BufReader<R> { } } +// This is only used by a test which asserts that the initialization-tracking is correct. +#[cfg(test)] +impl<R> BufReader<R> { + pub fn initialized(&self) -> usize { + self.buf.initialized() + } +} + impl<R: Seek> BufReader<R> { /// Seeks relative to the current position. If the new position lies within the buffer, /// the buffer will not be flushed, allowing for more efficient seeks. @@ -266,21 +274,21 @@ impl<R: Read> Read for BufReader<R> { Ok(nread) } - fn read_buf(&mut self, buf: &mut ReadBuf<'_>) -> io::Result<()> { + fn read_buf(&mut self, mut cursor: BorrowedCursor<'_>) -> io::Result<()> { // If we don't have any buffered data and we're doing a massive read // (larger than our internal buffer), bypass our internal buffer // entirely. - if self.buf.pos() == self.buf.filled() && buf.remaining() >= self.capacity() { + if self.buf.pos() == self.buf.filled() && cursor.capacity() >= self.capacity() { self.discard_buffer(); - return self.inner.read_buf(buf); + return self.inner.read_buf(cursor); } - let prev = buf.filled_len(); + let prev = cursor.written(); let mut rem = self.fill_buf()?; - rem.read_buf(buf)?; + rem.read_buf(cursor.reborrow())?; - self.consume(buf.filled_len() - prev); //slice impl of read_buf known to never unfill buf + self.consume(cursor.written() - prev); //slice impl of read_buf known to never unfill buf Ok(()) } diff --git a/library/std/src/io/buffered/bufreader/buffer.rs b/library/std/src/io/buffered/bufreader/buffer.rs index 8ae01f3b0..e9e29d60c 100644 --- a/library/std/src/io/buffered/bufreader/buffer.rs +++ b/library/std/src/io/buffered/bufreader/buffer.rs @@ -9,7 +9,7 @@ /// that user code which wants to do reads from a `BufReader` via `buffer` + `consume` can do so /// without encountering any runtime bounds checks. use crate::cmp; -use crate::io::{self, Read, ReadBuf}; +use crate::io::{self, BorrowedBuf, Read}; use crate::mem::MaybeUninit; pub struct Buffer { @@ -20,13 +20,19 @@ pub struct Buffer { // Each call to `fill_buf` sets `filled` to indicate how many bytes at the start of `buf` are // initialized with bytes from a read. filled: usize, + // This is the max number of bytes returned across all `fill_buf` calls. We track this so that we + // can accurately tell `read_buf` how many bytes of buf are initialized, to bypass as much of its + // defensive initialization as possible. Note that while this often the same as `filled`, it + // doesn't need to be. Calls to `fill_buf` are not required to actually fill the buffer, and + // omitting this is a huge perf regression for `Read` impls that do not. + initialized: usize, } impl Buffer { #[inline] pub fn with_capacity(capacity: usize) -> Self { let buf = Box::new_uninit_slice(capacity); - Self { buf, pos: 0, filled: 0 } + Self { buf, pos: 0, filled: 0, initialized: 0 } } #[inline] @@ -51,6 +57,12 @@ impl Buffer { self.pos } + // This is only used by a test which asserts that the initialization-tracking is correct. + #[cfg(test)] + pub fn initialized(&self) -> usize { + self.initialized + } + #[inline] pub fn discard_buffer(&mut self) { self.pos = 0; @@ -93,12 +105,17 @@ impl Buffer { if self.pos >= self.filled { debug_assert!(self.pos == self.filled); - let mut readbuf = ReadBuf::uninit(&mut self.buf); + let mut buf = BorrowedBuf::from(&mut *self.buf); + // SAFETY: `self.filled` bytes will always have been initialized. + unsafe { + buf.set_init(self.initialized); + } - reader.read_buf(&mut readbuf)?; + reader.read_buf(buf.unfilled())?; - self.filled = readbuf.filled_len(); self.pos = 0; + self.filled = buf.len(); + self.initialized = buf.init_len(); } Ok(self.buffer()) } diff --git a/library/std/src/io/buffered/tests.rs b/library/std/src/io/buffered/tests.rs index fe45b1326..f4e688eb9 100644 --- a/library/std/src/io/buffered/tests.rs +++ b/library/std/src/io/buffered/tests.rs @@ -1,5 +1,7 @@ use crate::io::prelude::*; -use crate::io::{self, BufReader, BufWriter, ErrorKind, IoSlice, LineWriter, ReadBuf, SeekFrom}; +use crate::io::{ + self, BorrowedBuf, BufReader, BufWriter, ErrorKind, IoSlice, LineWriter, SeekFrom, +}; use crate::mem::MaybeUninit; use crate::panic; use crate::sync::atomic::{AtomicUsize, Ordering}; @@ -61,48 +63,48 @@ fn test_buffered_reader_read_buf() { let inner: &[u8] = &[5, 6, 7, 0, 1, 2, 3, 4]; let mut reader = BufReader::with_capacity(2, inner); - let mut buf = [MaybeUninit::uninit(); 3]; - let mut buf = ReadBuf::uninit(&mut buf); + let buf: &mut [_] = &mut [MaybeUninit::uninit(); 3]; + let mut buf: BorrowedBuf<'_> = buf.into(); - reader.read_buf(&mut buf).unwrap(); + reader.read_buf(buf.unfilled()).unwrap(); assert_eq!(buf.filled(), [5, 6, 7]); assert_eq!(reader.buffer(), []); - let mut buf = [MaybeUninit::uninit(); 2]; - let mut buf = ReadBuf::uninit(&mut buf); + let buf: &mut [_] = &mut [MaybeUninit::uninit(); 2]; + let mut buf: BorrowedBuf<'_> = buf.into(); - reader.read_buf(&mut buf).unwrap(); + reader.read_buf(buf.unfilled()).unwrap(); assert_eq!(buf.filled(), [0, 1]); assert_eq!(reader.buffer(), []); - let mut buf = [MaybeUninit::uninit(); 1]; - let mut buf = ReadBuf::uninit(&mut buf); + let buf: &mut [_] = &mut [MaybeUninit::uninit(); 1]; + let mut buf: BorrowedBuf<'_> = buf.into(); - reader.read_buf(&mut buf).unwrap(); + reader.read_buf(buf.unfilled()).unwrap(); assert_eq!(buf.filled(), [2]); assert_eq!(reader.buffer(), [3]); - let mut buf = [MaybeUninit::uninit(); 3]; - let mut buf = ReadBuf::uninit(&mut buf); + let buf: &mut [_] = &mut [MaybeUninit::uninit(); 3]; + let mut buf: BorrowedBuf<'_> = buf.into(); - reader.read_buf(&mut buf).unwrap(); + reader.read_buf(buf.unfilled()).unwrap(); assert_eq!(buf.filled(), [3]); assert_eq!(reader.buffer(), []); - reader.read_buf(&mut buf).unwrap(); + reader.read_buf(buf.unfilled()).unwrap(); assert_eq!(buf.filled(), [3, 4]); assert_eq!(reader.buffer(), []); buf.clear(); - reader.read_buf(&mut buf).unwrap(); + reader.read_buf(buf.unfilled()).unwrap(); - assert_eq!(buf.filled_len(), 0); + assert!(buf.filled().is_empty()); } #[test] @@ -1037,3 +1039,27 @@ fn single_formatted_write() { writeln!(&mut writer, "{}, {}!", "hello", "world").unwrap(); assert_eq!(writer.get_ref().events, [RecordedEvent::Write("hello, world!\n".to_string())]); } + +#[test] +fn bufreader_full_initialize() { + struct OneByteReader; + impl Read for OneByteReader { + fn read(&mut self, buf: &mut [u8]) -> crate::io::Result<usize> { + if buf.len() > 0 { + buf[0] = 0; + Ok(1) + } else { + Ok(0) + } + } + } + let mut reader = BufReader::new(OneByteReader); + // Nothing is initialized yet. + assert_eq!(reader.initialized(), 0); + + let buf = reader.fill_buf().unwrap(); + // We read one byte... + assert_eq!(buf.len(), 1); + // But we initialized the whole buffer! + assert_eq!(reader.initialized(), reader.capacity()); +} diff --git a/library/std/src/io/copy.rs b/library/std/src/io/copy.rs index 1a10245e4..38b98afff 100644 --- a/library/std/src/io/copy.rs +++ b/library/std/src/io/copy.rs @@ -1,4 +1,4 @@ -use super::{BufWriter, ErrorKind, Read, ReadBuf, Result, Write, DEFAULT_BUF_SIZE}; +use super::{BorrowedBuf, BufWriter, ErrorKind, Read, Result, Write, DEFAULT_BUF_SIZE}; use crate::mem::MaybeUninit; /// Copies the entire contents of a reader into a writer. @@ -97,37 +97,39 @@ impl<I: Write> BufferedCopySpec for BufWriter<I> { loop { let buf = writer.buffer_mut(); - let mut read_buf = ReadBuf::uninit(buf.spare_capacity_mut()); + let mut read_buf: BorrowedBuf<'_> = buf.spare_capacity_mut().into(); - // SAFETY: init is either 0 or the initialized_len of the previous iteration unsafe { - read_buf.assume_init(init); + // SAFETY: init is either 0 or the init_len from the previous iteration. + read_buf.set_init(init); } if read_buf.capacity() >= DEFAULT_BUF_SIZE { - match reader.read_buf(&mut read_buf) { + let mut cursor = read_buf.unfilled(); + match reader.read_buf(cursor.reborrow()) { Ok(()) => { - let bytes_read = read_buf.filled_len(); + let bytes_read = cursor.written(); if bytes_read == 0 { return Ok(len); } - init = read_buf.initialized_len() - bytes_read; + init = read_buf.init_len() - bytes_read; + len += bytes_read as u64; - // SAFETY: ReadBuf guarantees all of its filled bytes are init + // SAFETY: BorrowedBuf guarantees all of its filled bytes are init unsafe { buf.set_len(buf.len() + bytes_read) }; - len += bytes_read as u64; + // Read again if the buffer still has enough capacity, as BufWriter itself would do // This will occur if the reader returns short reads - continue; } - Err(ref e) if e.kind() == ErrorKind::Interrupted => continue, + Err(ref e) if e.kind() == ErrorKind::Interrupted => {} Err(e) => return Err(e), } + } else { + writer.flush_buf()?; + init = 0; } - - writer.flush_buf()?; } } } @@ -136,13 +138,13 @@ fn stack_buffer_copy<R: Read + ?Sized, W: Write + ?Sized>( reader: &mut R, writer: &mut W, ) -> Result<u64> { - let mut buf = [MaybeUninit::uninit(); DEFAULT_BUF_SIZE]; - let mut buf = ReadBuf::uninit(&mut buf); + let buf: &mut [_] = &mut [MaybeUninit::uninit(); DEFAULT_BUF_SIZE]; + let mut buf: BorrowedBuf<'_> = buf.into(); let mut len = 0; loop { - match reader.read_buf(&mut buf) { + match reader.read_buf(buf.unfilled()) { Ok(()) => {} Err(e) if e.kind() == ErrorKind::Interrupted => continue, Err(e) => return Err(e), diff --git a/library/std/src/io/cursor.rs b/library/std/src/io/cursor.rs index f3fbfc447..d98ab021c 100644 --- a/library/std/src/io/cursor.rs +++ b/library/std/src/io/cursor.rs @@ -5,7 +5,7 @@ use crate::io::prelude::*; use crate::alloc::Allocator; use crate::cmp; -use crate::io::{self, ErrorKind, IoSlice, IoSliceMut, ReadBuf, SeekFrom}; +use crate::io::{self, BorrowedCursor, ErrorKind, IoSlice, IoSliceMut, SeekFrom}; /// A `Cursor` wraps an in-memory buffer and provides it with a /// [`Seek`] implementation. @@ -323,12 +323,12 @@ where Ok(n) } - fn read_buf(&mut self, buf: &mut ReadBuf<'_>) -> io::Result<()> { - let prev_filled = buf.filled_len(); + fn read_buf(&mut self, mut cursor: BorrowedCursor<'_>) -> io::Result<()> { + let prev_written = cursor.written(); - Read::read_buf(&mut self.fill_buf()?, buf)?; + Read::read_buf(&mut self.fill_buf()?, cursor.reborrow())?; - self.pos += (buf.filled_len() - prev_filled) as u64; + self.pos += (cursor.written() - prev_written) as u64; Ok(()) } diff --git a/library/std/src/io/error.rs b/library/std/src/io/error.rs index ff7fdcae1..29b09fcc5 100644 --- a/library/std/src/io/error.rs +++ b/library/std/src/io/error.rs @@ -76,6 +76,15 @@ impl fmt::Debug for Error { } } +#[cfg(not(bootstrap))] +#[stable(feature = "rust1", since = "1.0.0")] +impl From<alloc::ffi::NulError> for Error { + /// Converts a [`alloc::ffi::NulError`] into a [`Error`]. + fn from(_: alloc::ffi::NulError) -> Error { + const_io_error!(ErrorKind::InvalidInput, "data provided contains a nul byte") + } +} + // Only derive debug in tests, to make sure it // doesn't accidentally get printed. #[cfg_attr(test, derive(Debug))] @@ -564,6 +573,8 @@ impl Error { /// println!("last OS error: {os_error:?}"); /// ``` #[stable(feature = "rust1", since = "1.0.0")] + #[doc(alias = "GetLastError")] + #[doc(alias = "errno")] #[must_use] #[inline] pub fn last_os_error() -> Error { diff --git a/library/std/src/io/error/repr_bitpacked.rs b/library/std/src/io/error/repr_bitpacked.rs index 292bf4826..781ae03ad 100644 --- a/library/std/src/io/error/repr_bitpacked.rs +++ b/library/std/src/io/error/repr_bitpacked.rs @@ -269,10 +269,10 @@ where } TAG_SIMPLE_MESSAGE => ErrorData::SimpleMessage(&*ptr.cast::<SimpleMessage>().as_ptr()), TAG_CUSTOM => { - // It would be correct for us to use `ptr::sub` here (see the + // It would be correct for us to use `ptr::byte_sub` here (see the // comment above the `wrapping_add` call in `new_custom` for why), // but it isn't clear that it makes a difference, so we don't. - let custom = ptr.as_ptr().cast::<u8>().wrapping_sub(TAG_CUSTOM).cast::<Custom>(); + let custom = ptr.as_ptr().wrapping_byte_sub(TAG_CUSTOM).cast::<Custom>(); ErrorData::Custom(make_custom(custom)) } _ => { diff --git a/library/std/src/io/impls.rs b/library/std/src/io/impls.rs index 950725473..e5048dcc8 100644 --- a/library/std/src/io/impls.rs +++ b/library/std/src/io/impls.rs @@ -6,7 +6,7 @@ use crate::cmp; use crate::collections::VecDeque; use crate::fmt; use crate::io::{ - self, BufRead, ErrorKind, IoSlice, IoSliceMut, Read, ReadBuf, Seek, SeekFrom, Write, + self, BorrowedCursor, BufRead, ErrorKind, IoSlice, IoSliceMut, Read, Seek, SeekFrom, Write, }; use crate::mem; @@ -21,8 +21,8 @@ impl<R: Read + ?Sized> Read for &mut R { } #[inline] - fn read_buf(&mut self, buf: &mut ReadBuf<'_>) -> io::Result<()> { - (**self).read_buf(buf) + fn read_buf(&mut self, cursor: BorrowedCursor<'_>) -> io::Result<()> { + (**self).read_buf(cursor) } #[inline] @@ -125,8 +125,8 @@ impl<R: Read + ?Sized> Read for Box<R> { } #[inline] - fn read_buf(&mut self, buf: &mut ReadBuf<'_>) -> io::Result<()> { - (**self).read_buf(buf) + fn read_buf(&mut self, cursor: BorrowedCursor<'_>) -> io::Result<()> { + (**self).read_buf(cursor) } #[inline] @@ -249,11 +249,11 @@ impl Read for &[u8] { } #[inline] - fn read_buf(&mut self, buf: &mut ReadBuf<'_>) -> io::Result<()> { - let amt = cmp::min(buf.remaining(), self.len()); + fn read_buf(&mut self, mut cursor: BorrowedCursor<'_>) -> io::Result<()> { + let amt = cmp::min(cursor.capacity(), self.len()); let (a, b) = self.split_at(amt); - buf.append(a); + cursor.append(a); *self = b; Ok(()) @@ -427,10 +427,10 @@ impl<A: Allocator> Read for VecDeque<u8, A> { } #[inline] - fn read_buf(&mut self, buf: &mut ReadBuf<'_>) -> io::Result<()> { + fn read_buf(&mut self, cursor: BorrowedCursor<'_>) -> io::Result<()> { let (ref mut front, _) = self.as_slices(); - let n = cmp::min(buf.remaining(), front.len()); - Read::read_buf(front, buf)?; + let n = cmp::min(cursor.capacity(), front.len()); + Read::read_buf(front, cursor)?; self.drain(..n); Ok(()) } diff --git a/library/std/src/io/mod.rs b/library/std/src/io/mod.rs index 96addbd1a..eeace2c43 100644 --- a/library/std/src/io/mod.rs +++ b/library/std/src/io/mod.rs @@ -278,7 +278,7 @@ pub use self::{ }; #[unstable(feature = "read_buf", issue = "78485")] -pub use self::readbuf::ReadBuf; +pub use self::readbuf::{BorrowedBuf, BorrowedCursor}; pub(crate) use error::const_io_error; mod buffered; @@ -362,29 +362,30 @@ pub(crate) fn default_read_to_end<R: Read + ?Sized>(r: &mut R, buf: &mut Vec<u8> buf.reserve(32); // buf is full, need more space } - let mut read_buf = ReadBuf::uninit(buf.spare_capacity_mut()); + let mut read_buf: BorrowedBuf<'_> = buf.spare_capacity_mut().into(); // SAFETY: These bytes were initialized but not filled in the previous loop unsafe { - read_buf.assume_init(initialized); + read_buf.set_init(initialized); } - match r.read_buf(&mut read_buf) { + let mut cursor = read_buf.unfilled(); + match r.read_buf(cursor.reborrow()) { Ok(()) => {} Err(e) if e.kind() == ErrorKind::Interrupted => continue, Err(e) => return Err(e), } - if read_buf.filled_len() == 0 { + if cursor.written() == 0 { return Ok(buf.len() - start_len); } // store how much was initialized but not filled - initialized = read_buf.initialized_len() - read_buf.filled_len(); - let new_len = read_buf.filled_len() + buf.len(); + initialized = cursor.init_ref().len(); - // SAFETY: ReadBuf's invariants mean this much memory is init + // SAFETY: BorrowedBuf's invariants mean this much memory is initialized. unsafe { + let new_len = read_buf.filled().len() + buf.len(); buf.set_len(new_len); } @@ -461,12 +462,15 @@ pub(crate) fn default_read_exact<R: Read + ?Sized>(this: &mut R, mut buf: &mut [ } } -pub(crate) fn default_read_buf<F>(read: F, buf: &mut ReadBuf<'_>) -> Result<()> +pub(crate) fn default_read_buf<F>(read: F, mut cursor: BorrowedCursor<'_>) -> Result<()> where F: FnOnce(&mut [u8]) -> Result<usize>, { - let n = read(buf.initialize_unfilled())?; - buf.add_filled(n); + let n = read(cursor.ensure_init().init_mut())?; + unsafe { + // SAFETY: we initialised using `ensure_init` so there is no uninit data to advance to. + cursor.advance(n); + } Ok(()) } @@ -803,30 +807,30 @@ pub trait Read { /// Pull some bytes from this source into the specified buffer. /// - /// This is equivalent to the [`read`](Read::read) method, except that it is passed a [`ReadBuf`] rather than `[u8]` to allow use + /// This is equivalent to the [`read`](Read::read) method, except that it is passed a [`BorrowedCursor`] rather than `[u8]` to allow use /// with uninitialized buffers. The new data will be appended to any existing contents of `buf`. /// /// The default implementation delegates to `read`. #[unstable(feature = "read_buf", issue = "78485")] - fn read_buf(&mut self, buf: &mut ReadBuf<'_>) -> Result<()> { + fn read_buf(&mut self, buf: BorrowedCursor<'_>) -> Result<()> { default_read_buf(|b| self.read(b), buf) } - /// Read the exact number of bytes required to fill `buf`. + /// Read the exact number of bytes required to fill `cursor`. /// - /// This is equivalent to the [`read_exact`](Read::read_exact) method, except that it is passed a [`ReadBuf`] rather than `[u8]` to + /// This is equivalent to the [`read_exact`](Read::read_exact) method, except that it is passed a [`BorrowedCursor`] rather than `[u8]` to /// allow use with uninitialized buffers. #[unstable(feature = "read_buf", issue = "78485")] - fn read_buf_exact(&mut self, buf: &mut ReadBuf<'_>) -> Result<()> { - while buf.remaining() > 0 { - let prev_filled = buf.filled().len(); - match self.read_buf(buf) { + fn read_buf_exact(&mut self, mut cursor: BorrowedCursor<'_>) -> Result<()> { + while cursor.capacity() > 0 { + let prev_written = cursor.written(); + match self.read_buf(cursor.reborrow()) { Ok(()) => {} Err(e) if e.kind() == ErrorKind::Interrupted => continue, Err(e) => return Err(e), } - if buf.filled().len() == prev_filled { + if cursor.written() == prev_written { return Err(Error::new(ErrorKind::UnexpectedEof, "failed to fill buffer")); } } @@ -883,6 +887,10 @@ pub trait Read { /// The yielded item is [`Ok`] if a byte was successfully read and [`Err`] /// otherwise. EOF is mapped to returning [`None`] from this iterator. /// + /// The default implementation calls `read` for each byte, + /// which can be very inefficient for data that's not in memory, + /// such as [`File`]. Consider using a [`BufReader`] in such cases. + /// /// # Examples /// /// [`File`]s implement `Read`: @@ -895,10 +903,11 @@ pub trait Read { /// ```no_run /// use std::io; /// use std::io::prelude::*; + /// use std::io::BufReader; /// use std::fs::File; /// /// fn main() -> io::Result<()> { - /// let f = File::open("foo.txt")?; + /// let f = BufReader::new(File::open("foo.txt")?); /// /// for byte in f.bytes() { /// println!("{}", byte.unwrap()); @@ -1028,8 +1037,6 @@ pub trait Read { /// # Examples /// /// ```no_run -/// #![feature(io_read_to_string)] -/// /// # use std::io; /// fn main() -> io::Result<()> { /// let stdin = io::read_to_string(io::stdin())?; @@ -1038,7 +1045,7 @@ pub trait Read { /// Ok(()) /// } /// ``` -#[unstable(feature = "io_read_to_string", issue = "80218")] +#[stable(feature = "io_read_to_string", since = "1.65.0")] pub fn read_to_string<R: Read>(mut reader: R) -> Result<String> { let mut buf = String::new(); reader.read_to_string(&mut buf)?; @@ -2582,50 +2589,48 @@ impl<T: Read> Read for Take<T> { Ok(n) } - fn read_buf(&mut self, buf: &mut ReadBuf<'_>) -> Result<()> { + fn read_buf(&mut self, mut buf: BorrowedCursor<'_>) -> Result<()> { // Don't call into inner reader at all at EOF because it may still block if self.limit == 0 { return Ok(()); } - let prev_filled = buf.filled_len(); - - if self.limit <= buf.remaining() as u64 { + if self.limit <= buf.capacity() as u64 { // if we just use an as cast to convert, limit may wrap around on a 32 bit target let limit = cmp::min(self.limit, usize::MAX as u64) as usize; - let extra_init = cmp::min(limit as usize, buf.initialized_len() - buf.filled_len()); + let extra_init = cmp::min(limit as usize, buf.init_ref().len()); // SAFETY: no uninit data is written to ibuf - let ibuf = unsafe { &mut buf.unfilled_mut()[..limit] }; + let ibuf = unsafe { &mut buf.as_mut()[..limit] }; - let mut sliced_buf = ReadBuf::uninit(ibuf); + let mut sliced_buf: BorrowedBuf<'_> = ibuf.into(); // SAFETY: extra_init bytes of ibuf are known to be initialized unsafe { - sliced_buf.assume_init(extra_init); + sliced_buf.set_init(extra_init); } - self.inner.read_buf(&mut sliced_buf)?; + let mut cursor = sliced_buf.unfilled(); + self.inner.read_buf(cursor.reborrow())?; - let new_init = sliced_buf.initialized_len(); - let filled = sliced_buf.filled_len(); + let new_init = cursor.init_ref().len(); + let filled = sliced_buf.len(); - // sliced_buf / ibuf must drop here + // cursor / sliced_buf / ibuf must drop here - // SAFETY: new_init bytes of buf's unfilled buffer have been initialized unsafe { - buf.assume_init(new_init); + // SAFETY: filled bytes have been filled and therefore initialized + buf.advance(filled); + // SAFETY: new_init bytes of buf's unfilled buffer have been initialized + buf.set_init(new_init); } - buf.add_filled(filled); - self.limit -= filled as u64; } else { - self.inner.read_buf(buf)?; - - //inner may unfill - self.limit -= buf.filled_len().saturating_sub(prev_filled) as u64; + let written = buf.written(); + self.inner.read_buf(buf.reborrow())?; + self.limit -= (buf.written() - written) as u64; } Ok(()) diff --git a/library/std/src/io/readbuf.rs b/library/std/src/io/readbuf.rs index 78d1113f8..b1a84095f 100644 --- a/library/std/src/io/readbuf.rs +++ b/library/std/src/io/readbuf.rs @@ -5,9 +5,10 @@ mod tests; use crate::cmp; use crate::fmt::{self, Debug, Formatter}; -use crate::mem::MaybeUninit; +use crate::io::{Result, Write}; +use crate::mem::{self, MaybeUninit}; -/// A wrapper around a byte buffer that is incrementally filled and initialized. +/// A borrowed byte buffer which is incrementally filled and initialized. /// /// This type is a sort of "double cursor". It tracks three regions in the buffer: a region at the beginning of the /// buffer that has been logically filled with data, a region that has been initialized at some point but not yet @@ -20,230 +21,286 @@ use crate::mem::MaybeUninit; /// [ filled | unfilled ] /// [ initialized | uninitialized ] /// ``` -pub struct ReadBuf<'a> { - buf: &'a mut [MaybeUninit<u8>], +/// +/// A `BorrowedBuf` is created around some existing data (or capacity for data) via a unique reference +/// (`&mut`). The `BorrowedBuf` can be configured (e.g., using `clear` or `set_init`), but cannot be +/// directly written. To write into the buffer, use `unfilled` to create a `BorrowedCursor`. The cursor +/// has write-only access to the unfilled portion of the buffer (you can think of it as a +/// write-only iterator). +/// +/// The lifetime `'data` is a bound on the lifetime of the underlying data. +pub struct BorrowedBuf<'data> { + /// The buffer's underlying data. + buf: &'data mut [MaybeUninit<u8>], + /// The length of `self.buf` which is known to be filled. filled: usize, - initialized: usize, + /// The length of `self.buf` which is known to be initialized. + init: usize, } -impl Debug for ReadBuf<'_> { +impl Debug for BorrowedBuf<'_> { fn fmt(&self, f: &mut Formatter<'_>) -> fmt::Result { - f.debug_struct("ReadBuf") - .field("init", &self.initialized()) + f.debug_struct("BorrowedBuf") + .field("init", &self.init) .field("filled", &self.filled) .field("capacity", &self.capacity()) .finish() } } -impl<'a> ReadBuf<'a> { - /// Creates a new `ReadBuf` from a fully initialized buffer. +/// Create a new `BorrowedBuf` from a fully initialized slice. +impl<'data> From<&'data mut [u8]> for BorrowedBuf<'data> { #[inline] - pub fn new(buf: &'a mut [u8]) -> ReadBuf<'a> { - let len = buf.len(); + fn from(slice: &'data mut [u8]) -> BorrowedBuf<'data> { + let len = slice.len(); - ReadBuf { - //SAFETY: initialized data never becoming uninitialized is an invariant of ReadBuf - buf: unsafe { (buf as *mut [u8]).as_uninit_slice_mut().unwrap() }, + BorrowedBuf { + // SAFETY: initialized data never becoming uninitialized is an invariant of BorrowedBuf + buf: unsafe { (slice as *mut [u8]).as_uninit_slice_mut().unwrap() }, filled: 0, - initialized: len, + init: len, } } +} - /// Creates a new `ReadBuf` from a fully uninitialized buffer. - /// - /// Use `assume_init` if part of the buffer is known to be already initialized. +/// Create a new `BorrowedBuf` from an uninitialized buffer. +/// +/// Use `set_init` if part of the buffer is known to be already initialized. +impl<'data> From<&'data mut [MaybeUninit<u8>]> for BorrowedBuf<'data> { #[inline] - pub fn uninit(buf: &'a mut [MaybeUninit<u8>]) -> ReadBuf<'a> { - ReadBuf { buf, filled: 0, initialized: 0 } + fn from(buf: &'data mut [MaybeUninit<u8>]) -> BorrowedBuf<'data> { + BorrowedBuf { buf, filled: 0, init: 0 } } +} +impl<'data> BorrowedBuf<'data> { /// Returns the total capacity of the buffer. #[inline] pub fn capacity(&self) -> usize { self.buf.len() } + /// Returns the length of the filled part of the buffer. + #[inline] + pub fn len(&self) -> usize { + self.filled + } + + /// Returns the length of the initialized part of the buffer. + #[inline] + pub fn init_len(&self) -> usize { + self.init + } + /// Returns a shared reference to the filled portion of the buffer. #[inline] pub fn filled(&self) -> &[u8] { - //SAFETY: We only slice the filled part of the buffer, which is always valid + // SAFETY: We only slice the filled part of the buffer, which is always valid unsafe { MaybeUninit::slice_assume_init_ref(&self.buf[0..self.filled]) } } - /// Returns a mutable reference to the filled portion of the buffer. + /// Returns a cursor over the unfilled part of the buffer. #[inline] - pub fn filled_mut(&mut self) -> &mut [u8] { - //SAFETY: We only slice the filled part of the buffer, which is always valid - unsafe { MaybeUninit::slice_assume_init_mut(&mut self.buf[0..self.filled]) } + pub fn unfilled<'this>(&'this mut self) -> BorrowedCursor<'this> { + BorrowedCursor { + start: self.filled, + // SAFETY: we never assign into `BorrowedCursor::buf`, so treating its + // lifetime covariantly is safe. + buf: unsafe { + mem::transmute::<&'this mut BorrowedBuf<'data>, &'this mut BorrowedBuf<'this>>(self) + }, + } } - /// Returns a shared reference to the initialized portion of the buffer. + /// Clears the buffer, resetting the filled region to empty. /// - /// This includes the filled portion. + /// The number of initialized bytes is not changed, and the contents of the buffer are not modified. #[inline] - pub fn initialized(&self) -> &[u8] { - //SAFETY: We only slice the initialized part of the buffer, which is always valid - unsafe { MaybeUninit::slice_assume_init_ref(&self.buf[0..self.initialized]) } + pub fn clear(&mut self) -> &mut Self { + self.filled = 0; + self } - /// Returns a mutable reference to the initialized portion of the buffer. + /// Asserts that the first `n` bytes of the buffer are initialized. /// - /// This includes the filled portion. - #[inline] - pub fn initialized_mut(&mut self) -> &mut [u8] { - //SAFETY: We only slice the initialized part of the buffer, which is always valid - unsafe { MaybeUninit::slice_assume_init_mut(&mut self.buf[0..self.initialized]) } - } - - /// Returns a mutable reference to the unfilled part of the buffer without ensuring that it has been fully - /// initialized. + /// `BorrowedBuf` assumes that bytes are never de-initialized, so this method does nothing when called with fewer + /// bytes than are already known to be initialized. /// /// # Safety /// - /// The caller must not de-initialize portions of the buffer that have already been initialized. + /// The caller must ensure that the first `n` unfilled bytes of the buffer have already been initialized. #[inline] - pub unsafe fn unfilled_mut(&mut self) -> &mut [MaybeUninit<u8>] { - &mut self.buf[self.filled..] + pub unsafe fn set_init(&mut self, n: usize) -> &mut Self { + self.init = cmp::max(self.init, n); + self } +} - /// Returns a mutable reference to the uninitialized part of the buffer. +/// A writeable view of the unfilled portion of a [`BorrowedBuf`](BorrowedBuf). +/// +/// Provides access to the initialized and uninitialized parts of the underlying `BorrowedBuf`. +/// Data can be written directly to the cursor by using [`append`](BorrowedCursor::append) or +/// indirectly by getting a slice of part or all of the cursor and writing into the slice. In the +/// indirect case, the caller must call [`advance`](BorrowedCursor::advance) after writing to inform +/// the cursor how many bytes have been written. +/// +/// Once data is written to the cursor, it becomes part of the filled portion of the underlying +/// `BorrowedBuf` and can no longer be accessed or re-written by the cursor. I.e., the cursor tracks +/// the unfilled part of the underlying `BorrowedBuf`. +/// +/// The lifetime `'a` is a bound on the lifetime of the underlying buffer (which means it is a bound +/// on the data in that buffer by transitivity). +#[derive(Debug)] +pub struct BorrowedCursor<'a> { + /// The underlying buffer. + // Safety invariant: we treat the type of buf as covariant in the lifetime of `BorrowedBuf` when + // we create a `BorrowedCursor`. This is only safe if we never replace `buf` by assigning into + // it, so don't do that! + buf: &'a mut BorrowedBuf<'a>, + /// The length of the filled portion of the underlying buffer at the time of the cursor's + /// creation. + start: usize, +} + +impl<'a> BorrowedCursor<'a> { + /// Reborrow this cursor by cloning it with a smaller lifetime. /// - /// It is safe to uninitialize any of these bytes. + /// Since a cursor maintains unique access to its underlying buffer, the borrowed cursor is + /// not accessible while the new cursor exists. #[inline] - pub fn uninitialized_mut(&mut self) -> &mut [MaybeUninit<u8>] { - &mut self.buf[self.initialized..] + pub fn reborrow<'this>(&'this mut self) -> BorrowedCursor<'this> { + BorrowedCursor { + // SAFETY: we never assign into `BorrowedCursor::buf`, so treating its + // lifetime covariantly is safe. + buf: unsafe { + mem::transmute::<&'this mut BorrowedBuf<'a>, &'this mut BorrowedBuf<'this>>( + self.buf, + ) + }, + start: self.start, + } } - /// Returns a mutable reference to the unfilled part of the buffer, ensuring it is fully initialized. - /// - /// Since `ReadBuf` tracks the region of the buffer that has been initialized, this is effectively "free" after - /// the first use. + /// Returns the available space in the cursor. #[inline] - pub fn initialize_unfilled(&mut self) -> &mut [u8] { - // should optimize out the assertion - self.initialize_unfilled_to(self.remaining()) + pub fn capacity(&self) -> usize { + self.buf.capacity() - self.buf.filled } - /// Returns a mutable reference to the first `n` bytes of the unfilled part of the buffer, ensuring it is - /// fully initialized. - /// - /// # Panics + /// Returns the number of bytes written to this cursor since it was created from a `BorrowedBuf`. /// - /// Panics if `self.remaining()` is less than `n`. + /// Note that if this cursor is a reborrowed clone of another, then the count returned is the + /// count written via either cursor, not the count since the cursor was reborrowed. #[inline] - pub fn initialize_unfilled_to(&mut self, n: usize) -> &mut [u8] { - assert!(self.remaining() >= n); - - let extra_init = self.initialized - self.filled; - // If we don't have enough initialized, do zeroing - if n > extra_init { - let uninit = n - extra_init; - let unfilled = &mut self.uninitialized_mut()[0..uninit]; - - for byte in unfilled.iter_mut() { - byte.write(0); - } - - // SAFETY: we just initialized uninit bytes, and the previous bytes were already init - unsafe { - self.assume_init(n); - } - } - - let filled = self.filled; + pub fn written(&self) -> usize { + self.buf.filled - self.start + } - &mut self.initialized_mut()[filled..filled + n] + /// Returns a shared reference to the initialized portion of the cursor. + #[inline] + pub fn init_ref(&self) -> &[u8] { + // SAFETY: We only slice the initialized part of the buffer, which is always valid + unsafe { MaybeUninit::slice_assume_init_ref(&self.buf.buf[self.buf.filled..self.buf.init]) } } - /// Returns the number of bytes at the end of the slice that have not yet been filled. + /// Returns a mutable reference to the initialized portion of the cursor. #[inline] - pub fn remaining(&self) -> usize { - self.capacity() - self.filled + pub fn init_mut(&mut self) -> &mut [u8] { + // SAFETY: We only slice the initialized part of the buffer, which is always valid + unsafe { + MaybeUninit::slice_assume_init_mut(&mut self.buf.buf[self.buf.filled..self.buf.init]) + } } - /// Clears the buffer, resetting the filled region to empty. + /// Returns a mutable reference to the uninitialized part of the cursor. /// - /// The number of initialized bytes is not changed, and the contents of the buffer are not modified. + /// It is safe to uninitialize any of these bytes. #[inline] - pub fn clear(&mut self) -> &mut Self { - self.set_filled(0) // The assertion in `set_filled` is optimized out + pub fn uninit_mut(&mut self) -> &mut [MaybeUninit<u8>] { + &mut self.buf.buf[self.buf.init..] } - /// Increases the size of the filled region of the buffer. - /// - /// The number of initialized bytes is not changed. + /// Returns a mutable reference to the whole cursor. /// - /// # Panics + /// # Safety /// - /// Panics if the filled region of the buffer would become larger than the initialized region. + /// The caller must not uninitialize any bytes in the initialized portion of the cursor. #[inline] - pub fn add_filled(&mut self, n: usize) -> &mut Self { - self.set_filled(self.filled + n) + pub unsafe fn as_mut(&mut self) -> &mut [MaybeUninit<u8>] { + &mut self.buf.buf[self.buf.filled..] } - /// Sets the size of the filled region of the buffer. + /// Advance the cursor by asserting that `n` bytes have been filled. /// - /// The number of initialized bytes is not changed. + /// After advancing, the `n` bytes are no longer accessible via the cursor and can only be + /// accessed via the underlying buffer. I.e., the buffer's filled portion grows by `n` elements + /// and its unfilled portion (and the capacity of this cursor) shrinks by `n` elements. /// - /// Note that this can be used to *shrink* the filled region of the buffer in addition to growing it (for - /// example, by a `Read` implementation that compresses data in-place). - /// - /// # Panics + /// # Safety /// - /// Panics if the filled region of the buffer would become larger than the initialized region. + /// The caller must ensure that the first `n` bytes of the cursor have been properly + /// initialised. + #[inline] + pub unsafe fn advance(&mut self, n: usize) -> &mut Self { + self.buf.filled += n; + self.buf.init = cmp::max(self.buf.init, self.buf.filled); + self + } + + /// Initializes all bytes in the cursor. #[inline] - pub fn set_filled(&mut self, n: usize) -> &mut Self { - assert!(n <= self.initialized); + pub fn ensure_init(&mut self) -> &mut Self { + for byte in self.uninit_mut() { + byte.write(0); + } + self.buf.init = self.buf.capacity(); - self.filled = n; self } - /// Asserts that the first `n` unfilled bytes of the buffer are initialized. + /// Asserts that the first `n` unfilled bytes of the cursor are initialized. /// - /// `ReadBuf` assumes that bytes are never de-initialized, so this method does nothing when called with fewer - /// bytes than are already known to be initialized. + /// `BorrowedBuf` assumes that bytes are never de-initialized, so this method does nothing when + /// called with fewer bytes than are already known to be initialized. /// /// # Safety /// - /// The caller must ensure that the first `n` unfilled bytes of the buffer have already been initialized. + /// The caller must ensure that the first `n` bytes of the buffer have already been initialized. #[inline] - pub unsafe fn assume_init(&mut self, n: usize) -> &mut Self { - self.initialized = cmp::max(self.initialized, self.filled + n); + pub unsafe fn set_init(&mut self, n: usize) -> &mut Self { + self.buf.init = cmp::max(self.buf.init, self.buf.filled + n); self } - /// Appends data to the buffer, advancing the written position and possibly also the initialized position. + /// Appends data to the cursor, advancing position within its buffer. /// /// # Panics /// - /// Panics if `self.remaining()` is less than `buf.len()`. + /// Panics if `self.capacity()` is less than `buf.len()`. #[inline] pub fn append(&mut self, buf: &[u8]) { - assert!(self.remaining() >= buf.len()); + assert!(self.capacity() >= buf.len()); // SAFETY: we do not de-initialize any of the elements of the slice unsafe { - MaybeUninit::write_slice(&mut self.unfilled_mut()[..buf.len()], buf); + MaybeUninit::write_slice(&mut self.as_mut()[..buf.len()], buf); } // SAFETY: We just added the entire contents of buf to the filled section. unsafe { - self.assume_init(buf.len()); + self.set_init(buf.len()); } - self.add_filled(buf.len()); + self.buf.filled += buf.len(); } +} - /// Returns the amount of bytes that have been filled. - #[inline] - pub fn filled_len(&self) -> usize { - self.filled +impl<'a> Write for BorrowedCursor<'a> { + fn write(&mut self, buf: &[u8]) -> Result<usize> { + self.append(buf); + Ok(buf.len()) } - /// Returns the amount of bytes that have been initialized. - #[inline] - pub fn initialized_len(&self) -> usize { - self.initialized + fn flush(&mut self) -> Result<()> { + Ok(()) } } diff --git a/library/std/src/io/readbuf/tests.rs b/library/std/src/io/readbuf/tests.rs index 3b7a5a56d..cc1b423f2 100644 --- a/library/std/src/io/readbuf/tests.rs +++ b/library/std/src/io/readbuf/tests.rs @@ -1,181 +1,175 @@ -use super::ReadBuf; +use super::BorrowedBuf; use crate::mem::MaybeUninit; -/// Test that ReadBuf has the correct numbers when created with new +/// Test that BorrowedBuf has the correct numbers when created with new #[test] fn new() { - let mut buf = [0; 16]; - let rbuf = ReadBuf::new(&mut buf); + let buf: &mut [_] = &mut [0; 16]; + let mut rbuf: BorrowedBuf<'_> = buf.into(); - assert_eq!(rbuf.filled_len(), 0); - assert_eq!(rbuf.initialized_len(), 16); + assert_eq!(rbuf.filled().len(), 0); + assert_eq!(rbuf.init_len(), 16); assert_eq!(rbuf.capacity(), 16); - assert_eq!(rbuf.remaining(), 16); + assert_eq!(rbuf.unfilled().capacity(), 16); } -/// Test that ReadBuf has the correct numbers when created with uninit +/// Test that BorrowedBuf has the correct numbers when created with uninit #[test] fn uninit() { - let mut buf = [MaybeUninit::uninit(); 16]; - let rbuf = ReadBuf::uninit(&mut buf); + let buf: &mut [_] = &mut [MaybeUninit::uninit(); 16]; + let mut rbuf: BorrowedBuf<'_> = buf.into(); - assert_eq!(rbuf.filled_len(), 0); - assert_eq!(rbuf.initialized_len(), 0); + assert_eq!(rbuf.filled().len(), 0); + assert_eq!(rbuf.init_len(), 0); assert_eq!(rbuf.capacity(), 16); - assert_eq!(rbuf.remaining(), 16); + assert_eq!(rbuf.unfilled().capacity(), 16); } #[test] fn initialize_unfilled() { - let mut buf = [MaybeUninit::uninit(); 16]; - let mut rbuf = ReadBuf::uninit(&mut buf); + let buf: &mut [_] = &mut [MaybeUninit::uninit(); 16]; + let mut rbuf: BorrowedBuf<'_> = buf.into(); - rbuf.initialize_unfilled(); + rbuf.unfilled().ensure_init(); - assert_eq!(rbuf.initialized_len(), 16); + assert_eq!(rbuf.init_len(), 16); } #[test] -fn initialize_unfilled_to() { - let mut buf = [MaybeUninit::uninit(); 16]; - let mut rbuf = ReadBuf::uninit(&mut buf); +fn addvance_filled() { + let buf: &mut [_] = &mut [0; 16]; + let mut rbuf: BorrowedBuf<'_> = buf.into(); - rbuf.initialize_unfilled_to(8); - - assert_eq!(rbuf.initialized_len(), 8); - - rbuf.initialize_unfilled_to(4); - - assert_eq!(rbuf.initialized_len(), 8); - - rbuf.set_filled(8); - - rbuf.initialize_unfilled_to(6); - - assert_eq!(rbuf.initialized_len(), 14); - - rbuf.initialize_unfilled_to(8); - - assert_eq!(rbuf.initialized_len(), 16); -} - -#[test] -fn add_filled() { - let mut buf = [0; 16]; - let mut rbuf = ReadBuf::new(&mut buf); - - rbuf.add_filled(1); - - assert_eq!(rbuf.filled_len(), 1); - assert_eq!(rbuf.remaining(), 15); -} - -#[test] -#[should_panic] -fn add_filled_panic() { - let mut buf = [MaybeUninit::uninit(); 16]; - let mut rbuf = ReadBuf::uninit(&mut buf); - - rbuf.add_filled(1); -} - -#[test] -fn set_filled() { - let mut buf = [0; 16]; - let mut rbuf = ReadBuf::new(&mut buf); - - rbuf.set_filled(16); - - assert_eq!(rbuf.filled_len(), 16); - assert_eq!(rbuf.remaining(), 0); - - rbuf.set_filled(6); - - assert_eq!(rbuf.filled_len(), 6); - assert_eq!(rbuf.remaining(), 10); -} - -#[test] -#[should_panic] -fn set_filled_panic() { - let mut buf = [MaybeUninit::uninit(); 16]; - let mut rbuf = ReadBuf::uninit(&mut buf); + unsafe { + rbuf.unfilled().advance(1); + } - rbuf.set_filled(16); + assert_eq!(rbuf.filled().len(), 1); + assert_eq!(rbuf.unfilled().capacity(), 15); } #[test] fn clear() { - let mut buf = [255; 16]; - let mut rbuf = ReadBuf::new(&mut buf); + let buf: &mut [_] = &mut [255; 16]; + let mut rbuf: BorrowedBuf<'_> = buf.into(); - rbuf.set_filled(16); + unsafe { + rbuf.unfilled().advance(16); + } - assert_eq!(rbuf.filled_len(), 16); - assert_eq!(rbuf.remaining(), 0); + assert_eq!(rbuf.filled().len(), 16); + assert_eq!(rbuf.unfilled().capacity(), 0); rbuf.clear(); - assert_eq!(rbuf.filled_len(), 0); - assert_eq!(rbuf.remaining(), 16); + assert_eq!(rbuf.filled().len(), 0); + assert_eq!(rbuf.unfilled().capacity(), 16); - assert_eq!(rbuf.initialized(), [255; 16]); + assert_eq!(rbuf.unfilled().init_ref(), [255; 16]); } #[test] -fn assume_init() { - let mut buf = [MaybeUninit::uninit(); 16]; - let mut rbuf = ReadBuf::uninit(&mut buf); +fn set_init() { + let buf: &mut [_] = &mut [MaybeUninit::uninit(); 16]; + let mut rbuf: BorrowedBuf<'_> = buf.into(); unsafe { - rbuf.assume_init(8); + rbuf.set_init(8); } - assert_eq!(rbuf.initialized_len(), 8); + assert_eq!(rbuf.init_len(), 8); - rbuf.add_filled(4); + unsafe { + rbuf.unfilled().advance(4); + } unsafe { - rbuf.assume_init(2); + rbuf.set_init(2); } - assert_eq!(rbuf.initialized_len(), 8); + assert_eq!(rbuf.init_len(), 8); unsafe { - rbuf.assume_init(8); + rbuf.set_init(8); } - assert_eq!(rbuf.initialized_len(), 12); + assert_eq!(rbuf.init_len(), 8); } #[test] fn append() { - let mut buf = [MaybeUninit::new(255); 16]; - let mut rbuf = ReadBuf::uninit(&mut buf); + let buf: &mut [_] = &mut [MaybeUninit::new(255); 16]; + let mut rbuf: BorrowedBuf<'_> = buf.into(); - rbuf.append(&[0; 8]); + rbuf.unfilled().append(&[0; 8]); - assert_eq!(rbuf.initialized_len(), 8); - assert_eq!(rbuf.filled_len(), 8); + assert_eq!(rbuf.init_len(), 8); + assert_eq!(rbuf.filled().len(), 8); assert_eq!(rbuf.filled(), [0; 8]); rbuf.clear(); - rbuf.append(&[1; 16]); + rbuf.unfilled().append(&[1; 16]); - assert_eq!(rbuf.initialized_len(), 16); - assert_eq!(rbuf.filled_len(), 16); + assert_eq!(rbuf.init_len(), 16); + assert_eq!(rbuf.filled().len(), 16); assert_eq!(rbuf.filled(), [1; 16]); } #[test] -fn filled_mut() { - let mut buf = [0; 16]; - let mut rbuf = ReadBuf::new(&mut buf); +fn reborrow_written() { + let buf: &mut [_] = &mut [MaybeUninit::new(0); 32]; + let mut buf: BorrowedBuf<'_> = buf.into(); + + let mut cursor = buf.unfilled(); + cursor.append(&[1; 16]); + + let mut cursor2 = cursor.reborrow(); + cursor2.append(&[2; 16]); + + assert_eq!(cursor2.written(), 32); + assert_eq!(cursor.written(), 32); + + assert_eq!(buf.unfilled().written(), 0); + assert_eq!(buf.init_len(), 32); + assert_eq!(buf.filled().len(), 32); + let filled = buf.filled(); + assert_eq!(&filled[..16], [1; 16]); + assert_eq!(&filled[16..], [2; 16]); +} + +#[test] +fn cursor_set_init() { + let buf: &mut [_] = &mut [MaybeUninit::uninit(); 16]; + let mut rbuf: BorrowedBuf<'_> = buf.into(); + + unsafe { + rbuf.unfilled().set_init(8); + } - rbuf.add_filled(8); + assert_eq!(rbuf.init_len(), 8); + assert_eq!(rbuf.unfilled().init_ref().len(), 8); + assert_eq!(rbuf.unfilled().init_mut().len(), 8); + assert_eq!(rbuf.unfilled().uninit_mut().len(), 8); + assert_eq!(unsafe { rbuf.unfilled().as_mut() }.len(), 16); + + unsafe { + rbuf.unfilled().advance(4); + } - let filled = rbuf.filled().to_vec(); + unsafe { + rbuf.unfilled().set_init(2); + } + + assert_eq!(rbuf.init_len(), 8); + + unsafe { + rbuf.unfilled().set_init(8); + } - assert_eq!(&*filled, &*rbuf.filled_mut()); + assert_eq!(rbuf.init_len(), 12); + assert_eq!(rbuf.unfilled().init_ref().len(), 8); + assert_eq!(rbuf.unfilled().init_mut().len(), 8); + assert_eq!(rbuf.unfilled().uninit_mut().len(), 4); + assert_eq!(unsafe { rbuf.unfilled().as_mut() }.len(), 12); } diff --git a/library/std/src/io/stdio.rs b/library/std/src/io/stdio.rs index 4d3736f79..2dc12a18a 100644 --- a/library/std/src/io/stdio.rs +++ b/library/std/src/io/stdio.rs @@ -8,7 +8,6 @@ use crate::io::prelude::*; use crate::cell::{Cell, RefCell}; use crate::fmt; use crate::io::{self, BufReader, IoSlice, IoSliceMut, LineWriter, Lines}; -use crate::pin::Pin; use crate::sync::atomic::{AtomicBool, Ordering}; use crate::sync::{Arc, Mutex, MutexGuard, OnceLock}; use crate::sys::stdio; @@ -526,7 +525,7 @@ pub struct Stdout { // FIXME: this should be LineWriter or BufWriter depending on the state of // stdout (tty or not). Note that if this is not line buffered it // should also flush-on-panic or some form of flush-on-abort. - inner: Pin<&'static ReentrantMutex<RefCell<LineWriter<StdoutRaw>>>>, + inner: &'static ReentrantMutex<RefCell<LineWriter<StdoutRaw>>>, } /// A locked reference to the [`Stdout`] handle. @@ -603,22 +602,27 @@ static STDOUT: OnceLock<ReentrantMutex<RefCell<LineWriter<StdoutRaw>>>> = OnceLo #[stable(feature = "rust1", since = "1.0.0")] pub fn stdout() -> Stdout { Stdout { - inner: Pin::static_ref(&STDOUT).get_or_init_pin( - || unsafe { ReentrantMutex::new(RefCell::new(LineWriter::new(stdout_raw()))) }, - |mutex| unsafe { mutex.init() }, - ), + inner: STDOUT + .get_or_init(|| ReentrantMutex::new(RefCell::new(LineWriter::new(stdout_raw())))), } } +// Flush the data and disable buffering during shutdown +// by replacing the line writer by one with zero +// buffering capacity. pub fn cleanup() { - if let Some(instance) = STDOUT.get() { - // Flush the data and disable buffering during shutdown - // by replacing the line writer by one with zero - // buffering capacity. + let mut initialized = false; + let stdout = STDOUT.get_or_init(|| { + initialized = true; + ReentrantMutex::new(RefCell::new(LineWriter::with_capacity(0, stdout_raw()))) + }); + + if !initialized { + // The buffer was previously initialized, overwrite it here. // We use try_lock() instead of lock(), because someone // might have leaked a StdoutLock, which would // otherwise cause a deadlock here. - if let Some(lock) = Pin::static_ref(instance).try_lock() { + if let Some(lock) = stdout.try_lock() { *lock.borrow_mut() = LineWriter::with_capacity(0, stdout_raw()); } } @@ -761,7 +765,7 @@ impl fmt::Debug for StdoutLock<'_> { /// standard library or via raw Windows API calls, will fail. #[stable(feature = "rust1", since = "1.0.0")] pub struct Stderr { - inner: Pin<&'static ReentrantMutex<RefCell<StderrRaw>>>, + inner: &'static ReentrantMutex<RefCell<StderrRaw>>, } /// A locked reference to the [`Stderr`] handle. @@ -834,16 +838,12 @@ pub struct StderrLock<'a> { #[stable(feature = "rust1", since = "1.0.0")] pub fn stderr() -> Stderr { // Note that unlike `stdout()` we don't use `at_exit` here to register a - // destructor. Stderr is not buffered , so there's no need to run a + // destructor. Stderr is not buffered, so there's no need to run a // destructor for flushing the buffer - static INSTANCE: OnceLock<ReentrantMutex<RefCell<StderrRaw>>> = OnceLock::new(); + static INSTANCE: ReentrantMutex<RefCell<StderrRaw>> = + ReentrantMutex::new(RefCell::new(stderr_raw())); - Stderr { - inner: Pin::static_ref(&INSTANCE).get_or_init_pin( - || unsafe { ReentrantMutex::new(RefCell::new(stderr_raw())) }, - |mutex| unsafe { mutex.init() }, - ), - } + Stderr { inner: &INSTANCE } } impl Stderr { @@ -986,12 +986,15 @@ pub fn set_output_capture(sink: Option<LocalStream>) -> Option<LocalStream> { /// otherwise. `label` identifies the stream in a panic message. /// /// This function is used to print error messages, so it takes extra -/// care to avoid causing a panic when `local_s` is unusable. -/// For instance, if the TLS key for the local stream is -/// already destroyed, or if the local stream is locked by another -/// thread, it will just fall back to the global stream. +/// care to avoid causing a panic when `OUTPUT_CAPTURE` is unusable. +/// For instance, if the TLS key for output capturing is already destroyed, or +/// if the local stream is in use by another thread, it will just fall back to +/// the global stream. /// /// However, if the actual I/O causes an error, this function does panic. +/// +/// Writing to non-blocking stdout/stderr can cause an error, which will lead +/// this function to panic. fn print_to<T>(args: fmt::Arguments<'_>, global_s: fn() -> T, label: &str) where T: Write, diff --git a/library/std/src/io/tests.rs b/library/std/src/io/tests.rs index f357f33ec..f4a886d88 100644 --- a/library/std/src/io/tests.rs +++ b/library/std/src/io/tests.rs @@ -1,4 +1,4 @@ -use super::{repeat, Cursor, ReadBuf, SeekFrom}; +use super::{repeat, BorrowedBuf, Cursor, SeekFrom}; use crate::cmp::{self, min}; use crate::io::{self, IoSlice, IoSliceMut}; use crate::io::{BufRead, BufReader, Read, Seek, Write}; @@ -94,7 +94,7 @@ fn read_to_end() { assert_eq!(c.read_to_end(&mut v).unwrap(), 1); assert_eq!(v, b"1"); - let cap = 1024 * 1024; + let cap = if cfg!(miri) { 1024 } else { 1024 * 1024 }; let data = (0..cap).map(|i| (i / 3) as u8).collect::<Vec<_>>(); let mut v = Vec::new(); let (a, b) = data.split_at(data.len() / 2); @@ -159,24 +159,24 @@ fn read_exact_slice() { #[test] fn read_buf_exact() { - let mut buf = [0; 4]; - let mut buf = ReadBuf::new(&mut buf); + let buf: &mut [_] = &mut [0; 4]; + let mut buf: BorrowedBuf<'_> = buf.into(); let mut c = Cursor::new(&b""[..]); - assert_eq!(c.read_buf_exact(&mut buf).unwrap_err().kind(), io::ErrorKind::UnexpectedEof); + assert_eq!(c.read_buf_exact(buf.unfilled()).unwrap_err().kind(), io::ErrorKind::UnexpectedEof); let mut c = Cursor::new(&b"123456789"[..]); - c.read_buf_exact(&mut buf).unwrap(); + c.read_buf_exact(buf.unfilled()).unwrap(); assert_eq!(buf.filled(), b"1234"); buf.clear(); - c.read_buf_exact(&mut buf).unwrap(); + c.read_buf_exact(buf.unfilled()).unwrap(); assert_eq!(buf.filled(), b"5678"); buf.clear(); - assert_eq!(c.read_buf_exact(&mut buf).unwrap_err().kind(), io::ErrorKind::UnexpectedEof); + assert_eq!(c.read_buf_exact(buf.unfilled()).unwrap_err().kind(), io::ErrorKind::UnexpectedEof); } #[test] @@ -309,6 +309,7 @@ fn chain_zero_length_read_is_not_eof() { #[bench] #[cfg_attr(target_os = "emscripten", ignore)] +#[cfg_attr(miri, ignore)] // Miri isn't fast... fn bench_read_to_end(b: &mut test::Bencher) { b.iter(|| { let mut lr = repeat(1).take(10000000); @@ -614,10 +615,10 @@ fn bench_take_read(b: &mut test::Bencher) { #[bench] fn bench_take_read_buf(b: &mut test::Bencher) { b.iter(|| { - let mut buf = [MaybeUninit::uninit(); 64]; + let buf: &mut [_] = &mut [MaybeUninit::uninit(); 64]; - let mut rbuf = ReadBuf::uninit(&mut buf); + let mut buf: BorrowedBuf<'_> = buf.into(); - [255; 128].take(64).read_buf(&mut rbuf).unwrap(); + [255; 128].take(64).read_buf(buf.unfilled()).unwrap(); }); } diff --git a/library/std/src/io/util.rs b/library/std/src/io/util.rs index c1300cd67..f076ee092 100644 --- a/library/std/src/io/util.rs +++ b/library/std/src/io/util.rs @@ -5,7 +5,7 @@ mod tests; use crate::fmt; use crate::io::{ - self, BufRead, IoSlice, IoSliceMut, Read, ReadBuf, Seek, SeekFrom, SizeHint, Write, + self, BorrowedCursor, BufRead, IoSlice, IoSliceMut, Read, Seek, SeekFrom, SizeHint, Write, }; /// A reader which is always at EOF. @@ -47,7 +47,7 @@ impl Read for Empty { } #[inline] - fn read_buf(&mut self, _buf: &mut ReadBuf<'_>) -> io::Result<()> { + fn read_buf(&mut self, _cursor: BorrowedCursor<'_>) -> io::Result<()> { Ok(()) } } @@ -130,21 +130,19 @@ impl Read for Repeat { Ok(buf.len()) } - fn read_buf(&mut self, buf: &mut ReadBuf<'_>) -> io::Result<()> { + fn read_buf(&mut self, mut buf: BorrowedCursor<'_>) -> io::Result<()> { // SAFETY: No uninit bytes are being written - for slot in unsafe { buf.unfilled_mut() } { + for slot in unsafe { buf.as_mut() } { slot.write(self.byte); } - let remaining = buf.remaining(); + let remaining = buf.capacity(); // SAFETY: the entire unfilled portion of buf has been initialized unsafe { - buf.assume_init(remaining); + buf.advance(remaining); } - buf.add_filled(remaining); - Ok(()) } diff --git a/library/std/src/io/util/tests.rs b/library/std/src/io/util/tests.rs index 08972a59a..ce5e2c9da 100644 --- a/library/std/src/io/util/tests.rs +++ b/library/std/src/io/util/tests.rs @@ -1,7 +1,7 @@ use crate::cmp::{max, min}; use crate::io::prelude::*; use crate::io::{ - copy, empty, repeat, sink, BufWriter, Empty, ReadBuf, Repeat, Result, SeekFrom, Sink, + copy, empty, repeat, sink, BorrowedBuf, BufWriter, Empty, Repeat, Result, SeekFrom, Sink, DEFAULT_BUF_SIZE, }; @@ -79,29 +79,29 @@ fn empty_reads() { assert_eq!(e.read(&mut [0; 1024]).unwrap(), 0); assert_eq!(e.by_ref().read(&mut [0; 1024]).unwrap(), 0); - let mut buf = []; - let mut buf = ReadBuf::uninit(&mut buf); - e.read_buf(&mut buf).unwrap(); - assert_eq!(buf.filled_len(), 0); - assert_eq!(buf.initialized_len(), 0); - - let mut buf = [MaybeUninit::uninit()]; - let mut buf = ReadBuf::uninit(&mut buf); - e.read_buf(&mut buf).unwrap(); - assert_eq!(buf.filled_len(), 0); - assert_eq!(buf.initialized_len(), 0); - - let mut buf = [MaybeUninit::uninit(); 1024]; - let mut buf = ReadBuf::uninit(&mut buf); - e.read_buf(&mut buf).unwrap(); - assert_eq!(buf.filled_len(), 0); - assert_eq!(buf.initialized_len(), 0); - - let mut buf = [MaybeUninit::uninit(); 1024]; - let mut buf = ReadBuf::uninit(&mut buf); - e.by_ref().read_buf(&mut buf).unwrap(); - assert_eq!(buf.filled_len(), 0); - assert_eq!(buf.initialized_len(), 0); + let buf: &mut [MaybeUninit<_>] = &mut []; + let mut buf: BorrowedBuf<'_> = buf.into(); + e.read_buf(buf.unfilled()).unwrap(); + assert_eq!(buf.len(), 0); + assert_eq!(buf.init_len(), 0); + + let buf: &mut [_] = &mut [MaybeUninit::uninit()]; + let mut buf: BorrowedBuf<'_> = buf.into(); + e.read_buf(buf.unfilled()).unwrap(); + assert_eq!(buf.len(), 0); + assert_eq!(buf.init_len(), 0); + + let buf: &mut [_] = &mut [MaybeUninit::uninit(); 1024]; + let mut buf: BorrowedBuf<'_> = buf.into(); + e.read_buf(buf.unfilled()).unwrap(); + assert_eq!(buf.len(), 0); + assert_eq!(buf.init_len(), 0); + + let buf: &mut [_] = &mut [MaybeUninit::uninit(); 1024]; + let mut buf: BorrowedBuf<'_> = buf.into(); + e.by_ref().read_buf(buf.unfilled()).unwrap(); + assert_eq!(buf.len(), 0); + assert_eq!(buf.init_len(), 0); } #[test] diff --git a/library/std/src/keyword_docs.rs b/library/std/src/keyword_docs.rs index 7157b5af0..a4b0522b0 100644 --- a/library/std/src/keyword_docs.rs +++ b/library/std/src/keyword_docs.rs @@ -1921,7 +1921,7 @@ mod type_keyword {} /// and [proposal]s exist to use `unsafe {}` blocks inside such functions when /// making `unsafe` operations. /// -/// See the [Rustnomicon] and the [Reference] for more informations. +/// See the [Rustnomicon] and the [Reference] for more information. /// /// # Examples /// @@ -2113,7 +2113,7 @@ mod use_keyword {} /// Add constraints that must be upheld to use an item. /// /// `where` allows specifying constraints on lifetime and generic parameters. -/// The [RFC] introducing `where` contains detailed informations about the +/// The [RFC] introducing `where` contains detailed information about the /// keyword. /// /// # Examples @@ -2355,7 +2355,7 @@ mod dyn_keyword {} /// println!("f = {f} and i = {i}"); /// ``` /// -/// See the [Reference][union] for more informations on `union`s. +/// See the [Reference][union] for more information on `union`s. /// /// [`struct`]: keyword.struct.html /// [union]: ../reference/items/unions.html diff --git a/library/std/src/lib.rs b/library/std/src/lib.rs index 20d25a608..bc4f1b27c 100644 --- a/library/std/src/lib.rs +++ b/library/std/src/lib.rs @@ -187,6 +187,7 @@ //! [rust-discord]: https://discord.gg/rust-lang //! [array]: prim@array //! [slice]: prim@slice + #![cfg_attr(not(feature = "restricted-std"), stable(feature = "rust1", since = "1.0.0"))] #![cfg_attr(feature = "restricted-std", unstable(feature = "restricted_std", issue = "none"))] #![doc( @@ -201,25 +202,35 @@ no_global_oom_handling, not(no_global_oom_handling) ))] +// To run libstd tests without x.py without ending up with two copies of libstd, Miri needs to be +// able to "empty" this crate. See <https://github.com/rust-lang/miri-test-libstd/issues/4>. +// rustc itself never sets the feature, so this line has no affect there. +#![cfg(any(not(feature = "miri-test-libstd"), test, doctest))] +// miri-test-libstd also prefers to make std use the sysroot versions of the dependencies. +#![cfg_attr(feature = "miri-test-libstd", feature(rustc_private))] // Don't link to std. We are std. #![no_std] +// Tell the compiler to link to either panic_abort or panic_unwind +#![needs_panic_runtime] +// +// Lints: #![warn(deprecated_in_future)] #![warn(missing_docs)] #![warn(missing_debug_implementations)] #![allow(explicit_outlives_requirements)] #![allow(unused_lifetimes)] -// Tell the compiler to link to either panic_abort or panic_unwind -#![needs_panic_runtime] +#![deny(rustc::existing_doc_keyword)] // Ensure that std can be linked against panic_abort despite compiled with `-C panic=unwind` -#![cfg_attr(not(bootstrap), deny(ffi_unwind_calls))] +#![deny(ffi_unwind_calls)] // std may use features in a platform-specific way #![allow(unused_features)] +// +// Features: #![cfg_attr(test, feature(internal_output_capture, print_internals, update_panic_count, rt))] #![cfg_attr( all(target_vendor = "fortanix", target_env = "sgx"), feature(slice_index_methods, coerce_unsized, sgx_platform) )] -#![deny(rustc::existing_doc_keyword)] // // Language features: #![feature(alloc_error_handler)] @@ -241,11 +252,12 @@ #![feature(dropck_eyepatch)] #![feature(exhaustive_patterns)] #![feature(intra_doc_pointers)] -#![feature(label_break_value)] +#![cfg_attr(bootstrap, feature(label_break_value))] #![feature(lang_items)] #![feature(let_chains)] -#![feature(let_else)] +#![cfg_attr(bootstrap, feature(let_else))] #![feature(linkage)] +#![feature(link_cfg)] #![feature(min_specialization)] #![feature(must_not_suspend)] #![feature(needs_panic_runtime)] @@ -258,6 +270,7 @@ #![feature(staged_api)] #![feature(thread_local)] #![feature(try_blocks)] +#![feature(utf8_chunks)] // // Library features (core): #![feature(array_error_internals)] @@ -269,10 +282,14 @@ #![feature(cstr_internals)] #![feature(duration_checked_float)] #![feature(duration_constants)] +#![cfg_attr(not(bootstrap), feature(error_generic_member_access))] +#![cfg_attr(not(bootstrap), feature(error_in_core))] +#![cfg_attr(not(bootstrap), feature(error_iter))] #![feature(exact_size_is_empty)] #![feature(exclusive_wrapper)] #![feature(extend_one)] #![feature(float_minimum_maximum)] +#![feature(float_next_up_down)] #![feature(hasher_prefixfree_extras)] #![feature(hashmap_internals)] #![feature(int_error_internals)] @@ -284,6 +301,8 @@ #![feature(panic_can_unwind)] #![feature(panic_info_message)] #![feature(panic_internals)] +#![feature(pointer_byte_offsets)] +#![feature(pointer_is_aligned)] #![feature(portable_simd)] #![feature(prelude_2024)] #![feature(provide_any)] @@ -294,6 +313,8 @@ #![feature(std_internals)] #![feature(str_internals)] #![feature(strict_provenance)] +#![feature(maybe_uninit_uninit_array)] +#![feature(const_maybe_uninit_uninit_array)] // // Library features (alloc): #![feature(alloc_layout_extra)] @@ -576,6 +597,7 @@ pub mod alloc; // Private support modules mod panicking; +mod personality; #[path = "../../backtrace/src/lib.rs"] #[allow(dead_code, unused_attributes)] diff --git a/library/std/src/macros.rs b/library/std/src/macros.rs index 0cb21ef53..6e4ba1404 100644 --- a/library/std/src/macros.rs +++ b/library/std/src/macros.rs @@ -27,17 +27,31 @@ macro_rules! panic { /// necessary to use [`io::stdout().flush()`][flush] to ensure the output is emitted /// immediately. /// +/// The `print!` macro will lock the standard output on each call. If you call +/// `print!` within a hot loop, this behavior may be the bottleneck of the loop. +/// To avoid this, lock stdout with [`io::stdout().lock()`][lock]: +/// ``` +/// use std::io::{stdout, Write}; +/// +/// let mut lock = stdout().lock(); +/// write!(lock, "hello world").unwrap(); +/// ``` +/// /// Use `print!` only for the primary output of your program. Use /// [`eprint!`] instead to print error and progress messages. /// /// [flush]: crate::io::Write::flush /// [`println!`]: crate::println /// [`eprint!`]: crate::eprint +/// [lock]: crate::io::Stdout /// /// # Panics /// /// Panics if writing to `io::stdout()` fails. /// +/// Writing to non-blocking stdout can cause an error, which will lead +/// this macro to panic. +/// /// # Examples /// /// ``` @@ -75,16 +89,30 @@ macro_rules! print { /// This macro uses the same syntax as [`format!`], but writes to the standard output instead. /// See [`std::fmt`] for more information. /// +/// The `println!` macro will lock the standard output on each call. If you call +/// `println!` within a hot loop, this behavior may be the bottleneck of the loop. +/// To avoid this, lock stdout with [`io::stdout().lock()`][lock]: +/// ``` +/// use std::io::{stdout, Write}; +/// +/// let mut lock = stdout().lock(); +/// writeln!(lock, "hello world").unwrap(); +/// ``` +/// /// Use `println!` only for the primary output of your program. Use /// [`eprintln!`] instead to print error and progress messages. /// /// [`std::fmt`]: crate::fmt /// [`eprintln!`]: crate::eprintln +/// [lock]: crate::io::Stdout /// /// # Panics /// /// Panics if writing to [`io::stdout`] fails. /// +/// Writing to non-blocking stdout can cause an error, which will lead +/// this macro to panic. +/// /// [`io::stdout`]: crate::io::stdout /// /// # Examples @@ -93,6 +121,8 @@ macro_rules! print { /// println!(); // prints just a newline /// println!("hello there!"); /// println!("format {} arguments", "some"); +/// let local_variable = "some"; +/// println!("format {local_variable} arguments"); /// ``` #[macro_export] #[stable(feature = "rust1", since = "1.0.0")] @@ -123,6 +153,9 @@ macro_rules! println { /// /// Panics if writing to `io::stderr` fails. /// +/// Writing to non-blocking stdout can cause an error, which will lead +/// this macro to panic. +/// /// # Examples /// /// ``` @@ -155,6 +188,9 @@ macro_rules! eprint { /// /// Panics if writing to `io::stderr` fails. /// +/// Writing to non-blocking stdout can cause an error, which will lead +/// this macro to panic. +/// /// # Examples /// /// ``` diff --git a/library/std/src/net/display_buffer.rs b/library/std/src/net/display_buffer.rs new file mode 100644 index 000000000..7aadf06e9 --- /dev/null +++ b/library/std/src/net/display_buffer.rs @@ -0,0 +1,40 @@ +use crate::fmt; +use crate::mem::MaybeUninit; +use crate::str; + +/// Used for slow path in `Display` implementations when alignment is required. +pub struct DisplayBuffer<const SIZE: usize> { + buf: [MaybeUninit<u8>; SIZE], + len: usize, +} + +impl<const SIZE: usize> DisplayBuffer<SIZE> { + #[inline] + pub const fn new() -> Self { + Self { buf: MaybeUninit::uninit_array(), len: 0 } + } + + #[inline] + pub fn as_str(&self) -> &str { + // SAFETY: `buf` is only written to by the `fmt::Write::write_str` implementation + // which writes a valid UTF-8 string to `buf` and correctly sets `len`. + unsafe { + let s = MaybeUninit::slice_assume_init_ref(&self.buf[..self.len]); + str::from_utf8_unchecked(s) + } + } +} + +impl<const SIZE: usize> fmt::Write for DisplayBuffer<SIZE> { + fn write_str(&mut self, s: &str) -> fmt::Result { + let bytes = s.as_bytes(); + + if let Some(buf) = self.buf.get_mut(self.len..(self.len + bytes.len())) { + MaybeUninit::write_slice(buf, bytes); + self.len += bytes.len(); + Ok(()) + } else { + Err(fmt::Error) + } + } +} diff --git a/library/std/src/net/ip.rs b/library/std/src/net/ip_addr.rs index 41ca9ba84..4f14fc280 100644 --- a/library/std/src/net/ip.rs +++ b/library/std/src/net/ip_addr.rs @@ -3,12 +3,13 @@ mod tests; use crate::cmp::Ordering; -use crate::fmt::{self, Write as FmtWrite}; -use crate::io::Write as IoWrite; +use crate::fmt::{self, Write}; use crate::mem::transmute; use crate::sys::net::netc as c; use crate::sys_common::{FromInner, IntoInner}; +use super::display_buffer::DisplayBuffer; + /// An IP address, either IPv4 or IPv6. /// /// This enum can contain either an [`Ipv4Addr`] or an [`Ipv6Addr`], see their @@ -28,6 +29,7 @@ use crate::sys_common::{FromInner, IntoInner}; /// assert_eq!(localhost_v4.is_ipv6(), false); /// assert_eq!(localhost_v4.is_ipv4(), true); /// ``` +#[cfg_attr(not(test), rustc_diagnostic_item = "IpAddr")] #[stable(feature = "ip_addr", since = "1.7.0")] #[derive(Copy, Clone, Eq, PartialEq, Hash, PartialOrd, Ord)] pub enum IpAddr { @@ -71,6 +73,7 @@ pub enum IpAddr { /// assert!("0xcb.0x0.0x71.0x00".parse::<Ipv4Addr>().is_err()); // all octets are in hex /// ``` #[derive(Copy, Clone, PartialEq, Eq, Hash)] +#[cfg_attr(not(test), rustc_diagnostic_item = "Ipv4Addr")] #[stable(feature = "rust1", since = "1.0.0")] pub struct Ipv4Addr { octets: [u8; 4], @@ -153,6 +156,7 @@ pub struct Ipv4Addr { /// assert_eq!(localhost.is_loopback(), true); /// ``` #[derive(Copy, Clone, PartialEq, Eq, Hash)] +#[cfg_attr(not(test), rustc_diagnostic_item = "Ipv6Addr")] #[stable(feature = "rust1", since = "1.0.0")] pub struct Ipv6Addr { octets: [u8; 16], @@ -618,25 +622,31 @@ impl Ipv4Addr { matches!(self.octets(), [169, 254, ..]) } - /// Returns [`true`] if the address appears to be globally routable. - /// See [iana-ipv4-special-registry][ipv4-sr]. + /// Returns [`true`] if the address appears to be globally reachable + /// as specified by the [IANA IPv4 Special-Purpose Address Registry]. + /// Whether or not an address is practically reachable will depend on your network configuration. + /// + /// Most IPv4 addresses are globally reachable; + /// unless they are specifically defined as *not* globally reachable. /// - /// The following return [`false`]: + /// Non-exhaustive list of notable addresses that are not globally reachable: /// - /// - private addresses (see [`Ipv4Addr::is_private()`]) - /// - the loopback address (see [`Ipv4Addr::is_loopback()`]) - /// - the link-local address (see [`Ipv4Addr::is_link_local()`]) - /// - the broadcast address (see [`Ipv4Addr::is_broadcast()`]) - /// - addresses used for documentation (see [`Ipv4Addr::is_documentation()`]) - /// - the unspecified address (see [`Ipv4Addr::is_unspecified()`]), and the whole - /// `0.0.0.0/8` block - /// - addresses reserved for future protocols, except - /// `192.0.0.9/32` and `192.0.0.10/32` which are globally routable - /// - addresses reserved for future use (see [`Ipv4Addr::is_reserved()`] - /// - addresses reserved for networking devices benchmarking (see - /// [`Ipv4Addr::is_benchmarking()`]) + /// - The [unspecified address] ([`is_unspecified`](Ipv4Addr::is_unspecified)) + /// - Addresses reserved for private use ([`is_private`](Ipv4Addr::is_private)) + /// - Addresses in the shared address space ([`is_shared`](Ipv4Addr::is_shared)) + /// - Loopback addresses ([`is_loopback`](Ipv4Addr::is_loopback)) + /// - Link-local addresses ([`is_link_local`](Ipv4Addr::is_link_local)) + /// - Addresses reserved for documentation ([`is_documentation`](Ipv4Addr::is_documentation)) + /// - Addresses reserved for benchmarking ([`is_benchmarking`](Ipv4Addr::is_benchmarking)) + /// - Reserved addresses ([`is_reserved`](Ipv4Addr::is_reserved)) + /// - The [broadcast address] ([`is_broadcast`](Ipv4Addr::is_broadcast)) /// - /// [ipv4-sr]: https://www.iana.org/assignments/iana-ipv4-special-registry/iana-ipv4-special-registry.xhtml + /// For the complete overview of which addresses are globally reachable, see the table at the [IANA IPv4 Special-Purpose Address Registry]. + /// + /// [IANA IPv4 Special-Purpose Address Registry]: https://www.iana.org/assignments/iana-ipv4-special-registry/iana-ipv4-special-registry.xhtml + /// [unspecified address]: Ipv4Addr::UNSPECIFIED + /// [broadcast address]: Ipv4Addr::BROADCAST + /// /// # Examples /// @@ -645,71 +655,61 @@ impl Ipv4Addr { /// /// use std::net::Ipv4Addr; /// - /// // private addresses are not global + /// // Most IPv4 addresses are globally reachable: + /// assert_eq!(Ipv4Addr::new(80, 9, 12, 3).is_global(), true); + /// + /// // However some addresses have been assigned a special meaning + /// // that makes them not globally reachable. Some examples are: + /// + /// // The unspecified address (`0.0.0.0`) + /// assert_eq!(Ipv4Addr::UNSPECIFIED.is_global(), false); + /// + /// // Addresses reserved for private use (`10.0.0.0/8`, `172.16.0.0/12`, 192.168.0.0/16) /// assert_eq!(Ipv4Addr::new(10, 254, 0, 0).is_global(), false); /// assert_eq!(Ipv4Addr::new(192, 168, 10, 65).is_global(), false); /// assert_eq!(Ipv4Addr::new(172, 16, 10, 65).is_global(), false); /// - /// // the 0.0.0.0/8 block is not global - /// assert_eq!(Ipv4Addr::new(0, 1, 2, 3).is_global(), false); - /// // in particular, the unspecified address is not global - /// assert_eq!(Ipv4Addr::new(0, 0, 0, 0).is_global(), false); + /// // Addresses in the shared address space (`100.64.0.0/10`) + /// assert_eq!(Ipv4Addr::new(100, 100, 0, 0).is_global(), false); /// - /// // the loopback address is not global - /// assert_eq!(Ipv4Addr::new(127, 0, 0, 1).is_global(), false); + /// // The loopback addresses (`127.0.0.0/8`) + /// assert_eq!(Ipv4Addr::LOCALHOST.is_global(), false); /// - /// // link local addresses are not global + /// // Link-local addresses (`169.254.0.0/16`) /// assert_eq!(Ipv4Addr::new(169, 254, 45, 1).is_global(), false); /// - /// // the broadcast address is not global - /// assert_eq!(Ipv4Addr::new(255, 255, 255, 255).is_global(), false); - /// - /// // the address space designated for documentation is not global + /// // Addresses reserved for documentation (`192.0.2.0/24`, `198.51.100.0/24`, `203.0.113.0/24`) /// assert_eq!(Ipv4Addr::new(192, 0, 2, 255).is_global(), false); /// assert_eq!(Ipv4Addr::new(198, 51, 100, 65).is_global(), false); /// assert_eq!(Ipv4Addr::new(203, 0, 113, 6).is_global(), false); /// - /// // shared addresses are not global - /// assert_eq!(Ipv4Addr::new(100, 100, 0, 0).is_global(), false); - /// - /// // addresses reserved for protocol assignment are not global - /// assert_eq!(Ipv4Addr::new(192, 0, 0, 0).is_global(), false); - /// assert_eq!(Ipv4Addr::new(192, 0, 0, 255).is_global(), false); + /// // Addresses reserved for benchmarking (`198.18.0.0/15`) + /// assert_eq!(Ipv4Addr::new(198, 18, 0, 0).is_global(), false); /// - /// // addresses reserved for future use are not global + /// // Reserved addresses (`240.0.0.0/4`) /// assert_eq!(Ipv4Addr::new(250, 10, 20, 30).is_global(), false); /// - /// // addresses reserved for network devices benchmarking are not global - /// assert_eq!(Ipv4Addr::new(198, 18, 0, 0).is_global(), false); + /// // The broadcast address (`255.255.255.255`) + /// assert_eq!(Ipv4Addr::BROADCAST.is_global(), false); /// - /// // All the other addresses are global - /// assert_eq!(Ipv4Addr::new(1, 1, 1, 1).is_global(), true); - /// assert_eq!(Ipv4Addr::new(80, 9, 12, 3).is_global(), true); + /// // For a complete overview see the IANA IPv4 Special-Purpose Address Registry. /// ``` #[rustc_const_unstable(feature = "const_ipv4", issue = "76205")] #[unstable(feature = "ip", issue = "27709")] #[must_use] #[inline] pub const fn is_global(&self) -> bool { - // check if this address is 192.0.0.9 or 192.0.0.10. These addresses are the only two - // globally routable addresses in the 192.0.0.0/24 range. - if u32::from_be_bytes(self.octets()) == 0xc0000009 - || u32::from_be_bytes(self.octets()) == 0xc000000a - { - return true; - } - !self.is_private() - && !self.is_loopback() - && !self.is_link_local() - && !self.is_broadcast() - && !self.is_documentation() - && !self.is_shared() + !(self.octets()[0] == 0 // "This network" + || self.is_private() + || self.is_shared() + || self.is_loopback() + || self.is_link_local() // addresses reserved for future protocols (`192.0.0.0/24`) - && !(self.octets()[0] == 192 && self.octets()[1] == 0 && self.octets()[2] == 0) - && !self.is_reserved() - && !self.is_benchmarking() - // Make sure the address is not in 0.0.0.0/8 - && self.octets()[0] != 0 + ||(self.octets()[0] == 192 && self.octets()[1] == 0 && self.octets()[2] == 0) + || self.is_documentation() + || self.is_benchmarking() + || self.is_reserved() + || self.is_broadcast()) } /// Returns [`true`] if this address is part of the Shared Address Space defined in @@ -991,21 +991,19 @@ impl From<Ipv6Addr> for IpAddr { impl fmt::Display for Ipv4Addr { fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result { let octets = self.octets(); - // Fast Path: if there's no alignment stuff, write directly to the buffer + + // If there are no alignment requirements, write the IP address directly to `f`. + // Otherwise, write it to a local buffer and then use `f.pad`. if fmt.precision().is_none() && fmt.width().is_none() { write!(fmt, "{}.{}.{}.{}", octets[0], octets[1], octets[2], octets[3]) } else { - const IPV4_BUF_LEN: usize = 15; // Long enough for the longest possible IPv4 address - let mut buf = [0u8; IPV4_BUF_LEN]; - let mut buf_slice = &mut buf[..]; + const LONGEST_IPV4_ADDR: &str = "255.255.255.255"; - // Note: The call to write should never fail, hence the unwrap - write!(buf_slice, "{}.{}.{}.{}", octets[0], octets[1], octets[2], octets[3]).unwrap(); - let len = IPV4_BUF_LEN - buf_slice.len(); + let mut buf = DisplayBuffer::<{ LONGEST_IPV4_ADDR.len() }>::new(); + // Buffer is long enough for the longest possible IPv4 address, so this should never fail. + write!(buf, "{}.{}.{}.{}", octets[0], octets[1], octets[2], octets[3]).unwrap(); - // This unsafe is OK because we know what is being written to the buffer - let buf = unsafe { crate::str::from_utf8_unchecked(&buf[..len]) }; - fmt.pad(buf) + fmt.pad(buf.as_str()) } } } @@ -1300,13 +1298,33 @@ impl Ipv6Addr { u128::from_be_bytes(self.octets()) == u128::from_be_bytes(Ipv6Addr::LOCALHOST.octets()) } - /// Returns [`true`] if the address appears to be globally routable. + /// Returns [`true`] if the address appears to be globally reachable + /// as specified by the [IANA IPv6 Special-Purpose Address Registry]. + /// Whether or not an address is practically reachable will depend on your network configuration. /// - /// The following return [`false`]: + /// Most IPv6 addresses are globally reachable; + /// unless they are specifically defined as *not* globally reachable. /// - /// - the loopback address - /// - link-local and unique local unicast addresses - /// - interface-, link-, realm-, admin- and site-local multicast addresses + /// Non-exhaustive list of notable addresses that are not globally reachable: + /// - The [unspecified address] ([`is_unspecified`](Ipv6Addr::is_unspecified)) + /// - The [loopback address] ([`is_loopback`](Ipv6Addr::is_loopback)) + /// - IPv4-mapped addresses + /// - Addresses reserved for benchmarking + /// - Addresses reserved for documentation ([`is_documentation`](Ipv6Addr::is_documentation)) + /// - Unique local addresses ([`is_unique_local`](Ipv6Addr::is_unique_local)) + /// - Unicast addresses with link-local scope ([`is_unicast_link_local`](Ipv6Addr::is_unicast_link_local)) + /// + /// For the complete overview of which addresses are globally reachable, see the table at the [IANA IPv6 Special-Purpose Address Registry]. + /// + /// Note that an address having global scope is not the same as being globally reachable, + /// and there is no direct relation between the two concepts: There exist addresses with global scope + /// that are not globally reachable (for example unique local addresses), + /// and addresses that are globally reachable without having global scope + /// (multicast addresses with non-global scope). + /// + /// [IANA IPv6 Special-Purpose Address Registry]: https://www.iana.org/assignments/iana-ipv6-special-registry/iana-ipv6-special-registry.xhtml + /// [unspecified address]: Ipv6Addr::UNSPECIFIED + /// [loopback address]: Ipv6Addr::LOCALHOST /// /// # Examples /// @@ -1315,20 +1333,65 @@ impl Ipv6Addr { /// /// use std::net::Ipv6Addr; /// - /// assert_eq!(Ipv6Addr::new(0, 0, 0, 0, 0, 0xffff, 0xc00a, 0x2ff).is_global(), true); - /// assert_eq!(Ipv6Addr::new(0, 0, 0, 0, 0, 0, 0, 0x1).is_global(), false); - /// assert_eq!(Ipv6Addr::new(0, 0, 0x1c9, 0, 0, 0xafc8, 0, 0x1).is_global(), true); + /// // Most IPv6 addresses are globally reachable: + /// assert_eq!(Ipv6Addr::new(0x26, 0, 0x1c9, 0, 0, 0xafc8, 0x10, 0x1).is_global(), true); + /// + /// // However some addresses have been assigned a special meaning + /// // that makes them not globally reachable. Some examples are: + /// + /// // The unspecified address (`::`) + /// assert_eq!(Ipv6Addr::UNSPECIFIED.is_global(), false); + /// + /// // The loopback address (`::1`) + /// assert_eq!(Ipv6Addr::LOCALHOST.is_global(), false); + /// + /// // IPv4-mapped addresses (`::ffff:0:0/96`) + /// assert_eq!(Ipv6Addr::new(0, 0, 0, 0, 0, 0xffff, 0xc00a, 0x2ff).is_global(), false); + /// + /// // Addresses reserved for benchmarking (`2001:2::/48`) + /// assert_eq!(Ipv6Addr::new(0x2001, 2, 0, 0, 0, 0, 0, 1,).is_global(), false); + /// + /// // Addresses reserved for documentation (`2001:db8::/32`) + /// assert_eq!(Ipv6Addr::new(0x2001, 0xdb8, 0, 0, 0, 0, 0, 1).is_global(), false); + /// + /// // Unique local addresses (`fc00::/7`) + /// assert_eq!(Ipv6Addr::new(0xfc02, 0, 0, 0, 0, 0, 0, 1).is_global(), false); + /// + /// // Unicast addresses with link-local scope (`fe80::/10`) + /// assert_eq!(Ipv6Addr::new(0xfe81, 0, 0, 0, 0, 0, 0, 1).is_global(), false); + /// + /// // For a complete overview see the IANA IPv6 Special-Purpose Address Registry. /// ``` #[rustc_const_unstable(feature = "const_ipv6", issue = "76205")] #[unstable(feature = "ip", issue = "27709")] #[must_use] #[inline] pub const fn is_global(&self) -> bool { - match self.multicast_scope() { - Some(Ipv6MulticastScope::Global) => true, - None => self.is_unicast_global(), - _ => false, - } + !(self.is_unspecified() + || self.is_loopback() + // IPv4-mapped Address (`::ffff:0:0/96`) + || matches!(self.segments(), [0, 0, 0, 0, 0, 0xffff, _, _]) + // IPv4-IPv6 Translat. (`64:ff9b:1::/48`) + || matches!(self.segments(), [0x64, 0xff9b, 1, _, _, _, _, _]) + // Discard-Only Address Block (`100::/64`) + || matches!(self.segments(), [0x100, 0, 0, 0, _, _, _, _]) + // IETF Protocol Assignments (`2001::/23`) + || (matches!(self.segments(), [0x2001, b, _, _, _, _, _, _] if b < 0x200) + && !( + // Port Control Protocol Anycast (`2001:1::1`) + u128::from_be_bytes(self.octets()) == 0x2001_0001_0000_0000_0000_0000_0000_0001 + // Traversal Using Relays around NAT Anycast (`2001:1::2`) + || u128::from_be_bytes(self.octets()) == 0x2001_0001_0000_0000_0000_0000_0000_0002 + // AMT (`2001:3::/32`) + || matches!(self.segments(), [0x2001, 3, _, _, _, _, _, _]) + // AS112-v6 (`2001:4:112::/48`) + || matches!(self.segments(), [0x2001, 4, 0x112, _, _, _, _, _]) + // ORCHIDv2 (`2001:20::/28`) + || matches!(self.segments(), [0x2001, b, _, _, _, _, _, _] if b >= 0x20 && b <= 0x2F) + )) + || self.is_documentation() + || self.is_unique_local() + || self.is_unicast_link_local()) } /// Returns [`true`] if this is a unique local address (`fc00::/7`). @@ -1525,6 +1588,7 @@ impl Ipv6Addr { && !self.is_unique_local() && !self.is_unspecified() && !self.is_documentation() + && !self.is_benchmarking() } /// Returns the address's multicast scope if the address is multicast. @@ -1708,8 +1772,8 @@ impl Ipv6Addr { #[stable(feature = "rust1", since = "1.0.0")] impl fmt::Display for Ipv6Addr { fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { - // If there are no alignment requirements, write out the IP address to - // f. Otherwise, write it to a local buffer, then use f.pad. + // If there are no alignment requirements, write the IP address directly to `f`. + // Otherwise, write it to a local buffer and then use `f.pad`. if f.precision().is_none() && f.width().is_none() { let segments = self.segments(); @@ -1780,22 +1844,13 @@ impl fmt::Display for Ipv6Addr { } } } else { - // Slow path: write the address to a local buffer, then use f.pad. - // Defined recursively by using the fast path to write to the - // buffer. - - // This is the largest possible size of an IPv6 address - const IPV6_BUF_LEN: usize = (4 * 8) + 7; - let mut buf = [0u8; IPV6_BUF_LEN]; - let mut buf_slice = &mut buf[..]; - - // Note: This call to write should never fail, so unwrap is okay. - write!(buf_slice, "{}", self).unwrap(); - let len = IPV6_BUF_LEN - buf_slice.len(); - - // This is safe because we know exactly what can be in this buffer - let buf = unsafe { crate::str::from_utf8_unchecked(&buf[..len]) }; - f.pad(buf) + const LONGEST_IPV6_ADDR: &str = "ffff:ffff:ffff:ffff:ffff:ffff:ffff:ffff"; + + let mut buf = DisplayBuffer::<{ LONGEST_IPV6_ADDR.len() }>::new(); + // Buffer is long enough for the longest possible IPv6 address, so this should never fail. + write!(buf, "{}", self).unwrap(); + + f.pad(buf.as_str()) } } } diff --git a/library/std/src/net/ip/tests.rs b/library/std/src/net/ip_addr/tests.rs index c29509331..7c3430b2b 100644 --- a/library/std/src/net/ip/tests.rs +++ b/library/std/src/net/ip_addr/tests.rs @@ -321,15 +321,15 @@ fn ip_properties() { check!("fe80:ffff::"); check!("febf:ffff::"); check!("fec0::", global); - check!("ff01::", multicast); - check!("ff02::", multicast); - check!("ff03::", multicast); - check!("ff04::", multicast); - check!("ff05::", multicast); - check!("ff08::", multicast); + check!("ff01::", global | multicast); + check!("ff02::", global | multicast); + check!("ff03::", global | multicast); + check!("ff04::", global | multicast); + check!("ff05::", global | multicast); + check!("ff08::", global | multicast); check!("ff0e::", global | multicast); check!("2001:db8:85a3::8a2e:370:7334", doc); - check!("2001:2::ac32:23ff:21", global | benchmarking); + check!("2001:2::ac32:23ff:21", benchmarking); check!("102:304:506:708:90a:b0c:d0e:f10", global); } @@ -609,6 +609,60 @@ fn ipv6_properties() { check!("1::", &[0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], global | unicast_global); + check!( + "::ffff:127.0.0.1", + &[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0xff, 0xff, 0x7f, 0, 0, 1], + unicast_global + ); + + check!( + "64:ff9b:1::", + &[0, 0x64, 0xff, 0x9b, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], + unicast_global + ); + + check!("100::", &[0x01, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], unicast_global); + + check!("2001::", &[0x20, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], unicast_global); + + check!( + "2001:1::1", + &[0x20, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1], + global | unicast_global + ); + + check!( + "2001:1::2", + &[0x20, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2], + global | unicast_global + ); + + check!( + "2001:3::", + &[0x20, 1, 0, 3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], + global | unicast_global + ); + + check!( + "2001:4:112::", + &[0x20, 1, 0, 4, 1, 0x12, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], + global | unicast_global + ); + + check!( + "2001:20::", + &[0x20, 1, 0, 0x20, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], + global | unicast_global + ); + + check!("2001:30::", &[0x20, 1, 0, 0x30, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], unicast_global); + + check!( + "2001:200::", + &[0x20, 1, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], + global | unicast_global + ); + check!("fc00::", &[0xfc, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], unique_local); check!( @@ -666,21 +720,37 @@ fn ipv6_properties() { check!( "ff01::", &[0xff, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], - multicast_interface_local + multicast_interface_local | global ); - check!("ff02::", &[0xff, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], multicast_link_local); + check!( + "ff02::", + &[0xff, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], + multicast_link_local | global + ); - check!("ff03::", &[0xff, 3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], multicast_realm_local); + check!( + "ff03::", + &[0xff, 3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], + multicast_realm_local | global + ); - check!("ff04::", &[0xff, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], multicast_admin_local); + check!( + "ff04::", + &[0xff, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], + multicast_admin_local | global + ); - check!("ff05::", &[0xff, 5, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], multicast_site_local); + check!( + "ff05::", + &[0xff, 5, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], + multicast_site_local | global + ); check!( "ff08::", &[0xff, 8, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], - multicast_organization_local + multicast_organization_local | global ); check!( @@ -698,7 +768,7 @@ fn ipv6_properties() { check!( "2001:2::ac32:23ff:21", &[0x20, 1, 0, 2, 0, 0, 0, 0, 0, 0, 0xac, 0x32, 0x23, 0xff, 0, 0x21], - global | unicast_global | benchmarking + benchmarking ); check!( diff --git a/library/std/src/net/mod.rs b/library/std/src/net/mod.rs index e7a40bdaf..01e3db9de 100644 --- a/library/std/src/net/mod.rs +++ b/library/std/src/net/mod.rs @@ -24,11 +24,11 @@ use crate::io::{self, ErrorKind}; #[stable(feature = "rust1", since = "1.0.0")] -pub use self::addr::{SocketAddr, SocketAddrV4, SocketAddrV6, ToSocketAddrs}; -#[stable(feature = "rust1", since = "1.0.0")] -pub use self::ip::{IpAddr, Ipv4Addr, Ipv6Addr, Ipv6MulticastScope}; +pub use self::ip_addr::{IpAddr, Ipv4Addr, Ipv6Addr, Ipv6MulticastScope}; #[stable(feature = "rust1", since = "1.0.0")] pub use self::parser::AddrParseError; +#[stable(feature = "rust1", since = "1.0.0")] +pub use self::socket_addr::{SocketAddr, SocketAddrV4, SocketAddrV6, ToSocketAddrs}; #[unstable(feature = "tcplistener_into_incoming", issue = "88339")] pub use self::tcp::IntoIncoming; #[stable(feature = "rust1", since = "1.0.0")] @@ -36,12 +36,13 @@ pub use self::tcp::{Incoming, TcpListener, TcpStream}; #[stable(feature = "rust1", since = "1.0.0")] pub use self::udp::UdpSocket; -mod addr; -mod ip; +mod display_buffer; +mod ip_addr; mod parser; +mod socket_addr; mod tcp; #[cfg(test)] -mod test; +pub(crate) mod test; mod udp; /// Possible values which can be passed to the [`TcpStream::shutdown`] method. diff --git a/library/std/src/net/parser.rs b/library/std/src/net/parser.rs index 069b66099..a38031c48 100644 --- a/library/std/src/net/parser.rs +++ b/library/std/src/net/parser.rs @@ -39,8 +39,8 @@ struct Parser<'a> { } impl<'a> Parser<'a> { - fn new(input: &'a str) -> Parser<'a> { - Parser { state: input.as_bytes() } + fn new(input: &'a [u8]) -> Parser<'a> { + Parser { state: input } } /// Run a parser, and restore the pre-parse state if it fails. @@ -273,32 +273,106 @@ impl<'a> Parser<'a> { } } +impl IpAddr { + /// Parse an IP address from a slice of bytes. + /// + /// ``` + /// #![feature(addr_parse_ascii)] + /// + /// use std::net::{IpAddr, Ipv4Addr, Ipv6Addr}; + /// + /// let localhost_v4 = IpAddr::V4(Ipv4Addr::new(127, 0, 0, 1)); + /// let localhost_v6 = IpAddr::V6(Ipv6Addr::new(0, 0, 0, 0, 0, 0, 0, 1)); + /// + /// assert_eq!(IpAddr::parse_ascii(b"127.0.0.1"), Ok(localhost_v4)); + /// assert_eq!(IpAddr::parse_ascii(b"::1"), Ok(localhost_v6)); + /// ``` + #[unstable(feature = "addr_parse_ascii", issue = "101035")] + pub fn parse_ascii(b: &[u8]) -> Result<Self, AddrParseError> { + Parser::new(b).parse_with(|p| p.read_ip_addr(), AddrKind::Ip) + } +} + #[stable(feature = "ip_addr", since = "1.7.0")] impl FromStr for IpAddr { type Err = AddrParseError; fn from_str(s: &str) -> Result<IpAddr, AddrParseError> { - Parser::new(s).parse_with(|p| p.read_ip_addr(), AddrKind::Ip) + Self::parse_ascii(s.as_bytes()) } } -#[stable(feature = "rust1", since = "1.0.0")] -impl FromStr for Ipv4Addr { - type Err = AddrParseError; - fn from_str(s: &str) -> Result<Ipv4Addr, AddrParseError> { +impl Ipv4Addr { + /// Parse an IPv4 address from a slice of bytes. + /// + /// ``` + /// #![feature(addr_parse_ascii)] + /// + /// use std::net::Ipv4Addr; + /// + /// let localhost = Ipv4Addr::new(127, 0, 0, 1); + /// + /// assert_eq!(Ipv4Addr::parse_ascii(b"127.0.0.1"), Ok(localhost)); + /// ``` + #[unstable(feature = "addr_parse_ascii", issue = "101035")] + pub fn parse_ascii(b: &[u8]) -> Result<Self, AddrParseError> { // don't try to parse if too long - if s.len() > 15 { + if b.len() > 15 { Err(AddrParseError(AddrKind::Ipv4)) } else { - Parser::new(s).parse_with(|p| p.read_ipv4_addr(), AddrKind::Ipv4) + Parser::new(b).parse_with(|p| p.read_ipv4_addr(), AddrKind::Ipv4) } } } #[stable(feature = "rust1", since = "1.0.0")] +impl FromStr for Ipv4Addr { + type Err = AddrParseError; + fn from_str(s: &str) -> Result<Ipv4Addr, AddrParseError> { + Self::parse_ascii(s.as_bytes()) + } +} + +impl Ipv6Addr { + /// Parse an IPv6 address from a slice of bytes. + /// + /// ``` + /// #![feature(addr_parse_ascii)] + /// + /// use std::net::Ipv6Addr; + /// + /// let localhost = Ipv6Addr::new(0, 0, 0, 0, 0, 0, 0, 1); + /// + /// assert_eq!(Ipv6Addr::parse_ascii(b"::1"), Ok(localhost)); + /// ``` + #[unstable(feature = "addr_parse_ascii", issue = "101035")] + pub fn parse_ascii(b: &[u8]) -> Result<Self, AddrParseError> { + Parser::new(b).parse_with(|p| p.read_ipv6_addr(), AddrKind::Ipv6) + } +} + +#[stable(feature = "rust1", since = "1.0.0")] impl FromStr for Ipv6Addr { type Err = AddrParseError; fn from_str(s: &str) -> Result<Ipv6Addr, AddrParseError> { - Parser::new(s).parse_with(|p| p.read_ipv6_addr(), AddrKind::Ipv6) + Self::parse_ascii(s.as_bytes()) + } +} + +impl SocketAddrV4 { + /// Parse an IPv4 socket address from a slice of bytes. + /// + /// ``` + /// #![feature(addr_parse_ascii)] + /// + /// use std::net::{Ipv4Addr, SocketAddrV4}; + /// + /// let socket = SocketAddrV4::new(Ipv4Addr::new(127, 0, 0, 1), 8080); + /// + /// assert_eq!(SocketAddrV4::parse_ascii(b"127.0.0.1:8080"), Ok(socket)); + /// ``` + #[unstable(feature = "addr_parse_ascii", issue = "101035")] + pub fn parse_ascii(b: &[u8]) -> Result<Self, AddrParseError> { + Parser::new(b).parse_with(|p| p.read_socket_addr_v4(), AddrKind::SocketV4) } } @@ -306,7 +380,25 @@ impl FromStr for Ipv6Addr { impl FromStr for SocketAddrV4 { type Err = AddrParseError; fn from_str(s: &str) -> Result<SocketAddrV4, AddrParseError> { - Parser::new(s).parse_with(|p| p.read_socket_addr_v4(), AddrKind::SocketV4) + Self::parse_ascii(s.as_bytes()) + } +} + +impl SocketAddrV6 { + /// Parse an IPv6 socket address from a slice of bytes. + /// + /// ``` + /// #![feature(addr_parse_ascii)] + /// + /// use std::net::{Ipv6Addr, SocketAddrV6}; + /// + /// let socket = SocketAddrV6::new(Ipv6Addr::new(0x2001, 0xdb8, 0, 0, 0, 0, 0, 1), 8080, 0, 0); + /// + /// assert_eq!(SocketAddrV6::parse_ascii(b"[2001:db8::1]:8080"), Ok(socket)); + /// ``` + #[unstable(feature = "addr_parse_ascii", issue = "101035")] + pub fn parse_ascii(b: &[u8]) -> Result<Self, AddrParseError> { + Parser::new(b).parse_with(|p| p.read_socket_addr_v6(), AddrKind::SocketV6) } } @@ -314,7 +406,27 @@ impl FromStr for SocketAddrV4 { impl FromStr for SocketAddrV6 { type Err = AddrParseError; fn from_str(s: &str) -> Result<SocketAddrV6, AddrParseError> { - Parser::new(s).parse_with(|p| p.read_socket_addr_v6(), AddrKind::SocketV6) + Self::parse_ascii(s.as_bytes()) + } +} + +impl SocketAddr { + /// Parse a socket address from a slice of bytes. + /// + /// ``` + /// #![feature(addr_parse_ascii)] + /// + /// use std::net::{IpAddr, Ipv4Addr, Ipv6Addr, SocketAddr}; + /// + /// let socket_v4 = SocketAddr::new(IpAddr::V4(Ipv4Addr::new(127, 0, 0, 1)), 8080); + /// let socket_v6 = SocketAddr::new(IpAddr::V6(Ipv6Addr::new(0, 0, 0, 0, 0, 0, 0, 1)), 8080); + /// + /// assert_eq!(SocketAddr::parse_ascii(b"127.0.0.1:8080"), Ok(socket_v4)); + /// assert_eq!(SocketAddr::parse_ascii(b"[::1]:8080"), Ok(socket_v6)); + /// ``` + #[unstable(feature = "addr_parse_ascii", issue = "101035")] + pub fn parse_ascii(b: &[u8]) -> Result<Self, AddrParseError> { + Parser::new(b).parse_with(|p| p.read_socket_addr(), AddrKind::Socket) } } @@ -322,7 +434,7 @@ impl FromStr for SocketAddrV6 { impl FromStr for SocketAddr { type Err = AddrParseError; fn from_str(s: &str) -> Result<SocketAddr, AddrParseError> { - Parser::new(s).parse_with(|p| p.read_socket_addr(), AddrKind::Socket) + Self::parse_ascii(s.as_bytes()) } } diff --git a/library/std/src/net/addr.rs b/library/std/src/net/socket_addr.rs index 53fee952a..33b0dfa03 100644 --- a/library/std/src/net/addr.rs +++ b/library/std/src/net/socket_addr.rs @@ -2,9 +2,9 @@ mod tests; use crate::cmp::Ordering; -use crate::fmt; +use crate::fmt::{self, Write}; use crate::hash; -use crate::io::{self, Write}; +use crate::io; use crate::iter; use crate::mem; use crate::net::{IpAddr, Ipv4Addr, Ipv6Addr}; @@ -15,6 +15,8 @@ use crate::sys_common::net::LookupHost; use crate::sys_common::{FromInner, IntoInner}; use crate::vec; +use super::display_buffer::DisplayBuffer; + /// An internet socket address, either IPv4 or IPv6. /// /// Internet socket addresses consist of an [IP address], a 16-bit port number, as well @@ -616,25 +618,18 @@ impl fmt::Debug for SocketAddr { #[stable(feature = "rust1", since = "1.0.0")] impl fmt::Display for SocketAddrV4 { fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { - // Fast path: if there's no alignment stuff, write to the output buffer - // directly + // If there are no alignment requirements, write the socket address directly to `f`. + // Otherwise, write it to a local buffer and then use `f.pad`. if f.precision().is_none() && f.width().is_none() { write!(f, "{}:{}", self.ip(), self.port()) } else { - const IPV4_SOCKET_BUF_LEN: usize = (3 * 4) // the segments - + 3 // the separators - + 1 + 5; // the port - let mut buf = [0; IPV4_SOCKET_BUF_LEN]; - let mut buf_slice = &mut buf[..]; - - // Unwrap is fine because writing to a sufficiently-sized - // buffer is infallible - write!(buf_slice, "{}:{}", self.ip(), self.port()).unwrap(); - let len = IPV4_SOCKET_BUF_LEN - buf_slice.len(); - - // This unsafe is OK because we know what is being written to the buffer - let buf = unsafe { crate::str::from_utf8_unchecked(&buf[..len]) }; - f.pad(buf) + const LONGEST_IPV4_SOCKET_ADDR: &str = "255.255.255.255:65536"; + + let mut buf = DisplayBuffer::<{ LONGEST_IPV4_SOCKET_ADDR.len() }>::new(); + // Buffer is long enough for the longest possible IPv4 socket address, so this should never fail. + write!(buf, "{}:{}", self.ip(), self.port()).unwrap(); + + f.pad(buf.as_str()) } } } @@ -649,35 +644,26 @@ impl fmt::Debug for SocketAddrV4 { #[stable(feature = "rust1", since = "1.0.0")] impl fmt::Display for SocketAddrV6 { fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { - // Fast path: if there's no alignment stuff, write to the output - // buffer directly + // If there are no alignment requirements, write the socket address directly to `f`. + // Otherwise, write it to a local buffer and then use `f.pad`. if f.precision().is_none() && f.width().is_none() { match self.scope_id() { 0 => write!(f, "[{}]:{}", self.ip(), self.port()), scope_id => write!(f, "[{}%{}]:{}", self.ip(), scope_id, self.port()), } } else { - const IPV6_SOCKET_BUF_LEN: usize = (4 * 8) // The address - + 7 // The colon separators - + 2 // The brackets - + 1 + 10 // The scope id - + 1 + 5; // The port - - let mut buf = [0; IPV6_SOCKET_BUF_LEN]; - let mut buf_slice = &mut buf[..]; + const LONGEST_IPV6_SOCKET_ADDR: &str = + "[ffff:ffff:ffff:ffff:ffff:ffff:ffff:ffff%4294967296]:65536"; + let mut buf = DisplayBuffer::<{ LONGEST_IPV6_SOCKET_ADDR.len() }>::new(); match self.scope_id() { - 0 => write!(buf_slice, "[{}]:{}", self.ip(), self.port()), - scope_id => write!(buf_slice, "[{}%{}]:{}", self.ip(), scope_id, self.port()), + 0 => write!(buf, "[{}]:{}", self.ip(), self.port()), + scope_id => write!(buf, "[{}%{}]:{}", self.ip(), scope_id, self.port()), } - // Unwrap is fine because writing to a sufficiently-sized - // buffer is infallible + // Buffer is long enough for the longest possible IPv6 socket address, so this should never fail. .unwrap(); - let len = IPV6_SOCKET_BUF_LEN - buf_slice.len(); - // This unsafe is OK because we know what is being written to the buffer - let buf = unsafe { crate::str::from_utf8_unchecked(&buf[..len]) }; - f.pad(buf) + f.pad(buf.as_str()) } } } diff --git a/library/std/src/net/addr/tests.rs b/library/std/src/net/socket_addr/tests.rs index 585a17451..15211f819 100644 --- a/library/std/src/net/addr/tests.rs +++ b/library/std/src/net/socket_addr/tests.rs @@ -52,6 +52,75 @@ fn to_socket_addr_string() { } #[test] +fn ipv4_socket_addr_to_string() { + // Shortest possible IPv4 length. + assert_eq!(SocketAddrV4::new(Ipv4Addr::new(0, 0, 0, 0), 0).to_string(), "0.0.0.0:0"); + + // Longest possible IPv4 length. + assert_eq!( + SocketAddrV4::new(Ipv4Addr::new(255, 255, 255, 255), u16::MAX).to_string(), + "255.255.255.255:65535" + ); + + // Test padding. + assert_eq!( + &format!("{:16}", SocketAddrV4::new(Ipv4Addr::new(1, 1, 1, 1), 53)), + "1.1.1.1:53 " + ); + assert_eq!( + &format!("{:>16}", SocketAddrV4::new(Ipv4Addr::new(1, 1, 1, 1), 53)), + " 1.1.1.1:53" + ); +} + +#[test] +fn ipv6_socket_addr_to_string() { + // IPv4-mapped address. + assert_eq!( + SocketAddrV6::new(Ipv6Addr::new(0, 0, 0, 0, 0, 0xffff, 0xc000, 0x280), 8080, 0, 0) + .to_string(), + "[::ffff:192.0.2.128]:8080" + ); + + // IPv4-compatible address. + assert_eq!( + SocketAddrV6::new(Ipv6Addr::new(0, 0, 0, 0, 0, 0, 0xc000, 0x280), 8080, 0, 0).to_string(), + "[::192.0.2.128]:8080" + ); + + // IPv6 address with no zero segments. + assert_eq!( + SocketAddrV6::new(Ipv6Addr::new(8, 9, 10, 11, 12, 13, 14, 15), 80, 0, 0).to_string(), + "[8:9:a:b:c:d:e:f]:80" + ); + + // Shortest possible IPv6 length. + assert_eq!(SocketAddrV6::new(Ipv6Addr::UNSPECIFIED, 0, 0, 0).to_string(), "[::]:0"); + + // Longest possible IPv6 length. + assert_eq!( + SocketAddrV6::new( + Ipv6Addr::new(0x1111, 0x2222, 0x3333, 0x4444, 0x5555, 0x6666, 0x7777, 0x8888), + u16::MAX, + u32::MAX, + u32::MAX, + ) + .to_string(), + "[1111:2222:3333:4444:5555:6666:7777:8888%4294967295]:65535" + ); + + // Test padding. + assert_eq!( + &format!("{:22}", SocketAddrV6::new(Ipv6Addr::new(1, 2, 3, 4, 5, 6, 7, 8), 9, 0, 0)), + "[1:2:3:4:5:6:7:8]:9 " + ); + assert_eq!( + &format!("{:>22}", SocketAddrV6::new(Ipv6Addr::new(1, 2, 3, 4, 5, 6, 7, 8), 9, 0, 0)), + " [1:2:3:4:5:6:7:8]:9" + ); +} + +#[test] fn bind_udp_socket_bad() { // rust-lang/rust#53957: This is a regression test for a parsing problem // discovered as part of issue rust-lang/rust#23076, where we were diff --git a/library/std/src/os/android/mod.rs b/library/std/src/os/android/mod.rs index dbb0127f3..5adcb82b6 100644 --- a/library/std/src/os/android/mod.rs +++ b/library/std/src/os/android/mod.rs @@ -3,4 +3,5 @@ #![stable(feature = "raw_ext", since = "1.1.0")] pub mod fs; +pub mod net; pub mod raw; diff --git a/library/std/src/os/android/net.rs b/library/std/src/os/android/net.rs new file mode 100644 index 000000000..ff96125c3 --- /dev/null +++ b/library/std/src/os/android/net.rs @@ -0,0 +1,4 @@ +//! Linux and Android-specific definitions for socket options. + +#![unstable(feature = "tcp_quickack", issue = "96256")] +pub use crate::os::net::tcp::TcpStreamExt; diff --git a/library/std/src/os/fd/owned.rs b/library/std/src/os/fd/owned.rs index a463bc41d..71e33fb9e 100644 --- a/library/std/src/os/fd/owned.rs +++ b/library/std/src/os/fd/owned.rs @@ -104,7 +104,8 @@ impl BorrowedFd<'_> { #[cfg(target_os = "espidf")] let cmd = libc::F_DUPFD; - let fd = cvt(unsafe { libc::fcntl(self.as_raw_fd(), cmd, 0) })?; + // Avoid using file descriptors below 3 as they are used for stdio + let fd = cvt(unsafe { libc::fcntl(self.as_raw_fd(), cmd, 3) })?; Ok(unsafe { OwnedFd::from_raw_fd(fd) }) } diff --git a/library/std/src/os/fd/raw.rs b/library/std/src/os/fd/raw.rs index 081915ed1..1b3d11042 100644 --- a/library/std/src/os/fd/raw.rs +++ b/library/std/src/os/fd/raw.rs @@ -14,7 +14,7 @@ use crate::os::wasi::io::OwnedFd; use crate::sys_common::{AsInner, IntoInner}; /// Raw file descriptors. -#[cfg_attr(not(bootstrap), rustc_allowed_through_unstable_modules)] +#[rustc_allowed_through_unstable_modules] #[stable(feature = "rust1", since = "1.0.0")] pub type RawFd = raw::c_int; @@ -23,7 +23,7 @@ pub type RawFd = raw::c_int; /// This is only available on unix and WASI platforms and must be imported in /// order to call the method. Windows platforms have a corresponding /// `AsRawHandle` and `AsRawSocket` set of traits. -#[cfg_attr(not(bootstrap), rustc_allowed_through_unstable_modules)] +#[rustc_allowed_through_unstable_modules] #[stable(feature = "rust1", since = "1.0.0")] pub trait AsRawFd { /// Extracts the raw file descriptor. @@ -59,7 +59,7 @@ pub trait AsRawFd { /// A trait to express the ability to construct an object from a raw file /// descriptor. -#[cfg_attr(not(bootstrap), rustc_allowed_through_unstable_modules)] +#[rustc_allowed_through_unstable_modules] #[stable(feature = "from_raw_os", since = "1.1.0")] pub trait FromRawFd { /// Constructs a new instance of `Self` from the given raw file @@ -103,7 +103,7 @@ pub trait FromRawFd { /// A trait to express the ability to consume an object and acquire ownership of /// its raw file descriptor. -#[cfg_attr(not(bootstrap), rustc_allowed_through_unstable_modules)] +#[rustc_allowed_through_unstable_modules] #[stable(feature = "into_raw_os", since = "1.4.0")] pub trait IntoRawFd { /// Consumes this object, returning the raw underlying file descriptor. diff --git a/library/std/src/os/fortanix_sgx/mod.rs b/library/std/src/os/fortanix_sgx/mod.rs index a40dabe19..39a42f4e1 100644 --- a/library/std/src/os/fortanix_sgx/mod.rs +++ b/library/std/src/os/fortanix_sgx/mod.rs @@ -26,10 +26,13 @@ pub mod usercalls { free, insecure_time, launch_thread, read, read_alloc, send, wait, write, }; pub use crate::sys::abi::usercalls::raw::{do_usercall, Usercalls as UsercallNrs}; + pub use crate::sys::abi::usercalls::raw::{Register, RegisterArgument, ReturnValue}; // fortanix-sgx-abi re-exports pub use crate::sys::abi::usercalls::raw::Error; - pub use crate::sys::abi::usercalls::raw::{ByteBuffer, FifoDescriptor, Return, Usercall}; + pub use crate::sys::abi::usercalls::raw::{ + ByteBuffer, Cancel, FifoDescriptor, Return, Usercall, + }; pub use crate::sys::abi::usercalls::raw::{Fd, Result, Tcs}; pub use crate::sys::abi::usercalls::raw::{ EV_RETURNQ_NOT_EMPTY, EV_UNPARK, EV_USERCALLQ_NOT_FULL, FD_STDERR, FD_STDIN, FD_STDOUT, diff --git a/library/std/src/os/linux/mod.rs b/library/std/src/os/linux/mod.rs index 8e7776f66..c17053011 100644 --- a/library/std/src/os/linux/mod.rs +++ b/library/std/src/os/linux/mod.rs @@ -4,5 +4,6 @@ #![doc(cfg(target_os = "linux"))] pub mod fs; +pub mod net; pub mod process; pub mod raw; diff --git a/library/std/src/os/linux/net.rs b/library/std/src/os/linux/net.rs new file mode 100644 index 000000000..ff96125c3 --- /dev/null +++ b/library/std/src/os/linux/net.rs @@ -0,0 +1,4 @@ +//! Linux and Android-specific definitions for socket options. + +#![unstable(feature = "tcp_quickack", issue = "96256")] +pub use crate::os::net::tcp::TcpStreamExt; diff --git a/library/std/src/os/mod.rs b/library/std/src/os/mod.rs index 6fbaa42c7..18c64b510 100644 --- a/library/std/src/os/mod.rs +++ b/library/std/src/os/mod.rs @@ -148,3 +148,6 @@ pub mod vxworks; #[cfg(any(unix, target_os = "wasi", doc))] mod fd; + +#[cfg(any(target_os = "linux", target_os = "android", doc))] +mod net; diff --git a/library/std/src/os/net/mod.rs b/library/std/src/os/net/mod.rs new file mode 100644 index 000000000..d6d84d24e --- /dev/null +++ b/library/std/src/os/net/mod.rs @@ -0,0 +1,7 @@ +//! Linux and Android-specific definitions for socket options. + +#![unstable(feature = "tcp_quickack", issue = "96256")] +#![doc(cfg(any(target_os = "linux", target_os = "android",)))] +pub mod tcp; +#[cfg(test)] +mod tests; diff --git a/library/std/src/os/net/tcp.rs b/library/std/src/os/net/tcp.rs new file mode 100644 index 000000000..5e9ee65a4 --- /dev/null +++ b/library/std/src/os/net/tcp.rs @@ -0,0 +1,70 @@ +//! Linux and Android-specific tcp extensions to primitives in the [`std::net`] module. +//! +//! [`std::net`]: crate::net + +use crate::io; +use crate::net; +use crate::sealed::Sealed; +use crate::sys_common::AsInner; + +/// Os-specific extensions for [`TcpStream`] +/// +/// [`TcpStream`]: net::TcpStream +#[unstable(feature = "tcp_quickack", issue = "96256")] +pub trait TcpStreamExt: Sealed { + /// Enable or disable `TCP_QUICKACK`. + /// + /// This flag causes Linux to eagerly send ACKs rather than delaying them. + /// Linux may reset this flag after further operations on the socket. + /// + /// See [`man 7 tcp`](https://man7.org/linux/man-pages/man7/tcp.7.html) and + /// [TCP delayed acknowledgement](https://en.wikipedia.org/wiki/TCP_delayed_acknowledgment) + /// for more information. + /// + /// # Examples + /// + /// ```no_run + /// #![feature(tcp_quickack)] + /// use std::net::TcpStream; + /// use std::os::linux::net::TcpStreamExt; + /// + /// let stream = TcpStream::connect("127.0.0.1:8080") + /// .expect("Couldn't connect to the server..."); + /// stream.set_quickack(true).expect("set_quickack call failed"); + /// ``` + #[unstable(feature = "tcp_quickack", issue = "96256")] + fn set_quickack(&self, quickack: bool) -> io::Result<()>; + + /// Gets the value of the `TCP_QUICKACK` option on this socket. + /// + /// For more information about this option, see [`TcpStreamExt::set_quickack`]. + /// + /// # Examples + /// + /// ```no_run + /// #![feature(tcp_quickack)] + /// use std::net::TcpStream; + /// use std::os::linux::net::TcpStreamExt; + /// + /// let stream = TcpStream::connect("127.0.0.1:8080") + /// .expect("Couldn't connect to the server..."); + /// stream.set_quickack(true).expect("set_quickack call failed"); + /// assert_eq!(stream.quickack().unwrap_or(false), true); + /// ``` + #[unstable(feature = "tcp_quickack", issue = "96256")] + fn quickack(&self) -> io::Result<bool>; +} + +#[unstable(feature = "tcp_quickack", issue = "96256")] +impl Sealed for net::TcpStream {} + +#[unstable(feature = "tcp_quickack", issue = "96256")] +impl TcpStreamExt for net::TcpStream { + fn set_quickack(&self, quickack: bool) -> io::Result<()> { + self.as_inner().as_inner().set_quickack(quickack) + } + + fn quickack(&self) -> io::Result<bool> { + self.as_inner().as_inner().quickack() + } +} diff --git a/library/std/src/os/net/tests.rs b/library/std/src/os/net/tests.rs new file mode 100644 index 000000000..4704e3156 --- /dev/null +++ b/library/std/src/os/net/tests.rs @@ -0,0 +1,29 @@ +#[cfg(any(target_os = "android", target_os = "linux",))] +#[test] +fn quickack() { + use crate::{ + net::{test::next_test_ip4, TcpListener, TcpStream}, + os::net::tcp::TcpStreamExt, + }; + + macro_rules! t { + ($e:expr) => { + match $e { + Ok(t) => t, + Err(e) => panic!("received error for `{}`: {}", stringify!($e), e), + } + }; + } + + let addr = next_test_ip4(); + let _listener = t!(TcpListener::bind(&addr)); + + let stream = t!(TcpStream::connect(&("localhost", addr.port()))); + + t!(stream.set_quickack(false)); + assert_eq!(false, t!(stream.quickack())); + t!(stream.set_quickack(true)); + assert_eq!(true, t!(stream.quickack())); + t!(stream.set_quickack(false)); + assert_eq!(false, t!(stream.quickack())); +} diff --git a/library/std/src/os/unix/net/addr.rs b/library/std/src/os/unix/net/addr.rs index 9aeae4b2c..094085e19 100644 --- a/library/std/src/os/unix/net/addr.rs +++ b/library/std/src/os/unix/net/addr.rs @@ -2,7 +2,7 @@ use crate::ffi::OsStr; use crate::os::unix::ffi::OsStrExt; use crate::path::Path; use crate::sys::cvt; -use crate::{ascii, fmt, io, mem, ptr}; +use crate::{fmt, io, mem, ptr}; // FIXME(#43348): Make libc adapt #[doc(cfg(...))] so we don't need these fake definitions here? #[cfg(not(unix))] @@ -64,18 +64,6 @@ enum AddressKind<'a> { Abstract(&'a [u8]), } -struct AsciiEscaped<'a>(&'a [u8]); - -impl<'a> fmt::Display for AsciiEscaped<'a> { - fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result { - write!(fmt, "\"")?; - for byte in self.0.iter().cloned().flat_map(ascii::escape_default) { - write!(fmt, "{}", byte as char)?; - } - write!(fmt, "\"") - } -} - /// An address associated with a Unix socket. /// /// # Examples @@ -329,7 +317,7 @@ impl SocketAddr { crate::ptr::copy_nonoverlapping( namespace.as_ptr(), - addr.sun_path.as_mut_ptr().offset(1) as *mut u8, + addr.sun_path.as_mut_ptr().add(1) as *mut u8, namespace.len(), ); let len = (sun_path_offset(&addr) + 1 + namespace.len()) as libc::socklen_t; @@ -343,7 +331,7 @@ impl fmt::Debug for SocketAddr { fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result { match self.address() { AddressKind::Unnamed => write!(fmt, "(unnamed)"), - AddressKind::Abstract(name) => write!(fmt, "{} (abstract)", AsciiEscaped(name)), + AddressKind::Abstract(name) => write!(fmt, "\"{}\" (abstract)", name.escape_ascii()), AddressKind::Pathname(path) => write!(fmt, "{path:?} (pathname)"), } } diff --git a/library/std/src/os/unix/net/datagram.rs b/library/std/src/os/unix/net/datagram.rs index 8008acfd1..f758f88d0 100644 --- a/library/std/src/os/unix/net/datagram.rs +++ b/library/std/src/os/unix/net/datagram.rs @@ -838,6 +838,31 @@ impl UnixDatagram { self.0.passcred() } + /// Set the id of the socket for network filtering purpose + /// + #[cfg_attr( + any(target_os = "linux", target_os = "freebsd", target_os = "openbsd"), + doc = "```no_run" + )] + #[cfg_attr( + not(any(target_os = "linux", target_os = "freebsd", target_os = "openbsd")), + doc = "```ignore" + )] + /// #![feature(unix_set_mark)] + /// use std::os::unix::net::UnixDatagram; + /// + /// fn main() -> std::io::Result<()> { + /// let sock = UnixDatagram::unbound()?; + /// sock.set_mark(32)?; + /// Ok(()) + /// } + /// ``` + #[cfg(any(doc, target_os = "linux", target_os = "freebsd", target_os = "openbsd",))] + #[unstable(feature = "unix_set_mark", issue = "96467")] + pub fn set_mark(&self, mark: u32) -> io::Result<()> { + self.0.set_mark(mark) + } + /// Returns the value of the `SO_ERROR` option. /// /// # Examples diff --git a/library/std/src/os/unix/net/listener.rs b/library/std/src/os/unix/net/listener.rs index 7c0d53950..02090afc8 100644 --- a/library/std/src/os/unix/net/listener.rs +++ b/library/std/src/os/unix/net/listener.rs @@ -73,9 +73,11 @@ impl UnixListener { unsafe { let inner = Socket::new_raw(libc::AF_UNIX, libc::SOCK_STREAM)?; let (addr, len) = sockaddr_un(path.as_ref())?; + const backlog: libc::c_int = + if cfg!(any(target_os = "linux", target_os = "freebsd")) { -1 } else { 128 }; cvt(libc::bind(inner.as_inner().as_raw_fd(), &addr as *const _ as *const _, len as _))?; - cvt(libc::listen(inner.as_inner().as_raw_fd(), 128))?; + cvt(libc::listen(inner.as_inner().as_raw_fd(), backlog))?; Ok(UnixListener(inner)) } @@ -109,12 +111,16 @@ impl UnixListener { pub fn bind_addr(socket_addr: &SocketAddr) -> io::Result<UnixListener> { unsafe { let inner = Socket::new_raw(libc::AF_UNIX, libc::SOCK_STREAM)?; + #[cfg(target_os = "linux")] + const backlog: libc::c_int = -1; + #[cfg(not(target_os = "linux"))] + const backlog: libc::c_int = 128; cvt(libc::bind( inner.as_raw_fd(), &socket_addr.addr as *const _ as *const _, socket_addr.len as _, ))?; - cvt(libc::listen(inner.as_raw_fd(), 128))?; + cvt(libc::listen(inner.as_raw_fd(), backlog))?; Ok(UnixListener(inner)) } } diff --git a/library/std/src/os/unix/net/stream.rs b/library/std/src/os/unix/net/stream.rs index cc3a88587..dff8f6e85 100644 --- a/library/std/src/os/unix/net/stream.rs +++ b/library/std/src/os/unix/net/stream.rs @@ -427,6 +427,31 @@ impl UnixStream { self.0.passcred() } + /// Set the id of the socket for network filtering purpose + /// + #[cfg_attr( + any(target_os = "linux", target_os = "freebsd", target_os = "openbsd"), + doc = "```no_run" + )] + #[cfg_attr( + not(any(target_os = "linux", target_os = "freebsd", target_os = "openbsd")), + doc = "```ignore" + )] + /// #![feature(unix_set_mark)] + /// use std::os::unix::net::UnixStream; + /// + /// fn main() -> std::io::Result<()> { + /// let sock = UnixStream::connect("/tmp/sock")?; + /// sock.set_mark(32)?; + /// Ok(()) + /// } + /// ``` + #[cfg(any(doc, target_os = "linux", target_os = "freebsd", target_os = "openbsd",))] + #[unstable(feature = "unix_set_mark", issue = "96467")] + pub fn set_mark(&self, mark: u32) -> io::Result<()> { + self.0.set_mark(mark) + } + /// Returns the value of the `SO_ERROR` option. /// /// # Examples diff --git a/library/std/src/os/wasi/io/fd.rs b/library/std/src/os/wasi/io/fd.rs index 930aca887..75703af6a 100644 --- a/library/std/src/os/wasi/io/fd.rs +++ b/library/std/src/os/wasi/io/fd.rs @@ -1,9 +1,10 @@ //! Owned and borrowed file descriptors. -#![unstable(feature = "wasi_ext", issue = "71213")] +#![stable(feature = "io_safety_wasi", since = "1.65.0")] // Tests for this module #[cfg(test)] mod tests; +#[stable(feature = "io_safety_wasi", since = "1.65.0")] pub use crate::os::fd::owned::*; diff --git a/library/std/src/os/wasi/io/mod.rs b/library/std/src/os/wasi/io/mod.rs index 6c884e2ea..4f5cfbf9a 100644 --- a/library/std/src/os/wasi/io/mod.rs +++ b/library/std/src/os/wasi/io/mod.rs @@ -1,12 +1,12 @@ //! WASI-specific extensions to general I/O primitives. #![deny(unsafe_op_in_unsafe_fn)] -#![unstable(feature = "wasi_ext", issue = "71213")] +#![stable(feature = "io_safety_wasi", since = "1.65.0")] mod fd; mod raw; -#[unstable(feature = "wasi_ext", issue = "71213")] +#[stable(feature = "io_safety_wasi", since = "1.65.0")] pub use fd::*; -#[unstable(feature = "wasi_ext", issue = "71213")] +#[stable(feature = "io_safety_wasi", since = "1.65.0")] pub use raw::*; diff --git a/library/std/src/os/wasi/io/raw.rs b/library/std/src/os/wasi/io/raw.rs index da3b36ada..4ac792ee8 100644 --- a/library/std/src/os/wasi/io/raw.rs +++ b/library/std/src/os/wasi/io/raw.rs @@ -1,20 +1,6 @@ //! WASI-specific extensions to general I/O primitives. -#![unstable(feature = "wasi_ext", issue = "71213")] +#![stable(feature = "io_safety_wasi", since = "1.65.0")] -// NOTE: despite the fact that this module is unstable, -// stable Rust had the capability to access the stable -// re-exported items from os::fd::raw through this -// unstable module. -// In PR #95956 the stability checker was changed to check -// all path segments of an item rather than just the last, -// which caused the aforementioned stable usage to regress -// (see issue #99502). -// As a result, the items in os::fd::raw were given the -// rustc_allowed_through_unstable_modules attribute. -// No regression tests were added to ensure this property, -// as CI is not configured to test wasm32-wasi. -// If this module is stabilized, -// you may want to remove those attributes -// (assuming no other unstable modules need them). +#[stable(feature = "io_safety_wasi", since = "1.65.0")] pub use crate::os::fd::raw::*; diff --git a/library/std/src/panic.rs b/library/std/src/panic.rs index 45bc56efb..c4f022de0 100644 --- a/library/std/src/panic.rs +++ b/library/std/src/panic.rs @@ -295,23 +295,22 @@ pub fn get_backtrace_style() -> Option<BacktraceStyle> { return Some(style); } - // Setting environment variables for Fuchsia components isn't a standard - // or easily supported workflow. For now, display backtraces by default. - let format = if cfg!(target_os = "fuchsia") { - BacktraceStyle::Full - } else { - crate::env::var_os("RUST_BACKTRACE") - .map(|x| { - if &x == "0" { - BacktraceStyle::Off - } else if &x == "full" { - BacktraceStyle::Full - } else { - BacktraceStyle::Short - } - }) - .unwrap_or(BacktraceStyle::Off) - }; + let format = crate::env::var_os("RUST_BACKTRACE") + .map(|x| { + if &x == "0" { + BacktraceStyle::Off + } else if &x == "full" { + BacktraceStyle::Full + } else { + BacktraceStyle::Short + } + }) + .unwrap_or(if cfg!(target_os = "fuchsia") { + // Fuchsia components default to full backtrace. + BacktraceStyle::Full + } else { + BacktraceStyle::Off + }); set_backtrace_style(format); Some(format) } diff --git a/library/std/src/path/tests.rs b/library/std/src/path/tests.rs index 351cf6988..dd307022c 100644 --- a/library/std/src/path/tests.rs +++ b/library/std/src/path/tests.rs @@ -1768,6 +1768,7 @@ fn test_windows_absolute() { } #[bench] +#[cfg_attr(miri, ignore)] // Miri isn't fast... fn bench_path_cmp_fast_path_buf_sort(b: &mut test::Bencher) { let prefix = "my/home"; let mut paths: Vec<_> = @@ -1781,6 +1782,7 @@ fn bench_path_cmp_fast_path_buf_sort(b: &mut test::Bencher) { } #[bench] +#[cfg_attr(miri, ignore)] // Miri isn't fast... fn bench_path_cmp_fast_path_long(b: &mut test::Bencher) { let prefix = "/my/home/is/my/castle/and/my/castle/has/a/rusty/workbench/"; let paths: Vec<_> = @@ -1799,6 +1801,7 @@ fn bench_path_cmp_fast_path_long(b: &mut test::Bencher) { } #[bench] +#[cfg_attr(miri, ignore)] // Miri isn't fast... fn bench_path_cmp_fast_path_short(b: &mut test::Bencher) { let prefix = "my/home"; let paths: Vec<_> = @@ -1817,6 +1820,7 @@ fn bench_path_cmp_fast_path_short(b: &mut test::Bencher) { } #[bench] +#[cfg_attr(miri, ignore)] // Miri isn't fast... fn bench_path_hashset(b: &mut test::Bencher) { let prefix = "/my/home/is/my/castle/and/my/castle/has/a/rusty/workbench/"; let paths: Vec<_> = @@ -1835,6 +1839,7 @@ fn bench_path_hashset(b: &mut test::Bencher) { } #[bench] +#[cfg_attr(miri, ignore)] // Miri isn't fast... fn bench_path_hashset_miss(b: &mut test::Bencher) { let prefix = "/my/home/is/my/castle/and/my/castle/has/a/rusty/workbench/"; let paths: Vec<_> = diff --git a/library/std/src/personality.rs b/library/std/src/personality.rs new file mode 100644 index 000000000..63f0ad4f1 --- /dev/null +++ b/library/std/src/personality.rs @@ -0,0 +1,46 @@ +//! This module contains the implementation of the `eh_personality` lang item. +//! +//! The actual implementation is heavily dependent on the target since Rust +//! tries to use the native stack unwinding mechanism whenever possible. +//! +//! This personality function is still required with `-C panic=abort` because +//! it is used to catch foreign exceptions from `extern "C-unwind"` and turn +//! them into aborts. +//! +//! Additionally, ARM EHABI uses the personality function when generating +//! backtraces. + +mod dwarf; + +#[cfg(not(test))] +cfg_if::cfg_if! { + if #[cfg(target_os = "emscripten")] { + mod emcc; + } else if #[cfg(target_env = "msvc")] { + // This is required by the compiler to exist (e.g., it's a lang item), + // but it's never actually called by the compiler because + // _CxxFrameHandler3 is the personality function that is always used. + // Hence this is just an aborting stub. + #[lang = "eh_personality"] + fn rust_eh_personality() { + core::intrinsics::abort() + } + } else if #[cfg(any( + all(target_family = "windows", target_env = "gnu"), + target_os = "psp", + target_os = "solid_asp3", + all(target_family = "unix", not(target_os = "espidf")), + all(target_vendor = "fortanix", target_env = "sgx"), + ))] { + mod gcc; + } else { + // Targets that don't support unwinding. + // - family=wasm + // - os=none ("bare metal" targets) + // - os=uefi + // - os=espidf + // - os=hermit + // - nvptx64-nvidia-cuda + // - arch=avr + } +} diff --git a/library/std/src/personality/dwarf/eh.rs b/library/std/src/personality/dwarf/eh.rs new file mode 100644 index 000000000..8799137b7 --- /dev/null +++ b/library/std/src/personality/dwarf/eh.rs @@ -0,0 +1,192 @@ +//! Parsing of GCC-style Language-Specific Data Area (LSDA) +//! For details see: +//! * <https://refspecs.linuxfoundation.org/LSB_3.0.0/LSB-PDA/LSB-PDA/ehframechpt.html> +//! * <https://itanium-cxx-abi.github.io/cxx-abi/exceptions.pdf> +//! * <https://www.airs.com/blog/archives/460> +//! * <https://www.airs.com/blog/archives/464> +//! +//! A reference implementation may be found in the GCC source tree +//! (`<root>/libgcc/unwind-c.c` as of this writing). + +#![allow(non_upper_case_globals)] +#![allow(unused)] + +use super::DwarfReader; +use core::mem; + +pub const DW_EH_PE_omit: u8 = 0xFF; +pub const DW_EH_PE_absptr: u8 = 0x00; + +pub const DW_EH_PE_uleb128: u8 = 0x01; +pub const DW_EH_PE_udata2: u8 = 0x02; +pub const DW_EH_PE_udata4: u8 = 0x03; +pub const DW_EH_PE_udata8: u8 = 0x04; +pub const DW_EH_PE_sleb128: u8 = 0x09; +pub const DW_EH_PE_sdata2: u8 = 0x0A; +pub const DW_EH_PE_sdata4: u8 = 0x0B; +pub const DW_EH_PE_sdata8: u8 = 0x0C; + +pub const DW_EH_PE_pcrel: u8 = 0x10; +pub const DW_EH_PE_textrel: u8 = 0x20; +pub const DW_EH_PE_datarel: u8 = 0x30; +pub const DW_EH_PE_funcrel: u8 = 0x40; +pub const DW_EH_PE_aligned: u8 = 0x50; + +pub const DW_EH_PE_indirect: u8 = 0x80; + +#[derive(Copy, Clone)] +pub struct EHContext<'a> { + pub ip: usize, // Current instruction pointer + pub func_start: usize, // Address of the current function + pub get_text_start: &'a dyn Fn() -> usize, // Get address of the code section + pub get_data_start: &'a dyn Fn() -> usize, // Get address of the data section +} + +pub enum EHAction { + None, + Cleanup(usize), + Catch(usize), + Terminate, +} + +pub const USING_SJLJ_EXCEPTIONS: bool = cfg!(all(target_os = "ios", target_arch = "arm")); + +pub unsafe fn find_eh_action(lsda: *const u8, context: &EHContext<'_>) -> Result<EHAction, ()> { + if lsda.is_null() { + return Ok(EHAction::None); + } + + let func_start = context.func_start; + let mut reader = DwarfReader::new(lsda); + + let start_encoding = reader.read::<u8>(); + // base address for landing pad offsets + let lpad_base = if start_encoding != DW_EH_PE_omit { + read_encoded_pointer(&mut reader, context, start_encoding)? + } else { + func_start + }; + + let ttype_encoding = reader.read::<u8>(); + if ttype_encoding != DW_EH_PE_omit { + // Rust doesn't analyze exception types, so we don't care about the type table + reader.read_uleb128(); + } + + let call_site_encoding = reader.read::<u8>(); + let call_site_table_length = reader.read_uleb128(); + let action_table = reader.ptr.add(call_site_table_length as usize); + let ip = context.ip; + + if !USING_SJLJ_EXCEPTIONS { + while reader.ptr < action_table { + let cs_start = read_encoded_pointer(&mut reader, context, call_site_encoding)?; + let cs_len = read_encoded_pointer(&mut reader, context, call_site_encoding)?; + let cs_lpad = read_encoded_pointer(&mut reader, context, call_site_encoding)?; + let cs_action = reader.read_uleb128(); + // Callsite table is sorted by cs_start, so if we've passed the ip, we + // may stop searching. + if ip < func_start + cs_start { + break; + } + if ip < func_start + cs_start + cs_len { + if cs_lpad == 0 { + return Ok(EHAction::None); + } else { + let lpad = lpad_base + cs_lpad; + return Ok(interpret_cs_action(cs_action, lpad)); + } + } + } + // Ip is not present in the table. This should not happen... but it does: issue #35011. + // So rather than returning EHAction::Terminate, we do this. + Ok(EHAction::None) + } else { + // SjLj version: + // The "IP" is an index into the call-site table, with two exceptions: + // -1 means 'no-action', and 0 means 'terminate'. + match ip as isize { + -1 => return Ok(EHAction::None), + 0 => return Ok(EHAction::Terminate), + _ => (), + } + let mut idx = ip; + loop { + let cs_lpad = reader.read_uleb128(); + let cs_action = reader.read_uleb128(); + idx -= 1; + if idx == 0 { + // Can never have null landing pad for sjlj -- that would have + // been indicated by a -1 call site index. + let lpad = (cs_lpad + 1) as usize; + return Ok(interpret_cs_action(cs_action, lpad)); + } + } + } +} + +fn interpret_cs_action(cs_action: u64, lpad: usize) -> EHAction { + if cs_action == 0 { + // If cs_action is 0 then this is a cleanup (Drop::drop). We run these + // for both Rust panics and foreign exceptions. + EHAction::Cleanup(lpad) + } else { + // Stop unwinding Rust panics at catch_unwind. + EHAction::Catch(lpad) + } +} + +#[inline] +fn round_up(unrounded: usize, align: usize) -> Result<usize, ()> { + if align.is_power_of_two() { Ok((unrounded + align - 1) & !(align - 1)) } else { Err(()) } +} + +unsafe fn read_encoded_pointer( + reader: &mut DwarfReader, + context: &EHContext<'_>, + encoding: u8, +) -> Result<usize, ()> { + if encoding == DW_EH_PE_omit { + return Err(()); + } + + // DW_EH_PE_aligned implies it's an absolute pointer value + if encoding == DW_EH_PE_aligned { + reader.ptr = round_up(reader.ptr as usize, mem::size_of::<usize>())? as *const u8; + return Ok(reader.read::<usize>()); + } + + let mut result = match encoding & 0x0F { + DW_EH_PE_absptr => reader.read::<usize>(), + DW_EH_PE_uleb128 => reader.read_uleb128() as usize, + DW_EH_PE_udata2 => reader.read::<u16>() as usize, + DW_EH_PE_udata4 => reader.read::<u32>() as usize, + DW_EH_PE_udata8 => reader.read::<u64>() as usize, + DW_EH_PE_sleb128 => reader.read_sleb128() as usize, + DW_EH_PE_sdata2 => reader.read::<i16>() as usize, + DW_EH_PE_sdata4 => reader.read::<i32>() as usize, + DW_EH_PE_sdata8 => reader.read::<i64>() as usize, + _ => return Err(()), + }; + + result += match encoding & 0x70 { + DW_EH_PE_absptr => 0, + // relative to address of the encoded value, despite the name + DW_EH_PE_pcrel => reader.ptr as usize, + DW_EH_PE_funcrel => { + if context.func_start == 0 { + return Err(()); + } + context.func_start + } + DW_EH_PE_textrel => (*context.get_text_start)(), + DW_EH_PE_datarel => (*context.get_data_start)(), + _ => return Err(()), + }; + + if encoding & DW_EH_PE_indirect != 0 { + result = *(result as *const usize); + } + + Ok(result) +} diff --git a/library/std/src/personality/dwarf/mod.rs b/library/std/src/personality/dwarf/mod.rs new file mode 100644 index 000000000..652fbe95a --- /dev/null +++ b/library/std/src/personality/dwarf/mod.rs @@ -0,0 +1,73 @@ +//! Utilities for parsing DWARF-encoded data streams. +//! See <http://www.dwarfstd.org>, +//! DWARF-4 standard, Section 7 - "Data Representation" + +// This module is used only by x86_64-pc-windows-gnu for now, but we +// are compiling it everywhere to avoid regressions. +#![allow(unused)] + +#[cfg(test)] +mod tests; + +pub mod eh; + +use core::mem; + +pub struct DwarfReader { + pub ptr: *const u8, +} + +#[repr(C, packed)] +struct Unaligned<T>(T); + +impl DwarfReader { + pub fn new(ptr: *const u8) -> DwarfReader { + DwarfReader { ptr } + } + + // DWARF streams are packed, so e.g., a u32 would not necessarily be aligned + // on a 4-byte boundary. This may cause problems on platforms with strict + // alignment requirements. By wrapping data in a "packed" struct, we are + // telling the backend to generate "misalignment-safe" code. + pub unsafe fn read<T: Copy>(&mut self) -> T { + let Unaligned(result) = *(self.ptr as *const Unaligned<T>); + self.ptr = self.ptr.add(mem::size_of::<T>()); + result + } + + // ULEB128 and SLEB128 encodings are defined in Section 7.6 - "Variable + // Length Data". + pub unsafe fn read_uleb128(&mut self) -> u64 { + let mut shift: usize = 0; + let mut result: u64 = 0; + let mut byte: u8; + loop { + byte = self.read::<u8>(); + result |= ((byte & 0x7F) as u64) << shift; + shift += 7; + if byte & 0x80 == 0 { + break; + } + } + result + } + + pub unsafe fn read_sleb128(&mut self) -> i64 { + let mut shift: u32 = 0; + let mut result: u64 = 0; + let mut byte: u8; + loop { + byte = self.read::<u8>(); + result |= ((byte & 0x7F) as u64) << shift; + shift += 7; + if byte & 0x80 == 0 { + break; + } + } + // sign-extend + if shift < u64::BITS && (byte & 0x40) != 0 { + result |= (!0 as u64) << shift; + } + result as i64 + } +} diff --git a/library/std/src/personality/dwarf/tests.rs b/library/std/src/personality/dwarf/tests.rs new file mode 100644 index 000000000..1644f3708 --- /dev/null +++ b/library/std/src/personality/dwarf/tests.rs @@ -0,0 +1,19 @@ +use super::*; + +#[test] +fn dwarf_reader() { + let encoded: &[u8] = &[1, 2, 3, 4, 5, 6, 7, 0xE5, 0x8E, 0x26, 0x9B, 0xF1, 0x59, 0xFF, 0xFF]; + + let mut reader = DwarfReader::new(encoded.as_ptr()); + + unsafe { + assert!(reader.read::<u8>() == u8::to_be(1u8)); + assert!(reader.read::<u16>() == u16::to_be(0x0203)); + assert!(reader.read::<u32>() == u32::to_be(0x04050607)); + + assert!(reader.read_uleb128() == 624485); + assert!(reader.read_sleb128() == -624485); + + assert!(reader.read::<i8>() == i8::to_be(-1)); + } +} diff --git a/library/std/src/personality/emcc.rs b/library/std/src/personality/emcc.rs new file mode 100644 index 000000000..f942bdf18 --- /dev/null +++ b/library/std/src/personality/emcc.rs @@ -0,0 +1,20 @@ +//! On Emscripten Rust panics are wrapped in C++ exceptions, so we just forward +//! to `__gxx_personality_v0` which is provided by Emscripten. + +use libc::c_int; +use unwind as uw; + +// This is required by the compiler to exist (e.g., it's a lang item), but it's +// never actually called by the compiler. Emscripten EH doesn't use a +// personality function at all, it instead uses __cxa_find_matching_catch. +// Wasm error handling would use __gxx_personality_wasm0. +#[lang = "eh_personality"] +unsafe extern "C" fn rust_eh_personality( + _version: c_int, + _actions: uw::_Unwind_Action, + _exception_class: uw::_Unwind_Exception_Class, + _exception_object: *mut uw::_Unwind_Exception, + _context: *mut uw::_Unwind_Context, +) -> uw::_Unwind_Reason_Code { + core::intrinsics::abort() +} diff --git a/library/std/src/personality/gcc.rs b/library/std/src/personality/gcc.rs new file mode 100644 index 000000000..7f0b0439c --- /dev/null +++ b/library/std/src/personality/gcc.rs @@ -0,0 +1,279 @@ +//! Implementation of panics backed by libgcc/libunwind (in some form). +//! +//! For background on exception handling and stack unwinding please see +//! "Exception Handling in LLVM" (llvm.org/docs/ExceptionHandling.html) and +//! documents linked from it. +//! These are also good reads: +//! * <https://itanium-cxx-abi.github.io/cxx-abi/abi-eh.html> +//! * <https://monoinfinito.wordpress.com/series/exception-handling-in-c/> +//! * <https://www.airs.com/blog/index.php?s=exception+frames> +//! +//! ## A brief summary +//! +//! Exception handling happens in two phases: a search phase and a cleanup +//! phase. +//! +//! In both phases the unwinder walks stack frames from top to bottom using +//! information from the stack frame unwind sections of the current process's +//! modules ("module" here refers to an OS module, i.e., an executable or a +//! dynamic library). +//! +//! For each stack frame, it invokes the associated "personality routine", whose +//! address is also stored in the unwind info section. +//! +//! In the search phase, the job of a personality routine is to examine +//! exception object being thrown, and to decide whether it should be caught at +//! that stack frame. Once the handler frame has been identified, cleanup phase +//! begins. +//! +//! In the cleanup phase, the unwinder invokes each personality routine again. +//! This time it decides which (if any) cleanup code needs to be run for +//! the current stack frame. If so, the control is transferred to a special +//! branch in the function body, the "landing pad", which invokes destructors, +//! frees memory, etc. At the end of the landing pad, control is transferred +//! back to the unwinder and unwinding resumes. +//! +//! Once stack has been unwound down to the handler frame level, unwinding stops +//! and the last personality routine transfers control to the catch block. + +use super::dwarf::eh::{self, EHAction, EHContext}; +use libc::{c_int, uintptr_t}; +use unwind as uw; + +// Register ids were lifted from LLVM's TargetLowering::getExceptionPointerRegister() +// and TargetLowering::getExceptionSelectorRegister() for each architecture, +// then mapped to DWARF register numbers via register definition tables +// (typically <arch>RegisterInfo.td, search for "DwarfRegNum"). +// See also https://llvm.org/docs/WritingAnLLVMBackend.html#defining-a-register. + +#[cfg(target_arch = "x86")] +const UNWIND_DATA_REG: (i32, i32) = (0, 2); // EAX, EDX + +#[cfg(target_arch = "x86_64")] +const UNWIND_DATA_REG: (i32, i32) = (0, 1); // RAX, RDX + +#[cfg(any(target_arch = "arm", target_arch = "aarch64"))] +const UNWIND_DATA_REG: (i32, i32) = (0, 1); // R0, R1 / X0, X1 + +#[cfg(target_arch = "m68k")] +const UNWIND_DATA_REG: (i32, i32) = (0, 1); // D0, D1 + +#[cfg(any(target_arch = "mips", target_arch = "mips64"))] +const UNWIND_DATA_REG: (i32, i32) = (4, 5); // A0, A1 + +#[cfg(any(target_arch = "powerpc", target_arch = "powerpc64"))] +const UNWIND_DATA_REG: (i32, i32) = (3, 4); // R3, R4 / X3, X4 + +#[cfg(target_arch = "s390x")] +const UNWIND_DATA_REG: (i32, i32) = (6, 7); // R6, R7 + +#[cfg(any(target_arch = "sparc", target_arch = "sparc64"))] +const UNWIND_DATA_REG: (i32, i32) = (24, 25); // I0, I1 + +#[cfg(target_arch = "hexagon")] +const UNWIND_DATA_REG: (i32, i32) = (0, 1); // R0, R1 + +#[cfg(any(target_arch = "riscv64", target_arch = "riscv32"))] +const UNWIND_DATA_REG: (i32, i32) = (10, 11); // x10, x11 + +// The following code is based on GCC's C and C++ personality routines. For reference, see: +// https://github.com/gcc-mirror/gcc/blob/master/libstdc++-v3/libsupc++/eh_personality.cc +// https://github.com/gcc-mirror/gcc/blob/trunk/libgcc/unwind-c.c + +cfg_if::cfg_if! { + if #[cfg(all(target_arch = "arm", not(target_os = "ios"), not(target_os = "watchos"), not(target_os = "netbsd")))] { + // ARM EHABI personality routine. + // https://infocenter.arm.com/help/topic/com.arm.doc.ihi0038b/IHI0038B_ehabi.pdf + // + // iOS uses the default routine instead since it uses SjLj unwinding. + #[lang = "eh_personality"] + unsafe extern "C" fn rust_eh_personality( + state: uw::_Unwind_State, + exception_object: *mut uw::_Unwind_Exception, + context: *mut uw::_Unwind_Context, + ) -> uw::_Unwind_Reason_Code { + let state = state as c_int; + let action = state & uw::_US_ACTION_MASK as c_int; + let search_phase = if action == uw::_US_VIRTUAL_UNWIND_FRAME as c_int { + // Backtraces on ARM will call the personality routine with + // state == _US_VIRTUAL_UNWIND_FRAME | _US_FORCE_UNWIND. In those cases + // we want to continue unwinding the stack, otherwise all our backtraces + // would end at __rust_try + if state & uw::_US_FORCE_UNWIND as c_int != 0 { + return continue_unwind(exception_object, context); + } + true + } else if action == uw::_US_UNWIND_FRAME_STARTING as c_int { + false + } else if action == uw::_US_UNWIND_FRAME_RESUME as c_int { + return continue_unwind(exception_object, context); + } else { + return uw::_URC_FAILURE; + }; + + // The DWARF unwinder assumes that _Unwind_Context holds things like the function + // and LSDA pointers, however ARM EHABI places them into the exception object. + // To preserve signatures of functions like _Unwind_GetLanguageSpecificData(), which + // take only the context pointer, GCC personality routines stash a pointer to + // exception_object in the context, using location reserved for ARM's + // "scratch register" (r12). + uw::_Unwind_SetGR(context, uw::UNWIND_POINTER_REG, exception_object as uw::_Unwind_Ptr); + // ...A more principled approach would be to provide the full definition of ARM's + // _Unwind_Context in our libunwind bindings and fetch the required data from there + // directly, bypassing DWARF compatibility functions. + + let eh_action = match find_eh_action(context) { + Ok(action) => action, + Err(_) => return uw::_URC_FAILURE, + }; + if search_phase { + match eh_action { + EHAction::None | EHAction::Cleanup(_) => { + return continue_unwind(exception_object, context); + } + EHAction::Catch(_) => { + // EHABI requires the personality routine to update the + // SP value in the barrier cache of the exception object. + (*exception_object).private[5] = + uw::_Unwind_GetGR(context, uw::UNWIND_SP_REG); + return uw::_URC_HANDLER_FOUND; + } + EHAction::Terminate => return uw::_URC_FAILURE, + } + } else { + match eh_action { + EHAction::None => return continue_unwind(exception_object, context), + EHAction::Cleanup(lpad) | EHAction::Catch(lpad) => { + uw::_Unwind_SetGR( + context, + UNWIND_DATA_REG.0, + exception_object as uintptr_t, + ); + uw::_Unwind_SetGR(context, UNWIND_DATA_REG.1, 0); + uw::_Unwind_SetIP(context, lpad); + return uw::_URC_INSTALL_CONTEXT; + } + EHAction::Terminate => return uw::_URC_FAILURE, + } + } + + // On ARM EHABI the personality routine is responsible for actually + // unwinding a single stack frame before returning (ARM EHABI Sec. 6.1). + unsafe fn continue_unwind( + exception_object: *mut uw::_Unwind_Exception, + context: *mut uw::_Unwind_Context, + ) -> uw::_Unwind_Reason_Code { + if __gnu_unwind_frame(exception_object, context) == uw::_URC_NO_REASON { + uw::_URC_CONTINUE_UNWIND + } else { + uw::_URC_FAILURE + } + } + // defined in libgcc + extern "C" { + fn __gnu_unwind_frame( + exception_object: *mut uw::_Unwind_Exception, + context: *mut uw::_Unwind_Context, + ) -> uw::_Unwind_Reason_Code; + } + } + } else { + // Default personality routine, which is used directly on most targets + // and indirectly on Windows x86_64 via SEH. + unsafe extern "C" fn rust_eh_personality_impl( + version: c_int, + actions: uw::_Unwind_Action, + _exception_class: uw::_Unwind_Exception_Class, + exception_object: *mut uw::_Unwind_Exception, + context: *mut uw::_Unwind_Context, + ) -> uw::_Unwind_Reason_Code { + if version != 1 { + return uw::_URC_FATAL_PHASE1_ERROR; + } + let eh_action = match find_eh_action(context) { + Ok(action) => action, + Err(_) => return uw::_URC_FATAL_PHASE1_ERROR, + }; + if actions as i32 & uw::_UA_SEARCH_PHASE as i32 != 0 { + match eh_action { + EHAction::None | EHAction::Cleanup(_) => uw::_URC_CONTINUE_UNWIND, + EHAction::Catch(_) => uw::_URC_HANDLER_FOUND, + EHAction::Terminate => uw::_URC_FATAL_PHASE1_ERROR, + } + } else { + match eh_action { + EHAction::None => uw::_URC_CONTINUE_UNWIND, + EHAction::Cleanup(lpad) | EHAction::Catch(lpad) => { + uw::_Unwind_SetGR( + context, + UNWIND_DATA_REG.0, + exception_object as uintptr_t, + ); + uw::_Unwind_SetGR(context, UNWIND_DATA_REG.1, 0); + uw::_Unwind_SetIP(context, lpad); + uw::_URC_INSTALL_CONTEXT + } + EHAction::Terminate => uw::_URC_FATAL_PHASE2_ERROR, + } + } + } + + cfg_if::cfg_if! { + if #[cfg(all(windows, target_arch = "x86_64", target_env = "gnu"))] { + // On x86_64 MinGW targets, the unwinding mechanism is SEH however the unwind + // handler data (aka LSDA) uses GCC-compatible encoding. + #[lang = "eh_personality"] + #[allow(nonstandard_style)] + unsafe extern "C" fn rust_eh_personality( + exceptionRecord: *mut uw::EXCEPTION_RECORD, + establisherFrame: uw::LPVOID, + contextRecord: *mut uw::CONTEXT, + dispatcherContext: *mut uw::DISPATCHER_CONTEXT, + ) -> uw::EXCEPTION_DISPOSITION { + uw::_GCC_specific_handler( + exceptionRecord, + establisherFrame, + contextRecord, + dispatcherContext, + rust_eh_personality_impl, + ) + } + } else { + // The personality routine for most of our targets. + #[lang = "eh_personality"] + unsafe extern "C" fn rust_eh_personality( + version: c_int, + actions: uw::_Unwind_Action, + exception_class: uw::_Unwind_Exception_Class, + exception_object: *mut uw::_Unwind_Exception, + context: *mut uw::_Unwind_Context, + ) -> uw::_Unwind_Reason_Code { + rust_eh_personality_impl( + version, + actions, + exception_class, + exception_object, + context, + ) + } + } + } + } +} + +unsafe fn find_eh_action(context: *mut uw::_Unwind_Context) -> Result<EHAction, ()> { + let lsda = uw::_Unwind_GetLanguageSpecificData(context) as *const u8; + let mut ip_before_instr: c_int = 0; + let ip = uw::_Unwind_GetIPInfo(context, &mut ip_before_instr); + let eh_context = EHContext { + // The return address points 1 byte past the call instruction, + // which could be in the next IP range in LSDA range table. + // + // `ip = -1` has special meaning, so use wrapping sub to allow for that + ip: if ip_before_instr != 0 { ip } else { ip.wrapping_sub(1) }, + func_start: uw::_Unwind_GetRegionStart(context), + get_text_start: &|| uw::_Unwind_GetTextRelBase(context), + get_data_start: &|| uw::_Unwind_GetDataRelBase(context), + }; + eh::find_eh_action(lsda, &eh_context) +} diff --git a/library/std/src/primitive_docs.rs b/library/std/src/primitive_docs.rs index b8e546164..242f44ade 100644 --- a/library/std/src/primitive_docs.rs +++ b/library/std/src/primitive_docs.rs @@ -801,11 +801,53 @@ mod prim_array {} /// assert_eq!(2 * pointer_size, std::mem::size_of::<Box<[u8]>>()); /// assert_eq!(2 * pointer_size, std::mem::size_of::<Rc<[u8]>>()); /// ``` +/// +/// ## Trait Implementations +/// +/// Some traits are implemented for slices if the element type implements +/// that trait. This includes [`Eq`], [`Hash`] and [`Ord`]. +/// +/// ## Iteration +/// +/// The slices implement `IntoIterator`. The iterator yields references to the +/// slice elements. +/// +/// ``` +/// let numbers: &[i32] = &[0, 1, 2]; +/// for n in numbers { +/// println!("{n} is a number!"); +/// } +/// ``` +/// +/// The mutable slice yields mutable references to the elements: +/// +/// ``` +/// let mut scores: &mut [i32] = &mut [7, 8, 9]; +/// for score in scores { +/// *score += 1; +/// } +/// ``` +/// +/// This iterator yields mutable references to the slice's elements, so while +/// the element type of the slice is `i32`, the element type of the iterator is +/// `&mut i32`. +/// +/// * [`.iter`] and [`.iter_mut`] are the explicit methods to return the default +/// iterators. +/// * Further methods that return iterators are [`.split`], [`.splitn`], +/// [`.chunks`], [`.windows`] and more. +/// +/// [`Hash`]: core::hash::Hash +/// [`.iter`]: slice::iter +/// [`.iter_mut`]: slice::iter_mut +/// [`.split`]: slice::split +/// [`.splitn`]: slice::splitn +/// [`.chunks`]: slice::chunks +/// [`.windows`]: slice::windows #[stable(feature = "rust1", since = "1.0.0")] mod prim_slice {} #[doc(primitive = "str")] -// /// String slices. /// /// *[See also the `std::str` module](crate::str).* @@ -816,19 +858,22 @@ mod prim_slice {} /// /// String slices are always valid UTF-8. /// -/// # Examples +/// # Basic Usage /// /// String literals are string slices: /// /// ``` -/// let hello = "Hello, world!"; -/// -/// // with an explicit type annotation -/// let hello: &'static str = "Hello, world!"; +/// let hello_world = "Hello, World!"; /// ``` /// -/// They are `'static` because they're stored directly in the final binary, and -/// so will be valid for the `'static` duration. +/// Here we have declared a string slice initialized with a string literal. +/// String literals have a static lifetime, which means the string `hello_world` +/// is guaranteed to be valid for the duration of the entire program. +/// We can explicitly specify `hello_world`'s lifetime as well: +/// +/// ``` +/// let hello_world: &'static str = "Hello, world!"; +/// ``` /// /// # Representation /// @@ -996,7 +1041,7 @@ impl<T> (T,) {} // Fake impl that's only really used for docs. #[cfg(doc)] #[stable(feature = "rust1", since = "1.0.0")] -#[cfg_attr(not(bootstrap), doc(fake_variadic))] +#[doc(fake_variadic)] /// This trait is implemented on arbitrary-length tuples. impl<T: Clone> Clone for (T,) { fn clone(&self) -> Self { @@ -1007,7 +1052,7 @@ impl<T: Clone> Clone for (T,) { // Fake impl that's only really used for docs. #[cfg(doc)] #[stable(feature = "rust1", since = "1.0.0")] -#[cfg_attr(not(bootstrap), doc(fake_variadic))] +#[doc(fake_variadic)] /// This trait is implemented on arbitrary-length tuples. impl<T: Copy> Copy for (T,) { // empty @@ -1178,7 +1223,7 @@ mod prim_usize {} #[doc(alias = "&")] #[doc(alias = "&mut")] // -/// References, both shared and mutable. +/// References, `&T` and `&mut T`. /// /// A reference represents a borrow of some owned value. You can get one by using the `&` or `&mut` /// operators on a value, or by using a [`ref`](../std/keyword.ref.html) or @@ -1484,13 +1529,12 @@ mod prim_fn {} // Required to make auto trait impls render. // See src/librustdoc/passes/collect_trait_impls.rs:collect_trait_impls #[doc(hidden)] -#[cfg(not(bootstrap))] impl<Ret, T> fn(T) -> Ret {} // Fake impl that's only really used for docs. #[cfg(doc)] #[stable(feature = "rust1", since = "1.0.0")] -#[cfg_attr(not(bootstrap), doc(fake_variadic))] +#[doc(fake_variadic)] /// This trait is implemented on function pointers with any number of arguments. impl<Ret, T> Clone for fn(T) -> Ret { fn clone(&self) -> Self { @@ -1501,7 +1545,7 @@ impl<Ret, T> Clone for fn(T) -> Ret { // Fake impl that's only really used for docs. #[cfg(doc)] #[stable(feature = "rust1", since = "1.0.0")] -#[cfg_attr(not(bootstrap), doc(fake_variadic))] +#[doc(fake_variadic)] /// This trait is implemented on function pointers with any number of arguments. impl<Ret, T> Copy for fn(T) -> Ret { // empty diff --git a/library/std/src/process.rs b/library/std/src/process.rs index d6cba7e75..d91d4fa64 100644 --- a/library/std/src/process.rs +++ b/library/std/src/process.rs @@ -169,15 +169,15 @@ use crate::sys_common::{AsInner, AsInnerMut, FromInner, IntoInner}; pub struct Child { pub(crate) handle: imp::Process, - /// The handle for writing to the child's standard input (stdin), if it has - /// been captured. To avoid partially moving - /// the `child` and thus blocking yourself from calling - /// functions on `child` while using `stdin`, - /// you might find it helpful: + /// The handle for writing to the child's standard input (stdin), if it + /// has been captured. You might find it helpful to do /// /// ```compile_fail,E0425 /// let stdin = child.stdin.take().unwrap(); /// ``` + /// + /// to avoid partially moving the `child` and thus blocking yourself from calling + /// functions on `child` while using `stdin`. #[stable(feature = "process", since = "1.0.0")] pub stdin: Option<ChildStdin>, diff --git a/library/std/src/rt.rs b/library/std/src/rt.rs index 663537a05..98f6cc7aa 100644 --- a/library/std/src/rt.rs +++ b/library/std/src/rt.rs @@ -72,10 +72,29 @@ macro_rules! rtunwrap { // Runs before `main`. // SAFETY: must be called only once during runtime initialization. // NOTE: this is not guaranteed to run, for example when Rust code is called externally. +// +// # The `sigpipe` parameter +// +// Since 2014, the Rust runtime on Unix has set the `SIGPIPE` handler to +// `SIG_IGN`. Applications have good reasons to want a different behavior +// though, so there is a `#[unix_sigpipe = "..."]` attribute on `fn main()` that +// can be used to select how `SIGPIPE` shall be setup (if changed at all) before +// `fn main()` is called. See <https://github.com/rust-lang/rust/issues/97889> +// for more info. +// +// The `sigpipe` parameter to this function gets its value via the code that +// rustc generates to invoke `fn lang_start()`. The reason we have `sigpipe` for +// all platforms and not only Unix, is because std is not allowed to have `cfg` +// directives as this high level. See the module docs in +// `src/tools/tidy/src/pal.rs` for more info. On all other platforms, `sigpipe` +// has a value, but its value is ignored. +// +// Even though it is an `u8`, it only ever has 3 values. These are documented in +// `compiler/rustc_session/src/config/sigpipe.rs`. #[cfg_attr(test, allow(dead_code))] -unsafe fn init(argc: isize, argv: *const *const u8) { +unsafe fn init(argc: isize, argv: *const *const u8, sigpipe: u8) { unsafe { - sys::init(argc, argv); + sys::init(argc, argv, sigpipe); let main_guard = sys::thread::guard::init(); // Next, set up the current Thread with the guard information we just @@ -107,6 +126,7 @@ fn lang_start_internal( main: &(dyn Fn() -> i32 + Sync + crate::panic::RefUnwindSafe), argc: isize, argv: *const *const u8, + sigpipe: u8, ) -> Result<isize, !> { use crate::{mem, panic}; let rt_abort = move |e| { @@ -124,7 +144,7 @@ fn lang_start_internal( // prevent libstd from accidentally introducing a panic to these functions. Another is from // user code from `main` or, more nefariously, as described in e.g. issue #86030. // SAFETY: Only called once during runtime initialization. - panic::catch_unwind(move || unsafe { init(argc, argv) }).map_err(rt_abort)?; + panic::catch_unwind(move || unsafe { init(argc, argv, sigpipe) }).map_err(rt_abort)?; let ret_code = panic::catch_unwind(move || panic::catch_unwind(main).unwrap_or(101) as isize) .map_err(move |e| { mem::forget(e); @@ -140,11 +160,16 @@ fn lang_start<T: crate::process::Termination + 'static>( main: fn() -> T, argc: isize, argv: *const *const u8, + #[cfg(not(bootstrap))] sigpipe: u8, ) -> isize { let Ok(v) = lang_start_internal( &move || crate::sys_common::backtrace::__rust_begin_short_backtrace(main).report().to_i32(), argc, argv, + #[cfg(bootstrap)] + 2, // Temporary inlining of sigpipe::DEFAULT until bootstrap stops being special + #[cfg(not(bootstrap))] + sigpipe, ); v } diff --git a/library/std/src/sync/mpsc/mpsc_queue/tests.rs b/library/std/src/sync/mpsc/mpsc_queue/tests.rs index 9f4f31ed0..34b2a9a98 100644 --- a/library/std/src/sync/mpsc/mpsc_queue/tests.rs +++ b/library/std/src/sync/mpsc/mpsc_queue/tests.rs @@ -13,7 +13,7 @@ fn test_full() { #[test] fn test() { let nthreads = 8; - let nmsgs = 1000; + let nmsgs = if cfg!(miri) { 100 } else { 1000 }; let q = Queue::new(); match q.pop() { Empty => {} diff --git a/library/std/src/sync/mpsc/spsc_queue/tests.rs b/library/std/src/sync/mpsc/spsc_queue/tests.rs index 467ef3dbd..eb6d5c2cf 100644 --- a/library/std/src/sync/mpsc/spsc_queue/tests.rs +++ b/library/std/src/sync/mpsc/spsc_queue/tests.rs @@ -77,12 +77,13 @@ fn stress() { } unsafe fn stress_bound(bound: usize) { + let count = if cfg!(miri) { 1000 } else { 100000 }; let q = Arc::new(Queue::with_additions(bound, (), ())); let (tx, rx) = channel(); let q2 = q.clone(); let _t = thread::spawn(move || { - for _ in 0..100000 { + for _ in 0..count { loop { match q2.pop() { Some(1) => break, @@ -93,7 +94,7 @@ fn stress() { } tx.send(()).unwrap(); }); - for _ in 0..100000 { + for _ in 0..count { q.push(1); } rx.recv().unwrap(); diff --git a/library/std/src/sync/mpsc/sync_tests.rs b/library/std/src/sync/mpsc/sync_tests.rs index e58649bab..63c794369 100644 --- a/library/std/src/sync/mpsc/sync_tests.rs +++ b/library/std/src/sync/mpsc/sync_tests.rs @@ -113,23 +113,25 @@ fn chan_gone_concurrent() { #[test] fn stress() { + let count = if cfg!(miri) { 100 } else { 10000 }; let (tx, rx) = sync_channel::<i32>(0); thread::spawn(move || { - for _ in 0..10000 { + for _ in 0..count { tx.send(1).unwrap(); } }); - for _ in 0..10000 { + for _ in 0..count { assert_eq!(rx.recv().unwrap(), 1); } } #[test] fn stress_recv_timeout_two_threads() { + let count = if cfg!(miri) { 100 } else { 10000 }; let (tx, rx) = sync_channel::<i32>(0); thread::spawn(move || { - for _ in 0..10000 { + for _ in 0..count { tx.send(1).unwrap(); } }); @@ -146,12 +148,12 @@ fn stress_recv_timeout_two_threads() { } } - assert_eq!(recv_count, 10000); + assert_eq!(recv_count, count); } #[test] fn stress_recv_timeout_shared() { - const AMT: u32 = 1000; + const AMT: u32 = if cfg!(miri) { 100 } else { 1000 }; const NTHREADS: u32 = 8; let (tx, rx) = sync_channel::<i32>(0); let (dtx, drx) = sync_channel::<()>(0); @@ -191,7 +193,7 @@ fn stress_recv_timeout_shared() { #[test] fn stress_shared() { - const AMT: u32 = 1000; + const AMT: u32 = if cfg!(miri) { 100 } else { 1000 }; const NTHREADS: u32 = 8; let (tx, rx) = sync_channel::<i32>(0); let (dtx, drx) = sync_channel::<()>(0); @@ -438,12 +440,13 @@ fn stream_send_recv_stress() { #[test] fn recv_a_lot() { + let count = if cfg!(miri) { 1000 } else { 10000 }; // Regression test that we don't run out of stack in scheduler context - let (tx, rx) = sync_channel(10000); - for _ in 0..10000 { + let (tx, rx) = sync_channel(count); + for _ in 0..count { tx.send(()).unwrap(); } - for _ in 0..10000 { + for _ in 0..count { rx.recv().unwrap(); } } diff --git a/library/std/src/sync/mpsc/tests.rs b/library/std/src/sync/mpsc/tests.rs index 4deb3e596..f6d0796f6 100644 --- a/library/std/src/sync/mpsc/tests.rs +++ b/library/std/src/sync/mpsc/tests.rs @@ -120,13 +120,14 @@ fn chan_gone_concurrent() { #[test] fn stress() { + let count = if cfg!(miri) { 100 } else { 10000 }; let (tx, rx) = channel::<i32>(); let t = thread::spawn(move || { - for _ in 0..10000 { + for _ in 0..count { tx.send(1).unwrap(); } }); - for _ in 0..10000 { + for _ in 0..count { assert_eq!(rx.recv().unwrap(), 1); } t.join().ok().expect("thread panicked"); @@ -134,7 +135,7 @@ fn stress() { #[test] fn stress_shared() { - const AMT: u32 = 10000; + const AMT: u32 = if cfg!(miri) { 100 } else { 10000 }; const NTHREADS: u32 = 8; let (tx, rx) = channel::<i32>(); @@ -504,12 +505,13 @@ fn very_long_recv_timeout_wont_panic() { #[test] fn recv_a_lot() { + let count = if cfg!(miri) { 1000 } else { 10000 }; // Regression test that we don't run out of stack in scheduler context let (tx, rx) = channel(); - for _ in 0..10000 { + for _ in 0..count { tx.send(()).unwrap(); } - for _ in 0..10000 { + for _ in 0..count { rx.recv().unwrap(); } } diff --git a/library/std/src/sync/mutex.rs b/library/std/src/sync/mutex.rs index e0d13cd64..de851c8fb 100644 --- a/library/std/src/sync/mutex.rs +++ b/library/std/src/sync/mutex.rs @@ -192,6 +192,7 @@ unsafe impl<T: ?Sized + Send> Sync for Mutex<T> {} and cause Futures to not implement `Send`"] #[stable(feature = "rust1", since = "1.0.0")] #[clippy::has_significant_drop] +#[cfg_attr(not(test), rustc_diagnostic_item = "MutexGuard")] pub struct MutexGuard<'a, T: ?Sized + 'a> { lock: &'a Mutex<T>, poison: poison::Guard, diff --git a/library/std/src/sync/once_lock.rs b/library/std/src/sync/once_lock.rs index 813516040..37413ec62 100644 --- a/library/std/src/sync/once_lock.rs +++ b/library/std/src/sync/once_lock.rs @@ -3,7 +3,6 @@ use crate::fmt; use crate::marker::PhantomData; use crate::mem::MaybeUninit; use crate::panic::{RefUnwindSafe, UnwindSafe}; -use crate::pin::Pin; use crate::sync::Once; /// A synchronization primitive which can be written to only once. @@ -223,60 +222,6 @@ impl<T> OnceLock<T> { Ok(unsafe { self.get_unchecked() }) } - /// Internal-only API that gets the contents of the cell, initializing it - /// in two steps with `f` and `g` if the cell was empty. - /// - /// `f` is called to construct the value, which is then moved into the cell - /// and given as a (pinned) mutable reference to `g` to finish - /// initialization. - /// - /// This allows `g` to inspect an manipulate the value after it has been - /// moved into its final place in the cell, but before the cell is - /// considered initialized. - /// - /// # Panics - /// - /// If `f` or `g` panics, the panic is propagated to the caller, and the - /// cell remains uninitialized. - /// - /// With the current implementation, if `g` panics, the value from `f` will - /// not be dropped. This should probably be fixed if this is ever used for - /// a type where this matters. - /// - /// It is an error to reentrantly initialize the cell from `f`. The exact - /// outcome is unspecified. Current implementation deadlocks, but this may - /// be changed to a panic in the future. - pub(crate) fn get_or_init_pin<F, G>(self: Pin<&Self>, f: F, g: G) -> Pin<&T> - where - F: FnOnce() -> T, - G: FnOnce(Pin<&mut T>), - { - if let Some(value) = self.get_ref().get() { - // SAFETY: The inner value was already initialized, and will not be - // moved anymore. - return unsafe { Pin::new_unchecked(value) }; - } - - let slot = &self.value; - - // Ignore poisoning from other threads - // If another thread panics, then we'll be able to run our closure - self.once.call_once_force(|_| { - let value = f(); - // SAFETY: We use the Once (self.once) to guarantee unique access - // to the UnsafeCell (slot). - let value: &mut T = unsafe { (&mut *slot.get()).write(value) }; - // SAFETY: The value has been written to its final place in - // self.value. We do not to move it anymore, which we promise here - // with a Pin<&mut T>. - g(unsafe { Pin::new_unchecked(value) }); - }); - - // SAFETY: The inner value has been initialized, and will not be moved - // anymore. - unsafe { Pin::new_unchecked(self.get_ref().get_unchecked()) } - } - /// Consumes the `OnceLock`, returning the wrapped value. Returns /// `None` if the cell was empty. /// diff --git a/library/std/src/sync/rwlock.rs b/library/std/src/sync/rwlock.rs index 6e4a2cfc8..9ab781561 100644 --- a/library/std/src/sync/rwlock.rs +++ b/library/std/src/sync/rwlock.rs @@ -101,6 +101,7 @@ unsafe impl<T: ?Sized + Send + Sync> Sync for RwLock<T> {} and cause Futures to not implement `Send`"] #[stable(feature = "rust1", since = "1.0.0")] #[clippy::has_significant_drop] +#[cfg_attr(not(test), rustc_diagnostic_item = "RwLockReadGuard")] pub struct RwLockReadGuard<'a, T: ?Sized + 'a> { // NB: we use a pointer instead of `&'a T` to avoid `noalias` violations, because a // `Ref` argument doesn't hold immutability for its whole scope, only until it drops. @@ -130,6 +131,7 @@ unsafe impl<T: ?Sized + Sync> Sync for RwLockReadGuard<'_, T> {} and cause Future's to not implement `Send`"] #[stable(feature = "rust1", since = "1.0.0")] #[clippy::has_significant_drop] +#[cfg_attr(not(test), rustc_diagnostic_item = "RwLockWriteGuard")] pub struct RwLockWriteGuard<'a, T: ?Sized + 'a> { lock: &'a RwLock<T>, poison: poison::Guard, diff --git a/library/std/src/sync/rwlock/tests.rs b/library/std/src/sync/rwlock/tests.rs index 08255c985..b5b3ad989 100644 --- a/library/std/src/sync/rwlock/tests.rs +++ b/library/std/src/sync/rwlock/tests.rs @@ -19,7 +19,7 @@ fn smoke() { #[test] fn frob() { const N: u32 = 10; - const M: usize = 1000; + const M: usize = if cfg!(miri) { 100 } else { 1000 }; let r = Arc::new(RwLock::new(())); diff --git a/library/std/src/sys/hermit/condvar.rs b/library/std/src/sys/hermit/condvar.rs deleted file mode 100644 index 22059ca0d..000000000 --- a/library/std/src/sys/hermit/condvar.rs +++ /dev/null @@ -1,90 +0,0 @@ -use crate::ffi::c_void; -use crate::ptr; -use crate::sync::atomic::{AtomicUsize, Ordering::SeqCst}; -use crate::sys::hermit::abi; -use crate::sys::locks::Mutex; -use crate::sys_common::lazy_box::{LazyBox, LazyInit}; -use crate::time::Duration; - -// The implementation is inspired by Andrew D. Birrell's paper -// "Implementing Condition Variables with Semaphores" - -pub struct Condvar { - counter: AtomicUsize, - sem1: *const c_void, - sem2: *const c_void, -} - -pub(crate) type MovableCondvar = LazyBox<Condvar>; - -impl LazyInit for Condvar { - fn init() -> Box<Self> { - Box::new(Self::new()) - } -} - -unsafe impl Send for Condvar {} -unsafe impl Sync for Condvar {} - -impl Condvar { - pub fn new() -> Self { - let mut condvar = - Self { counter: AtomicUsize::new(0), sem1: ptr::null(), sem2: ptr::null() }; - unsafe { - let _ = abi::sem_init(&mut condvar.sem1, 0); - let _ = abi::sem_init(&mut condvar.sem2, 0); - } - condvar - } - - pub unsafe fn notify_one(&self) { - if self.counter.load(SeqCst) > 0 { - self.counter.fetch_sub(1, SeqCst); - abi::sem_post(self.sem1); - abi::sem_timedwait(self.sem2, 0); - } - } - - pub unsafe fn notify_all(&self) { - let counter = self.counter.swap(0, SeqCst); - for _ in 0..counter { - abi::sem_post(self.sem1); - } - for _ in 0..counter { - abi::sem_timedwait(self.sem2, 0); - } - } - - pub unsafe fn wait(&self, mutex: &Mutex) { - self.counter.fetch_add(1, SeqCst); - mutex.unlock(); - abi::sem_timedwait(self.sem1, 0); - abi::sem_post(self.sem2); - mutex.lock(); - } - - pub unsafe fn wait_timeout(&self, mutex: &Mutex, dur: Duration) -> bool { - self.counter.fetch_add(1, SeqCst); - mutex.unlock(); - let millis = dur.as_millis().min(u32::MAX as u128) as u32; - - let res = if millis > 0 { - abi::sem_timedwait(self.sem1, millis) - } else { - abi::sem_trywait(self.sem1) - }; - - abi::sem_post(self.sem2); - mutex.lock(); - res == 0 - } -} - -impl Drop for Condvar { - fn drop(&mut self) { - unsafe { - let _ = abi::sem_destroy(self.sem1); - let _ = abi::sem_destroy(self.sem2); - } - } -} diff --git a/library/std/src/sys/hermit/fs.rs b/library/std/src/sys/hermit/fs.rs index fa9a7fb19..f921839cf 100644 --- a/library/std/src/sys/hermit/fs.rs +++ b/library/std/src/sys/hermit/fs.rs @@ -2,7 +2,7 @@ use crate::ffi::{CStr, CString, OsString}; use crate::fmt; use crate::hash::{Hash, Hasher}; use crate::io::{self, Error, ErrorKind}; -use crate::io::{IoSlice, IoSliceMut, ReadBuf, SeekFrom}; +use crate::io::{BorrowedCursor, IoSlice, IoSliceMut, SeekFrom}; use crate::os::unix::ffi::OsStrExt; use crate::path::{Path, PathBuf}; use crate::sys::cvt; @@ -41,6 +41,9 @@ pub struct OpenOptions { mode: i32, } +#[derive(Copy, Clone, Debug, Default)] +pub struct FileTimes {} + pub struct FilePermissions(!); pub struct FileType(!); @@ -110,6 +113,11 @@ impl fmt::Debug for FilePermissions { } } +impl FileTimes { + pub fn set_accessed(&mut self, _t: SystemTime) {} + pub fn set_modified(&mut self, _t: SystemTime) {} +} + impl FileType { pub fn is_dir(&self) -> bool { self.0 @@ -312,8 +320,8 @@ impl File { false } - pub fn read_buf(&self, buf: &mut ReadBuf<'_>) -> io::Result<()> { - crate::io::default_read_buf(|buf| self.read(buf), buf) + pub fn read_buf(&self, cursor: BorrowedCursor<'_>) -> io::Result<()> { + crate::io::default_read_buf(|buf| self.read(buf), cursor) } pub fn write(&self, buf: &[u8]) -> io::Result<usize> { @@ -344,6 +352,10 @@ impl File { pub fn set_permissions(&self, _perm: FilePermissions) -> io::Result<()> { Err(Error::from_raw_os_error(22)) } + + pub fn set_times(&self, _times: FileTimes) -> io::Result<()> { + Err(Error::from_raw_os_error(22)) + } } impl DirBuilder { diff --git a/library/std/src/sys/hermit/futex.rs b/library/std/src/sys/hermit/futex.rs new file mode 100644 index 000000000..b64c174b0 --- /dev/null +++ b/library/std/src/sys/hermit/futex.rs @@ -0,0 +1,39 @@ +use super::abi; +use crate::ptr::null; +use crate::sync::atomic::AtomicU32; +use crate::time::Duration; + +pub fn futex_wait(futex: &AtomicU32, expected: u32, timeout: Option<Duration>) -> bool { + // Calculate the timeout as a relative timespec. + // + // Overflows are rounded up to an infinite timeout (None). + let timespec = timeout.and_then(|dur| { + Some(abi::timespec { + tv_sec: dur.as_secs().try_into().ok()?, + tv_nsec: dur.subsec_nanos().into(), + }) + }); + + let r = unsafe { + abi::futex_wait( + futex.as_mut_ptr(), + expected, + timespec.as_ref().map_or(null(), |t| t as *const abi::timespec), + abi::FUTEX_RELATIVE_TIMEOUT, + ) + }; + + r != -abi::errno::ETIMEDOUT +} + +#[inline] +pub fn futex_wake(futex: &AtomicU32) -> bool { + unsafe { abi::futex_wake(futex.as_mut_ptr(), 1) > 0 } +} + +#[inline] +pub fn futex_wake_all(futex: &AtomicU32) { + unsafe { + abi::futex_wake(futex.as_mut_ptr(), i32::MAX); + } +} diff --git a/library/std/src/sys/hermit/mod.rs b/library/std/src/sys/hermit/mod.rs index 60b7a973c..827d82900 100644 --- a/library/std/src/sys/hermit/mod.rs +++ b/library/std/src/sys/hermit/mod.rs @@ -25,6 +25,7 @@ pub mod cmath; pub mod env; pub mod fd; pub mod fs; +pub mod futex; #[path = "../unsupported/io.rs"] pub mod io; pub mod memchr; @@ -45,14 +46,14 @@ pub mod thread_local_dtor; pub mod thread_local_key; pub mod time; -mod condvar; -mod mutex; -mod rwlock; - +#[path = "../unix/locks"] pub mod locks { - pub use super::condvar::*; - pub use super::mutex::*; - pub use super::rwlock::*; + mod futex_condvar; + mod futex_mutex; + mod futex_rwlock; + pub(crate) use futex_condvar::MovableCondvar; + pub(crate) use futex_mutex::{MovableMutex, Mutex}; + pub(crate) use futex_rwlock::{MovableRwLock, RwLock}; } use crate::io::ErrorKind; @@ -98,7 +99,7 @@ pub extern "C" fn __rust_abort() { // SAFETY: must be called only once during runtime initialization. // NOTE: this is not guaranteed to run, for example when Rust code is called externally. -pub unsafe fn init(argc: isize, argv: *const *const u8) { +pub unsafe fn init(argc: isize, argv: *const *const u8, _sigpipe: u8) { let _ = net::init(); args::init(argc, argv); } diff --git a/library/std/src/sys/hermit/mutex.rs b/library/std/src/sys/hermit/mutex.rs deleted file mode 100644 index eb15a04ff..000000000 --- a/library/std/src/sys/hermit/mutex.rs +++ /dev/null @@ -1,216 +0,0 @@ -use crate::cell::UnsafeCell; -use crate::collections::VecDeque; -use crate::hint; -use crate::ops::{Deref, DerefMut, Drop}; -use crate::ptr; -use crate::sync::atomic::{AtomicUsize, Ordering}; -use crate::sys::hermit::abi; - -/// This type provides a lock based on busy waiting to realize mutual exclusion -/// -/// # Description -/// -/// This structure behaves a lot like a common mutex. There are some differences: -/// -/// - By using busy waiting, it can be used outside the runtime. -/// - It is a so called ticket lock and is completely fair. -#[cfg_attr(target_arch = "x86_64", repr(align(128)))] -#[cfg_attr(not(target_arch = "x86_64"), repr(align(64)))] -struct Spinlock<T: ?Sized> { - queue: AtomicUsize, - dequeue: AtomicUsize, - data: UnsafeCell<T>, -} - -unsafe impl<T: ?Sized + Send> Sync for Spinlock<T> {} -unsafe impl<T: ?Sized + Send> Send for Spinlock<T> {} - -/// A guard to which the protected data can be accessed -/// -/// When the guard falls out of scope it will release the lock. -struct SpinlockGuard<'a, T: ?Sized + 'a> { - dequeue: &'a AtomicUsize, - data: &'a mut T, -} - -impl<T> Spinlock<T> { - pub const fn new(user_data: T) -> Spinlock<T> { - Spinlock { - queue: AtomicUsize::new(0), - dequeue: AtomicUsize::new(1), - data: UnsafeCell::new(user_data), - } - } - - #[inline] - fn obtain_lock(&self) { - let ticket = self.queue.fetch_add(1, Ordering::SeqCst) + 1; - let mut counter: u16 = 0; - while self.dequeue.load(Ordering::SeqCst) != ticket { - counter += 1; - if counter < 100 { - hint::spin_loop(); - } else { - counter = 0; - unsafe { - abi::yield_now(); - } - } - } - } - - #[inline] - pub unsafe fn lock(&self) -> SpinlockGuard<'_, T> { - self.obtain_lock(); - SpinlockGuard { dequeue: &self.dequeue, data: &mut *self.data.get() } - } -} - -impl<T: ?Sized + Default> Default for Spinlock<T> { - fn default() -> Spinlock<T> { - Spinlock::new(Default::default()) - } -} - -impl<'a, T: ?Sized> Deref for SpinlockGuard<'a, T> { - type Target = T; - fn deref(&self) -> &T { - &*self.data - } -} - -impl<'a, T: ?Sized> DerefMut for SpinlockGuard<'a, T> { - fn deref_mut(&mut self) -> &mut T { - &mut *self.data - } -} - -impl<'a, T: ?Sized> Drop for SpinlockGuard<'a, T> { - /// The dropping of the SpinlockGuard will release the lock it was created from. - fn drop(&mut self) { - self.dequeue.fetch_add(1, Ordering::SeqCst); - } -} - -/// Realize a priority queue for tasks -struct PriorityQueue { - queues: [Option<VecDeque<abi::Tid>>; abi::NO_PRIORITIES], - prio_bitmap: u64, -} - -impl PriorityQueue { - pub const fn new() -> PriorityQueue { - PriorityQueue { - queues: [ - None, None, None, None, None, None, None, None, None, None, None, None, None, None, - None, None, None, None, None, None, None, None, None, None, None, None, None, None, - None, None, None, - ], - prio_bitmap: 0, - } - } - - /// Add a task id by its priority to the queue - pub fn push(&mut self, prio: abi::Priority, id: abi::Tid) { - let i: usize = prio.into().into(); - self.prio_bitmap |= (1 << i) as u64; - if let Some(queue) = &mut self.queues[i] { - queue.push_back(id); - } else { - let mut queue = VecDeque::new(); - queue.push_back(id); - self.queues[i] = Some(queue); - } - } - - fn pop_from_queue(&mut self, queue_index: usize) -> Option<abi::Tid> { - if let Some(queue) = &mut self.queues[queue_index] { - let id = queue.pop_front(); - - if queue.is_empty() { - self.prio_bitmap &= !(1 << queue_index as u64); - } - - id - } else { - None - } - } - - /// Pop the task handle with the highest priority from the queue - pub fn pop(&mut self) -> Option<abi::Tid> { - for i in 0..abi::NO_PRIORITIES { - if self.prio_bitmap & (1 << i) != 0 { - return self.pop_from_queue(i); - } - } - - None - } -} - -struct MutexInner { - locked: bool, - blocked_task: PriorityQueue, -} - -impl MutexInner { - pub const fn new() -> MutexInner { - MutexInner { locked: false, blocked_task: PriorityQueue::new() } - } -} - -pub struct Mutex { - inner: Spinlock<MutexInner>, -} - -pub type MovableMutex = Mutex; - -unsafe impl Send for Mutex {} -unsafe impl Sync for Mutex {} - -impl Mutex { - pub const fn new() -> Mutex { - Mutex { inner: Spinlock::new(MutexInner::new()) } - } - - #[inline] - pub unsafe fn init(&mut self) {} - - #[inline] - pub unsafe fn lock(&self) { - loop { - let mut guard = self.inner.lock(); - if guard.locked == false { - guard.locked = true; - return; - } else { - let prio = abi::get_priority(); - let id = abi::getpid(); - - guard.blocked_task.push(prio, id); - abi::block_current_task(); - drop(guard); - abi::yield_now(); - } - } - } - - #[inline] - pub unsafe fn unlock(&self) { - let mut guard = self.inner.lock(); - guard.locked = false; - if let Some(tid) = guard.blocked_task.pop() { - abi::wakeup_task(tid); - } - } - - #[inline] - pub unsafe fn try_lock(&self) -> bool { - let mut guard = self.inner.lock(); - if guard.locked == false { - guard.locked = true; - } - guard.locked - } -} diff --git a/library/std/src/sys/hermit/net.rs b/library/std/src/sys/hermit/net.rs index 745476171..8a13879d8 100644 --- a/library/std/src/sys/hermit/net.rs +++ b/library/std/src/sys/hermit/net.rs @@ -487,6 +487,4 @@ pub mod netc { #[derive(Copy, Clone)] pub struct sockaddr {} - - pub type socklen_t = usize; } diff --git a/library/std/src/sys/hermit/rwlock.rs b/library/std/src/sys/hermit/rwlock.rs deleted file mode 100644 index 9701bab1f..000000000 --- a/library/std/src/sys/hermit/rwlock.rs +++ /dev/null @@ -1,144 +0,0 @@ -use crate::cell::UnsafeCell; -use crate::sys::locks::{MovableCondvar, Mutex}; -use crate::sys_common::lazy_box::{LazyBox, LazyInit}; - -pub struct RwLock { - lock: Mutex, - cond: MovableCondvar, - state: UnsafeCell<State>, -} - -pub type MovableRwLock = RwLock; - -enum State { - Unlocked, - Reading(usize), - Writing, -} - -unsafe impl Send for RwLock {} -unsafe impl Sync for RwLock {} - -// This rwlock implementation is a relatively simple implementation which has a -// condition variable for readers/writers as well as a mutex protecting the -// internal state of the lock. A current downside of the implementation is that -// unlocking the lock will notify *all* waiters rather than just readers or just -// writers. This can cause lots of "thundering stampede" problems. While -// hopefully correct this implementation is very likely to want to be changed in -// the future. - -impl RwLock { - pub const fn new() -> RwLock { - RwLock { - lock: Mutex::new(), - cond: MovableCondvar::new(), - state: UnsafeCell::new(State::Unlocked), - } - } - - #[inline] - pub unsafe fn read(&self) { - self.lock.lock(); - while !(*self.state.get()).inc_readers() { - self.cond.wait(&self.lock); - } - self.lock.unlock(); - } - - #[inline] - pub unsafe fn try_read(&self) -> bool { - self.lock.lock(); - let ok = (*self.state.get()).inc_readers(); - self.lock.unlock(); - return ok; - } - - #[inline] - pub unsafe fn write(&self) { - self.lock.lock(); - while !(*self.state.get()).inc_writers() { - self.cond.wait(&self.lock); - } - self.lock.unlock(); - } - - #[inline] - pub unsafe fn try_write(&self) -> bool { - self.lock.lock(); - let ok = (*self.state.get()).inc_writers(); - self.lock.unlock(); - return ok; - } - - #[inline] - pub unsafe fn read_unlock(&self) { - self.lock.lock(); - let notify = (*self.state.get()).dec_readers(); - self.lock.unlock(); - if notify { - // FIXME: should only wake up one of these some of the time - self.cond.notify_all(); - } - } - - #[inline] - pub unsafe fn write_unlock(&self) { - self.lock.lock(); - (*self.state.get()).dec_writers(); - self.lock.unlock(); - // FIXME: should only wake up one of these some of the time - self.cond.notify_all(); - } -} - -impl State { - fn inc_readers(&mut self) -> bool { - match *self { - State::Unlocked => { - *self = State::Reading(1); - true - } - State::Reading(ref mut cnt) => { - *cnt += 1; - true - } - State::Writing => false, - } - } - - fn inc_writers(&mut self) -> bool { - match *self { - State::Unlocked => { - *self = State::Writing; - true - } - State::Reading(_) | State::Writing => false, - } - } - - fn dec_readers(&mut self) -> bool { - let zero = match *self { - State::Reading(ref mut cnt) => { - *cnt -= 1; - *cnt == 0 - } - State::Unlocked | State::Writing => invalid(), - }; - if zero { - *self = State::Unlocked; - } - zero - } - - fn dec_writers(&mut self) { - match *self { - State::Writing => {} - State::Unlocked | State::Reading(_) => invalid(), - } - *self = State::Unlocked; - } -} - -fn invalid() -> ! { - panic!("inconsistent rwlock"); -} diff --git a/library/std/src/sys/itron/mutex.rs b/library/std/src/sys/itron/mutex.rs index 715e94c3b..085662e6d 100644 --- a/library/std/src/sys/itron/mutex.rs +++ b/library/std/src/sys/itron/mutex.rs @@ -31,12 +31,6 @@ impl Mutex { Mutex { mtx: SpinIdOnceCell::new() } } - pub unsafe fn init(&mut self) { - // Initialize `self.mtx` eagerly - let id = new_mtx().unwrap_or_else(|e| fail(e, &"acre_mtx")); - unsafe { self.mtx.set_unchecked((id, ())) }; - } - /// Get the inner mutex's ID, which is lazily created. fn raw(&self) -> abi::ID { match self.mtx.get_or_try_init(|| new_mtx().map(|id| (id, ()))) { diff --git a/library/std/src/sys/sgx/abi/thread.rs b/library/std/src/sys/sgx/abi/thread.rs index ef55b821a..2b23e368c 100644 --- a/library/std/src/sys/sgx/abi/thread.rs +++ b/library/std/src/sys/sgx/abi/thread.rs @@ -7,7 +7,11 @@ use fortanix_sgx_abi::Tcs; #[unstable(feature = "sgx_platform", issue = "56975")] pub fn current() -> Tcs { extern "C" { - fn get_tcs_addr() -> Tcs; + fn get_tcs_addr() -> *mut u8; + } + let addr = unsafe { get_tcs_addr() }; + match Tcs::new(addr) { + Some(tcs) => tcs, + None => rtabort!("TCS must not be placed at address zero (this is a linker error)"), } - unsafe { get_tcs_addr() } } diff --git a/library/std/src/sys/sgx/abi/usercalls/alloc.rs b/library/std/src/sys/sgx/abi/usercalls/alloc.rs index ea24fedd0..5409bd177 100644 --- a/library/std/src/sys/sgx/abi/usercalls/alloc.rs +++ b/library/std/src/sys/sgx/abi/usercalls/alloc.rs @@ -56,6 +56,8 @@ unsafe impl UserSafeSized for Usercall {} #[unstable(feature = "sgx_platform", issue = "56975")] unsafe impl UserSafeSized for Return {} #[unstable(feature = "sgx_platform", issue = "56975")] +unsafe impl UserSafeSized for Cancel {} +#[unstable(feature = "sgx_platform", issue = "56975")] unsafe impl<T: UserSafeSized> UserSafeSized for [T; 2] {} /// A type that can be represented in memory as one or more `UserSafeSized`s. @@ -115,7 +117,7 @@ pub unsafe trait UserSafe { /// * the pointer is null. /// * the pointed-to range is not in user memory. unsafe fn check_ptr(ptr: *const Self) { - let is_aligned = |p| -> bool { 0 == (p as usize) & (Self::align_of() - 1) }; + let is_aligned = |p: *const u8| -> bool { p.is_aligned_to(Self::align_of()) }; assert!(is_aligned(ptr as *const u8)); assert!(is_user_range(ptr as _, mem::size_of_val(unsafe { &*ptr }))); @@ -305,6 +307,34 @@ where } } +// Split a memory region ptr..ptr + len into three parts: +// +--------+ +// | small0 | Chunk smaller than 8 bytes +// +--------+ +// | big | Chunk 8-byte aligned, and size a multiple of 8 bytes +// +--------+ +// | small1 | Chunk smaller than 8 bytes +// +--------+ +fn region_as_aligned_chunks(ptr: *const u8, len: usize) -> (usize, usize, usize) { + let small0_size = if ptr as usize % 8 == 0 { 0 } else { 8 - ptr as usize % 8 }; + let small1_size = (len - small0_size as usize) % 8; + let big_size = len - small0_size as usize - small1_size as usize; + + (small0_size, big_size, small1_size) +} + +unsafe fn copy_quadwords(src: *const u8, dst: *mut u8, len: usize) { + unsafe { + asm!( + "rep movsq (%rsi), (%rdi)", + inout("rcx") len / 8 => _, + inout("rdi") dst => _, + inout("rsi") src => _, + options(att_syntax, nostack, preserves_flags) + ); + } +} + /// Copies `len` bytes of data from enclave pointer `src` to userspace `dst` /// /// This function mitigates stale data vulnerabilities by ensuring all writes to untrusted memory are either: @@ -343,17 +373,6 @@ pub(crate) unsafe fn copy_to_userspace(src: *const u8, dst: *mut u8, len: usize) } } - unsafe fn copy_aligned_quadwords_to_userspace(src: *const u8, dst: *mut u8, len: usize) { - unsafe { - asm!( - "rep movsq (%rsi), (%rdi)", - inout("rcx") len / 8 => _, - inout("rdi") dst => _, - inout("rsi") src => _, - options(att_syntax, nostack, preserves_flags) - ); - } - } assert!(!src.is_null()); assert!(!dst.is_null()); assert!(is_enclave_range(src, len)); @@ -367,10 +386,10 @@ pub(crate) unsafe fn copy_to_userspace(src: *const u8, dst: *mut u8, len: usize) unsafe { copy_bytewise_to_userspace(src, dst, len); } - } else if len % 8 == 0 && dst as usize % 8 == 0 { + } else if len % 8 == 0 && dst.is_aligned_to(8) { // Copying 8-byte aligned quadwords: copy quad word per quad word unsafe { - copy_aligned_quadwords_to_userspace(src, dst, len); + copy_quadwords(src, dst, len); } } else { // Split copies into three parts: @@ -381,20 +400,16 @@ pub(crate) unsafe fn copy_to_userspace(src: *const u8, dst: *mut u8, len: usize) // +--------+ // | small1 | Chunk smaller than 8 bytes // +--------+ + let (small0_size, big_size, small1_size) = region_as_aligned_chunks(dst, len); unsafe { // Copy small0 - let small0_size = (8 - dst as usize % 8) as u8; - let small0_src = src; - let small0_dst = dst; - copy_bytewise_to_userspace(small0_src as _, small0_dst, small0_size as _); + copy_bytewise_to_userspace(src, dst, small0_size as _); // Copy big - let small1_size = ((len - small0_size as usize) % 8) as u8; - let big_size = len - small0_size as usize - small1_size as usize; let big_src = src.offset(small0_size as _); let big_dst = dst.offset(small0_size as _); - copy_aligned_quadwords_to_userspace(big_src as _, big_dst, big_size); + copy_quadwords(big_src as _, big_dst, big_size); // Copy small1 let small1_src = src.offset(big_size as isize + small0_size as isize); @@ -404,6 +419,106 @@ pub(crate) unsafe fn copy_to_userspace(src: *const u8, dst: *mut u8, len: usize) } } +/// Copies `len` bytes of data from userspace pointer `src` to enclave pointer `dst` +/// +/// This function mitigates AEPIC leak vulnerabilities by ensuring all reads from untrusted memory are 8-byte aligned +/// +/// # Panics +/// This function panics if: +/// +/// * The `src` pointer is null +/// * The `dst` pointer is null +/// * The `src` memory range is not in user memory +/// * The `dst` memory range is not in enclave memory +/// +/// # References +/// - https://www.intel.com/content/www/us/en/security-center/advisory/intel-sa-00657.html +/// - https://www.intel.com/content/www/us/en/developer/articles/technical/software-security-guidance/advisory-guidance/stale-data-read-from-xapic.html +pub(crate) unsafe fn copy_from_userspace(src: *const u8, dst: *mut u8, len: usize) { + // Copies memory region `src..src + len` to the enclave at `dst`. The source memory region + // is: + // - strictly less than 8 bytes in size and may be + // - located at a misaligned memory location + fn copy_misaligned_chunk_to_enclave(src: *const u8, dst: *mut u8, len: usize) { + let mut tmp_buff = [0u8; 16]; + + unsafe { + // Compute an aligned memory region to read from + // +--------+ <-- aligned_src + aligned_len (8B-aligned) + // | pad1 | + // +--------+ <-- src + len (misaligned) + // | | + // | | + // | | + // +--------+ <-- src (misaligned) + // | pad0 | + // +--------+ <-- aligned_src (8B-aligned) + let pad0_size = src as usize % 8; + let aligned_src = src.sub(pad0_size); + + let pad1_size = 8 - (src.add(len) as usize % 8); + let aligned_len = pad0_size + len + pad1_size; + + debug_assert!(len < 8); + debug_assert_eq!(aligned_src as usize % 8, 0); + debug_assert_eq!(aligned_len % 8, 0); + debug_assert!(aligned_len <= 16); + + // Copy the aligned buffer to a temporary buffer + // Note: copying from a slightly different memory location is a bit odd. In this case it + // can't lead to page faults or inadvertent copying from the enclave as we only ensured + // that the `src` pointer is aligned at an 8 byte boundary. As pages are 4096 bytes + // aligned, `aligned_src` must be on the same page as `src`. A similar argument can be made + // for `src + len` + copy_quadwords(aligned_src as _, tmp_buff.as_mut_ptr(), aligned_len); + + // Copy the correct parts of the temporary buffer to the destination + ptr::copy(tmp_buff.as_ptr().add(pad0_size), dst, len); + } + } + + assert!(!src.is_null()); + assert!(!dst.is_null()); + assert!(is_user_range(src, len)); + assert!(is_enclave_range(dst, len)); + assert!(!(src as usize).overflowing_add(len + 8).1); + assert!(!(dst as usize).overflowing_add(len + 8).1); + + if len < 8 { + copy_misaligned_chunk_to_enclave(src, dst, len); + } else if len % 8 == 0 && src as usize % 8 == 0 { + // Copying 8-byte aligned quadwords: copy quad word per quad word + unsafe { + copy_quadwords(src, dst, len); + } + } else { + // Split copies into three parts: + // +--------+ + // | small0 | Chunk smaller than 8 bytes + // +--------+ + // | big | Chunk 8-byte aligned, and size a multiple of 8 bytes + // +--------+ + // | small1 | Chunk smaller than 8 bytes + // +--------+ + let (small0_size, big_size, small1_size) = region_as_aligned_chunks(dst, len); + + unsafe { + // Copy small0 + copy_misaligned_chunk_to_enclave(src, dst, small0_size); + + // Copy big + let big_src = src.add(small0_size); + let big_dst = dst.add(small0_size); + copy_quadwords(big_src, big_dst, big_size); + + // Copy small1 + let small1_src = src.add(big_size + small0_size); + let small1_dst = dst.add(big_size + small0_size); + copy_misaligned_chunk_to_enclave(small1_src, small1_dst, small1_size); + } + } +} + #[unstable(feature = "sgx_platform", issue = "56975")] impl<T: ?Sized> UserRef<T> where @@ -468,7 +583,7 @@ where pub fn copy_to_enclave(&self, dest: &mut T) { unsafe { assert_eq!(mem::size_of_val(dest), mem::size_of_val(&*self.0.get())); - ptr::copy( + copy_from_userspace( self.0.get() as *const T as *const u8, dest as *mut T as *mut u8, mem::size_of_val(dest), @@ -494,7 +609,11 @@ where { /// Copies the value from user memory into enclave memory. pub fn to_enclave(&self) -> T { - unsafe { ptr::read(self.0.get()) } + unsafe { + let mut data: T = mem::MaybeUninit::uninit().assume_init(); + copy_from_userspace(self.0.get() as _, &mut data as *mut T as _, mem::size_of::<T>()); + data + } } } diff --git a/library/std/src/sys/sgx/abi/usercalls/mod.rs b/library/std/src/sys/sgx/abi/usercalls/mod.rs index 79d1db5e1..e19e84326 100644 --- a/library/std/src/sys/sgx/abi/usercalls/mod.rs +++ b/library/std/src/sys/sgx/abi/usercalls/mod.rs @@ -292,12 +292,17 @@ fn check_os_error(err: Result) -> i32 { } } -trait FromSgxResult { +/// Translate the raw result of an SGX usercall. +#[unstable(feature = "sgx_platform", issue = "56975")] +pub trait FromSgxResult { + /// Return type type Return; + /// Translate the raw result of an SGX usercall. fn from_sgx_result(self) -> IoResult<Self::Return>; } +#[unstable(feature = "sgx_platform", issue = "56975")] impl<T> FromSgxResult for (Result, T) { type Return = T; @@ -310,6 +315,7 @@ impl<T> FromSgxResult for (Result, T) { } } +#[unstable(feature = "sgx_platform", issue = "56975")] impl FromSgxResult for Result { type Return = (); diff --git a/library/std/src/sys/sgx/abi/usercalls/raw.rs b/library/std/src/sys/sgx/abi/usercalls/raw.rs index 4267b96cc..10c1456d4 100644 --- a/library/std/src/sys/sgx/abi/usercalls/raw.rs +++ b/library/std/src/sys/sgx/abi/usercalls/raw.rs @@ -37,14 +37,23 @@ pub unsafe fn do_usercall( (a, b) } -type Register = u64; +/// A value passed or returned in a CPU register. +#[unstable(feature = "sgx_platform", issue = "56975")] +pub type Register = u64; -trait RegisterArgument { +/// Translate a type from/to Register to be used as an argument. +#[unstable(feature = "sgx_platform", issue = "56975")] +pub trait RegisterArgument { + /// Translate a Register to Self. fn from_register(_: Register) -> Self; + /// Translate self to a Register. fn into_register(self) -> Register; } -trait ReturnValue { +/// Translate a pair of Registers to the raw usercall return value. +#[unstable(feature = "sgx_platform", issue = "56975")] +pub trait ReturnValue { + /// Translate a pair of Registers to the raw usercall return value. fn from_registers(call: &'static str, regs: (Register, Register)) -> Self; } @@ -68,6 +77,7 @@ macro_rules! define_usercalls { macro_rules! define_ra { (< $i:ident > $t:ty) => { + #[unstable(feature = "sgx_platform", issue = "56975")] impl<$i> RegisterArgument for $t { fn from_register(a: Register) -> Self { a as _ @@ -78,6 +88,7 @@ macro_rules! define_ra { } }; ($i:ty as $t:ty) => { + #[unstable(feature = "sgx_platform", issue = "56975")] impl RegisterArgument for $t { fn from_register(a: Register) -> Self { a as $i as _ @@ -88,6 +99,7 @@ macro_rules! define_ra { } }; ($t:ty) => { + #[unstable(feature = "sgx_platform", issue = "56975")] impl RegisterArgument for $t { fn from_register(a: Register) -> Self { a as _ @@ -112,6 +124,7 @@ define_ra!(usize as isize); define_ra!(<T> *const T); define_ra!(<T> *mut T); +#[unstable(feature = "sgx_platform", issue = "56975")] impl RegisterArgument for bool { fn from_register(a: Register) -> bool { if a != 0 { true } else { false } @@ -121,6 +134,7 @@ impl RegisterArgument for bool { } } +#[unstable(feature = "sgx_platform", issue = "56975")] impl<T: RegisterArgument> RegisterArgument for Option<NonNull<T>> { fn from_register(a: Register) -> Option<NonNull<T>> { NonNull::new(a as _) @@ -130,12 +144,14 @@ impl<T: RegisterArgument> RegisterArgument for Option<NonNull<T>> { } } +#[unstable(feature = "sgx_platform", issue = "56975")] impl ReturnValue for ! { fn from_registers(call: &'static str, _regs: (Register, Register)) -> Self { rtabort!("Usercall {call}: did not expect to be re-entered"); } } +#[unstable(feature = "sgx_platform", issue = "56975")] impl ReturnValue for () { fn from_registers(call: &'static str, usercall_retval: (Register, Register)) -> Self { rtassert!(usercall_retval.0 == 0); @@ -144,6 +160,7 @@ impl ReturnValue for () { } } +#[unstable(feature = "sgx_platform", issue = "56975")] impl<T: RegisterArgument> ReturnValue for T { fn from_registers(call: &'static str, usercall_retval: (Register, Register)) -> Self { rtassert!(usercall_retval.1 == 0); @@ -151,6 +168,7 @@ impl<T: RegisterArgument> ReturnValue for T { } } +#[unstable(feature = "sgx_platform", issue = "56975")] impl<T: RegisterArgument, U: RegisterArgument> ReturnValue for (T, U) { fn from_registers(_call: &'static str, regs: (Register, Register)) -> Self { (T::from_register(regs.0), U::from_register(regs.1)) diff --git a/library/std/src/sys/sgx/abi/usercalls/tests.rs b/library/std/src/sys/sgx/abi/usercalls/tests.rs index cbf7d7d54..58b8eb215 100644 --- a/library/std/src/sys/sgx/abi/usercalls/tests.rs +++ b/library/std/src/sys/sgx/abi/usercalls/tests.rs @@ -1,8 +1,8 @@ -use super::alloc::copy_to_userspace; use super::alloc::User; +use super::alloc::{copy_from_userspace, copy_to_userspace}; #[test] -fn test_copy_function() { +fn test_copy_to_userspace_function() { let mut src = [0u8; 100]; let mut dst = User::<[u8]>::uninitialized(100); @@ -17,12 +17,38 @@ fn test_copy_function() { dst.copy_from_enclave(&[0u8; 100]); // Copy src[0..size] to dst + offset - unsafe { copy_to_userspace(src.as_ptr(), dst.as_mut_ptr().offset(offset), size) }; + unsafe { copy_to_userspace(src.as_ptr(), dst.as_mut_ptr().add(offset), size) }; // Verify copy for byte in 0..size { unsafe { - assert_eq!(*dst.as_ptr().offset(offset + byte as isize), src[byte as usize]); + assert_eq!(*dst.as_ptr().add(offset + byte), src[byte as usize]); + } + } + } + } +} + +#[test] +fn test_copy_from_userspace_function() { + let mut dst = [0u8; 100]; + let mut src = User::<[u8]>::uninitialized(100); + + src.copy_from_enclave(&[0u8; 100]); + + for size in 0..48 { + // For all possible alignment + for offset in 0..8 { + // overwrite complete dst + dst = [0u8; 100]; + + // Copy src[0..size] to dst + offset + unsafe { copy_from_userspace(src.as_ptr().offset(offset), dst.as_mut_ptr(), size) }; + + // Verify copy + for byte in 0..size { + unsafe { + assert_eq!(dst[byte as usize], *src.as_ptr().offset(offset + byte as isize)); } } } diff --git a/library/std/src/sys/sgx/mod.rs b/library/std/src/sys/sgx/mod.rs index 696400670..b1d32929e 100644 --- a/library/std/src/sys/sgx/mod.rs +++ b/library/std/src/sys/sgx/mod.rs @@ -47,7 +47,7 @@ pub mod locks { // SAFETY: must be called only once during runtime initialization. // NOTE: this is not guaranteed to run, for example when Rust code is called externally. -pub unsafe fn init(argc: isize, argv: *const *const u8) { +pub unsafe fn init(argc: isize, argv: *const *const u8, _sigpipe: u8) { unsafe { args::init(argc, argv); } diff --git a/library/std/src/sys/sgx/mutex.rs b/library/std/src/sys/sgx/mutex.rs index 513cd77fd..aa747d56b 100644 --- a/library/std/src/sys/sgx/mutex.rs +++ b/library/std/src/sys/sgx/mutex.rs @@ -21,9 +21,6 @@ impl Mutex { } #[inline] - pub unsafe fn init(&mut self) {} - - #[inline] pub unsafe fn lock(&self) { let mut guard = self.inner.lock(); if *guard.lock_var() { diff --git a/library/std/src/sys/solid/fs.rs b/library/std/src/sys/solid/fs.rs index a2cbee4dc..969222253 100644 --- a/library/std/src/sys/solid/fs.rs +++ b/library/std/src/sys/solid/fs.rs @@ -2,7 +2,7 @@ use super::{abi, error}; use crate::{ ffi::{CStr, CString, OsStr, OsString}, fmt, - io::{self, IoSlice, IoSliceMut, ReadBuf, SeekFrom}, + io::{self, BorrowedCursor, IoSlice, IoSliceMut, SeekFrom}, mem::MaybeUninit, os::raw::{c_int, c_short}, os::solid::ffi::OsStrExt, @@ -77,6 +77,9 @@ pub struct OpenOptions { custom_flags: i32, } +#[derive(Copy, Clone, Debug, Default)] +pub struct FileTimes {} + #[derive(Clone, PartialEq, Eq, Debug)] pub struct FilePermissions(c_short); @@ -126,6 +129,11 @@ impl FilePermissions { } } +impl FileTimes { + pub fn set_accessed(&mut self, _t: SystemTime) {} + pub fn set_modified(&mut self, _t: SystemTime) {} +} + impl FileType { pub fn is_dir(&self) -> bool { self.is(abi::S_IFDIR) @@ -358,13 +366,13 @@ impl File { } } - pub fn read_buf(&self, buf: &mut ReadBuf<'_>) -> io::Result<()> { + pub fn read_buf(&self, mut cursor: BorrowedCursor<'_>) -> io::Result<()> { unsafe { - let len = buf.remaining(); + let len = cursor.capacity(); let mut out_num_bytes = MaybeUninit::uninit(); error::SolidError::err_if_negative(abi::SOLID_FS_Read( self.fd.raw(), - buf.unfilled_mut().as_mut_ptr() as *mut u8, + cursor.as_mut().as_mut_ptr() as *mut u8, len, out_num_bytes.as_mut_ptr(), )) @@ -376,9 +384,7 @@ impl File { // Safety: `num_bytes_read` bytes were written to the unfilled // portion of the buffer - buf.assume_init(num_bytes_read); - - buf.add_filled(num_bytes_read); + cursor.advance(num_bytes_read); Ok(()) } @@ -452,6 +458,10 @@ impl File { pub fn set_permissions(&self, _perm: FilePermissions) -> io::Result<()> { unsupported() } + + pub fn set_times(&self, _times: FileTimes) -> io::Result<()> { + unsupported() + } } impl Drop for File { diff --git a/library/std/src/sys/solid/mod.rs b/library/std/src/sys/solid/mod.rs index 778a589d1..5867979a2 100644 --- a/library/std/src/sys/solid/mod.rs +++ b/library/std/src/sys/solid/mod.rs @@ -56,7 +56,7 @@ pub mod locks { // SAFETY: must be called only once during runtime initialization. // NOTE: this is not guaranteed to run, for example when Rust code is called externally. -pub unsafe fn init(_argc: isize, _argv: *const *const u8) {} +pub unsafe fn init(_argc: isize, _argv: *const *const u8, _sigpipe: u8) {} // SAFETY: must be called only once during runtime cleanup. pub unsafe fn cleanup() {} diff --git a/library/std/src/sys/unix/fd.rs b/library/std/src/sys/unix/fd.rs index 30812dabb..dbaa3c33e 100644 --- a/library/std/src/sys/unix/fd.rs +++ b/library/std/src/sys/unix/fd.rs @@ -4,7 +4,7 @@ mod tests; use crate::cmp; -use crate::io::{self, IoSlice, IoSliceMut, Read, ReadBuf}; +use crate::io::{self, BorrowedCursor, IoSlice, IoSliceMut, Read}; use crate::os::unix::io::{AsFd, AsRawFd, BorrowedFd, FromRawFd, IntoRawFd, OwnedFd, RawFd}; use crate::sys::cvt; use crate::sys_common::{AsInner, FromInner, IntoInner}; @@ -131,20 +131,19 @@ impl FileDesc { } } - pub fn read_buf(&self, buf: &mut ReadBuf<'_>) -> io::Result<()> { + pub fn read_buf(&self, mut cursor: BorrowedCursor<'_>) -> io::Result<()> { let ret = cvt(unsafe { libc::read( self.as_raw_fd(), - buf.unfilled_mut().as_mut_ptr() as *mut libc::c_void, - cmp::min(buf.remaining(), READ_LIMIT), + cursor.as_mut().as_mut_ptr() as *mut libc::c_void, + cmp::min(cursor.capacity(), READ_LIMIT), ) })?; // Safety: `ret` bytes were written to the initialized portion of the buffer unsafe { - buf.assume_init(ret as usize); + cursor.advance(ret as usize); } - buf.add_filled(ret as usize); Ok(()) } diff --git a/library/std/src/sys/unix/fs.rs b/library/std/src/sys/unix/fs.rs index b5cc8038c..cc347e358 100644 --- a/library/std/src/sys/unix/fs.rs +++ b/library/std/src/sys/unix/fs.rs @@ -2,7 +2,7 @@ use crate::os::unix::prelude::*; use crate::ffi::{CStr, CString, OsStr, OsString}; use crate::fmt; -use crate::io::{self, Error, IoSlice, IoSliceMut, ReadBuf, SeekFrom}; +use crate::io::{self, BorrowedCursor, Error, IoSlice, IoSliceMut, SeekFrom}; use crate::mem; use crate::os::unix::io::{AsFd, AsRawFd, BorrowedFd, FromRawFd, IntoRawFd}; use crate::path::{Path, PathBuf}; @@ -544,11 +544,11 @@ impl Default for FileTimes { fn default() -> Self { // Redox doesn't appear to support `UTIME_OMIT`, so we stub it out here, and always return // an error in `set_times`. - // ESP-IDF does not support `futimens` at all and the behavior for that OS is therefore + // ESP-IDF and HorizonOS do not support `futimens` at all and the behavior for those OS is therefore // the same as for Redox. - #[cfg(any(target_os = "redox", target_os = "espidf"))] + #[cfg(any(target_os = "redox", target_os = "espidf", target_os = "horizon"))] let omit = libc::timespec { tv_sec: 0, tv_nsec: 0 }; - #[cfg(not(any(target_os = "redox", target_os = "espidf")))] + #[cfg(not(any(target_os = "redox", target_os = "espidf", target_os = "horizon")))] let omit = libc::timespec { tv_sec: 0, tv_nsec: libc::UTIME_OMIT as _ }; Self([omit; 2]) } @@ -687,7 +687,11 @@ impl Iterator for ReadDir { impl Drop for Dir { fn drop(&mut self) { let r = unsafe { libc::closedir(self.0) }; - debug_assert_eq!(r, 0); + assert!( + r == 0 || crate::io::Error::last_os_error().kind() == crate::io::ErrorKind::Interrupted, + "unexpected error during closedir: {:?}", + crate::io::Error::last_os_error() + ); } } @@ -825,6 +829,7 @@ impl DirEntry { target_os = "fuchsia", target_os = "redox" )))] + #[cfg_attr(miri, allow(unused))] fn name_cstr(&self) -> &CStr { unsafe { CStr::from_ptr(self.entry.d_name.as_ptr()) } } @@ -836,6 +841,7 @@ impl DirEntry { target_os = "fuchsia", target_os = "redox" ))] + #[cfg_attr(miri, allow(unused))] fn name_cstr(&self) -> &CStr { &self.name } @@ -1031,8 +1037,8 @@ impl File { self.0.read_at(buf, offset) } - pub fn read_buf(&self, buf: &mut ReadBuf<'_>) -> io::Result<()> { - self.0.read_buf(buf) + pub fn read_buf(&self, cursor: BorrowedCursor<'_>) -> io::Result<()> { + self.0.read_buf(cursor) } pub fn write(&self, buf: &[u8]) -> io::Result<usize> { @@ -1079,9 +1085,9 @@ impl File { pub fn set_times(&self, times: FileTimes) -> io::Result<()> { cfg_if::cfg_if! { - if #[cfg(any(target_os = "redox", target_os = "espidf"))] { + if #[cfg(any(target_os = "redox", target_os = "espidf", target_os = "horizon"))] { // Redox doesn't appear to support `UTIME_OMIT`. - // ESP-IDF does not support `futimens` at all and the behavior for that OS is therefore + // ESP-IDF and HorizonOS do not support `futimens` at all and the behavior for those OS is therefore // the same as for Redox. drop(times); Err(io::const_io_error!( diff --git a/library/std/src/sys/unix/locks/fuchsia_mutex.rs b/library/std/src/sys/unix/locks/fuchsia_mutex.rs index ce427599c..117611ce4 100644 --- a/library/std/src/sys/unix/locks/fuchsia_mutex.rs +++ b/library/std/src/sys/unix/locks/fuchsia_mutex.rs @@ -86,9 +86,6 @@ impl Mutex { } #[inline] - pub unsafe fn init(&mut self) {} - - #[inline] pub unsafe fn try_lock(&self) -> bool { let thread_self = zx_thread_self(); self.futex.compare_exchange(UNLOCKED, to_state(thread_self), Acquire, Relaxed).is_ok() @@ -138,7 +135,7 @@ impl Mutex { } } - // The state has changed or a wakeup occured, try to lock the mutex. + // The state has changed or a wakeup occurred, try to lock the mutex. match self.futex.compare_exchange(UNLOCKED, owned_state, Acquire, Relaxed) { Ok(_) => return, Err(updated) => state = updated, diff --git a/library/std/src/sys/unix/locks/futex_mutex.rs b/library/std/src/sys/unix/locks/futex_mutex.rs index 99ba86e5f..33b13dad4 100644 --- a/library/std/src/sys/unix/locks/futex_mutex.rs +++ b/library/std/src/sys/unix/locks/futex_mutex.rs @@ -20,9 +20,6 @@ impl Mutex { } #[inline] - pub unsafe fn init(&mut self) {} - - #[inline] pub unsafe fn try_lock(&self) -> bool { self.futex.compare_exchange(0, 1, Acquire, Relaxed).is_ok() } @@ -53,7 +50,7 @@ impl Mutex { // We avoid an unnecessary write if it as already set to 2, // to be friendlier for the caches. if state != 2 && self.futex.swap(2, Acquire) == 0 { - // We changed it from 0 to 2, so we just succesfully locked it. + // We changed it from 0 to 2, so we just successfully locked it. return; } diff --git a/library/std/src/sys/unix/locks/futex_rwlock.rs b/library/std/src/sys/unix/locks/futex_rwlock.rs index b3bbbf743..0cc92244e 100644 --- a/library/std/src/sys/unix/locks/futex_rwlock.rs +++ b/library/std/src/sys/unix/locks/futex_rwlock.rs @@ -54,7 +54,7 @@ fn is_read_lockable(state: u32) -> bool { // We don't allow read-locking if there's readers waiting, even if the lock is unlocked // and there's no writers waiting. The only situation when this happens is after unlocking, // at which point the unlocking thread might be waking up writers, which have priority over readers. - // The unlocking thread will clear the readers waiting bit and wake up readers, if necssary. + // The unlocking thread will clear the readers waiting bit and wake up readers, if necessary. state & MASK < MAX_READERS && !has_readers_waiting(state) && !has_writers_waiting(state) } diff --git a/library/std/src/sys/unix/locks/pthread_condvar.rs b/library/std/src/sys/unix/locks/pthread_condvar.rs index abf27e7db..4741c0c67 100644 --- a/library/std/src/sys/unix/locks/pthread_condvar.rs +++ b/library/std/src/sys/unix/locks/pthread_condvar.rs @@ -172,7 +172,7 @@ impl Condvar { let mut sys_now = libc::timeval { tv_sec: 0, tv_usec: 0 }; let stable_now = Instant::now(); let r = libc::gettimeofday(&mut sys_now, ptr::null_mut()); - debug_assert_eq!(r, 0); + assert_eq!(r, 0, "unexpected error: {:?}", crate::io::Error::last_os_error()); let nsec = dur.subsec_nanos() as libc::c_long + (sys_now.tv_usec * 1000) as libc::c_long; let extra = (nsec / 1_000_000_000) as libc::time_t; diff --git a/library/std/src/sys/unix/locks/pthread_mutex.rs b/library/std/src/sys/unix/locks/pthread_mutex.rs index 98afee69b..5964935dd 100644 --- a/library/std/src/sys/unix/locks/pthread_mutex.rs +++ b/library/std/src/sys/unix/locks/pthread_mutex.rs @@ -52,7 +52,7 @@ impl Mutex { Mutex { inner: UnsafeCell::new(libc::PTHREAD_MUTEX_INITIALIZER) } } #[inline] - pub unsafe fn init(&mut self) { + unsafe fn init(&mut self) { // Issue #33770 // // A pthread mutex initialized with PTHREAD_MUTEX_INITIALIZER will have diff --git a/library/std/src/sys/unix/mod.rs b/library/std/src/sys/unix/mod.rs index 3d0d91460..c84e292ea 100644 --- a/library/std/src/sys/unix/mod.rs +++ b/library/std/src/sys/unix/mod.rs @@ -44,12 +44,13 @@ pub mod thread_parker; pub mod time; #[cfg(target_os = "espidf")] -pub fn init(argc: isize, argv: *const *const u8) {} +pub fn init(argc: isize, argv: *const *const u8, _sigpipe: u8) {} #[cfg(not(target_os = "espidf"))] // SAFETY: must be called only once during runtime initialization. // NOTE: this is not guaranteed to run, for example when Rust code is called externally. -pub unsafe fn init(argc: isize, argv: *const *const u8) { +// See `fn init()` in `library/std/src/rt.rs` for docs on `sigpipe`. +pub unsafe fn init(argc: isize, argv: *const *const u8, sigpipe: u8) { // The standard streams might be closed on application startup. To prevent // std::io::{stdin, stdout,stderr} objects from using other unrelated file // resources opened later, we reopen standards streams when they are closed. @@ -61,8 +62,9 @@ pub unsafe fn init(argc: isize, argv: *const *const u8) { // want! // // Hence, we set SIGPIPE to ignore when the program starts up in order - // to prevent this problem. - reset_sigpipe(); + // to prevent this problem. Add `#[unix_sigpipe = "..."]` above `fn main()` to + // alter this behavior. + reset_sigpipe(sigpipe); stack_overflow::init(); args::init(argc, argv); @@ -151,9 +153,31 @@ pub unsafe fn init(argc: isize, argv: *const *const u8) { } } - unsafe fn reset_sigpipe() { + unsafe fn reset_sigpipe(#[allow(unused_variables)] sigpipe: u8) { #[cfg(not(any(target_os = "emscripten", target_os = "fuchsia", target_os = "horizon")))] - rtassert!(signal(libc::SIGPIPE, libc::SIG_IGN) != libc::SIG_ERR); + { + // We don't want to add this as a public type to libstd, nor do we + // want to `include!` a file from the compiler (which would break + // Miri and xargo for example), so we choose to duplicate these + // constants from `compiler/rustc_session/src/config/sigpipe.rs`. + // See the other file for docs. NOTE: Make sure to keep them in + // sync! + mod sigpipe { + pub const INHERIT: u8 = 1; + pub const SIG_IGN: u8 = 2; + pub const SIG_DFL: u8 = 3; + } + + let handler = match sigpipe { + sigpipe::INHERIT => None, + sigpipe::SIG_IGN => Some(libc::SIG_IGN), + sigpipe::SIG_DFL => Some(libc::SIG_DFL), + _ => unreachable!(), + }; + if let Some(handler) = handler { + rtassert!(signal(libc::SIGPIPE, handler) != libc::SIG_ERR); + } + } } } @@ -295,8 +319,10 @@ pub fn abort_internal() -> ! { cfg_if::cfg_if! { if #[cfg(target_os = "android")] { - #[link(name = "dl")] - #[link(name = "log")] + #[link(name = "dl", kind = "static", modifiers = "-bundle", + cfg(target_feature = "crt-static"))] + #[link(name = "dl", cfg(not(target_feature = "crt-static")))] + #[link(name = "log", cfg(not(target_feature = "crt-static")))] extern "C" {} } else if #[cfg(target_os = "freebsd")] { #[link(name = "execinfo")] diff --git a/library/std/src/sys/unix/net.rs b/library/std/src/sys/unix/net.rs index 462a45b01..b84bf8f92 100644 --- a/library/std/src/sys/unix/net.rs +++ b/library/std/src/sys/unix/net.rs @@ -393,6 +393,17 @@ impl Socket { } #[cfg(any(target_os = "android", target_os = "linux",))] + pub fn set_quickack(&self, quickack: bool) -> io::Result<()> { + setsockopt(self, libc::IPPROTO_TCP, libc::TCP_QUICKACK, quickack as c_int) + } + + #[cfg(any(target_os = "android", target_os = "linux",))] + pub fn quickack(&self) -> io::Result<bool> { + let raw: c_int = getsockopt(self, libc::IPPROTO_TCP, libc::TCP_QUICKACK)?; + Ok(raw != 0) + } + + #[cfg(any(target_os = "android", target_os = "linux",))] pub fn set_passcred(&self, passcred: bool) -> io::Result<()> { setsockopt(self, libc::SOL_SOCKET, libc::SO_PASSCRED, passcred as libc::c_int) } @@ -427,6 +438,17 @@ impl Socket { self.0.set_nonblocking(nonblocking) } + #[cfg(any(target_os = "linux", target_os = "freebsd", target_os = "openbsd"))] + pub fn set_mark(&self, mark: u32) -> io::Result<()> { + #[cfg(target_os = "linux")] + let option = libc::SO_MARK; + #[cfg(target_os = "freebsd")] + let option = libc::SO_USER_COOKIE; + #[cfg(target_os = "openbsd")] + let option = libc::SO_RTABLE; + setsockopt(self, libc::SOL_SOCKET, option, mark as libc::c_int) + } + pub fn take_error(&self) -> io::Result<Option<io::Error>> { let raw: c_int = getsockopt(self, libc::SOL_SOCKET, libc::SO_ERROR)?; if raw == 0 { Ok(None) } else { Ok(Some(io::Error::from_raw_os_error(raw as i32))) } diff --git a/library/std/src/sys/unix/os_str.rs b/library/std/src/sys/unix/os_str.rs index ccbc18224..017e2af29 100644 --- a/library/std/src/sys/unix/os_str.rs +++ b/library/std/src/sys/unix/os_str.rs @@ -11,7 +11,7 @@ use crate::str; use crate::sync::Arc; use crate::sys_common::{AsInner, IntoInner}; -use core::str::lossy::{Utf8Lossy, Utf8LossyChunk}; +use core::str::Utf8Chunks; #[cfg(test)] #[path = "../unix/os_str/tests.rs"] @@ -29,26 +29,32 @@ pub struct Slice { } impl fmt::Debug for Slice { - fn fmt(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result { - // Writes out a valid unicode string with the correct escape sequences - - formatter.write_str("\"")?; - for Utf8LossyChunk { valid, broken } in Utf8Lossy::from_bytes(&self.inner).chunks() { - for c in valid.chars().flat_map(|c| c.escape_debug()) { - formatter.write_char(c)? - } - - for b in broken { - write!(formatter, "\\x{:02X}", b)?; - } - } - formatter.write_str("\"") + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + fmt::Debug::fmt(&Utf8Chunks::new(&self.inner).debug(), f) } } impl fmt::Display for Slice { - fn fmt(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result { - fmt::Display::fmt(&Utf8Lossy::from_bytes(&self.inner), formatter) + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + // If we're the empty string then our iterator won't actually yield + // anything, so perform the formatting manually + if self.inner.is_empty() { + return "".fmt(f); + } + + for chunk in Utf8Chunks::new(&self.inner) { + let valid = chunk.valid(); + // If we successfully decoded the whole chunk as a valid string then + // we can return a direct formatting of the string which will also + // respect various formatting flags if possible. + if chunk.invalid().is_empty() { + return valid.fmt(f); + } + + f.write_str(valid)?; + f.write_char(char::REPLACEMENT_CHARACTER)?; + } + Ok(()) } } diff --git a/library/std/src/sys/unix/os_str/tests.rs b/library/std/src/sys/unix/os_str/tests.rs index 213277f01..22ba0c923 100644 --- a/library/std/src/sys/unix/os_str/tests.rs +++ b/library/std/src/sys/unix/os_str/tests.rs @@ -8,3 +8,11 @@ fn slice_debug_output() { assert_eq!(output, expected); } + +#[test] +fn display() { + assert_eq!( + "Hello\u{FFFD}\u{FFFD} There\u{FFFD} Goodbye", + Slice::from_u8_slice(b"Hello\xC0\x80 There\xE6\x83 Goodbye").to_string(), + ); +} diff --git a/library/std/src/sys/unix/process/process_common.rs b/library/std/src/sys/unix/process/process_common.rs index bca1b65a7..2834ee0ac 100644 --- a/library/std/src/sys/unix/process/process_common.rs +++ b/library/std/src/sys/unix/process/process_common.rs @@ -45,11 +45,31 @@ cfg_if::cfg_if! { } #[allow(dead_code)] pub unsafe fn sigaddset(set: *mut libc::sigset_t, signum: libc::c_int) -> libc::c_int { - use crate::{slice, mem}; + use crate::{ + mem::{align_of, size_of}, + slice, + }; + use libc::{c_ulong, sigset_t}; + + // The implementations from bionic (android libc) type pun `sigset_t` as an + // array of `c_ulong`. This works, but lets add a smoke check to make sure + // that doesn't change. + const _: () = assert!( + align_of::<c_ulong>() == align_of::<sigset_t>() + && (size_of::<sigset_t>() % size_of::<c_ulong>()) == 0 + ); - let raw = slice::from_raw_parts_mut(set as *mut u8, mem::size_of::<libc::sigset_t>()); let bit = (signum - 1) as usize; - raw[bit / 8] |= 1 << (bit % 8); + if set.is_null() || bit >= (8 * size_of::<sigset_t>()) { + crate::sys::unix::os::set_errno(libc::EINVAL); + return -1; + } + let raw = slice::from_raw_parts_mut( + set as *mut c_ulong, + size_of::<sigset_t>() / size_of::<c_ulong>(), + ); + const LONG_BIT: usize = size_of::<c_ulong>() * 8; + raw[bit / LONG_BIT] |= 1 << (bit % LONG_BIT); return 0; } } else { @@ -72,6 +92,7 @@ pub struct Command { argv: Argv, env: CommandEnv, + program_kind: ProgramKind, cwd: Option<CString>, uid: Option<uid_t>, gid: Option<gid_t>, @@ -128,15 +149,40 @@ pub enum Stdio { Fd(FileDesc), } +#[derive(Copy, Clone, Debug, Eq, PartialEq)] +pub enum ProgramKind { + /// A program that would be looked up on the PATH (e.g. `ls`) + PathLookup, + /// A relative path (e.g. `my-dir/foo`, `../foo`, `./foo`) + Relative, + /// An absolute path. + Absolute, +} + +impl ProgramKind { + fn new(program: &OsStr) -> Self { + if program.bytes().starts_with(b"/") { + Self::Absolute + } else if program.bytes().contains(&b'/') { + // If the program has more than one component in it, it is a relative path. + Self::Relative + } else { + Self::PathLookup + } + } +} + impl Command { #[cfg(not(target_os = "linux"))] pub fn new(program: &OsStr) -> Command { let mut saw_nul = false; + let program_kind = ProgramKind::new(program.as_ref()); let program = os2c(program, &mut saw_nul); Command { argv: Argv(vec![program.as_ptr(), ptr::null()]), args: vec![program.clone()], program, + program_kind, env: Default::default(), cwd: None, uid: None, @@ -154,11 +200,13 @@ impl Command { #[cfg(target_os = "linux")] pub fn new(program: &OsStr) -> Command { let mut saw_nul = false; + let program_kind = ProgramKind::new(program.as_ref()); let program = os2c(program, &mut saw_nul); Command { argv: Argv(vec![program.as_ptr(), ptr::null()]), args: vec![program.clone()], program, + program_kind, env: Default::default(), cwd: None, uid: None, @@ -234,6 +282,11 @@ impl Command { OsStr::from_bytes(self.program.as_bytes()) } + #[allow(dead_code)] + pub fn get_program_kind(&self) -> ProgramKind { + self.program_kind + } + pub fn get_args(&self) -> CommandArgs<'_> { let mut iter = self.args.iter(); iter.next(); diff --git a/library/std/src/sys/unix/process/process_common/tests.rs b/library/std/src/sys/unix/process/process_common/tests.rs index 1956b3692..d176b3401 100644 --- a/library/std/src/sys/unix/process/process_common/tests.rs +++ b/library/std/src/sys/unix/process/process_common/tests.rs @@ -122,3 +122,27 @@ fn test_process_group_no_posix_spawn() { t!(cat.wait()); } } + +#[test] +fn test_program_kind() { + let vectors = &[ + ("foo", ProgramKind::PathLookup), + ("foo.out", ProgramKind::PathLookup), + ("./foo", ProgramKind::Relative), + ("../foo", ProgramKind::Relative), + ("dir/foo", ProgramKind::Relative), + // Note that paths on Unix can't contain / in them, so this is actually the directory "fo\\" + // followed by the file "o". + ("fo\\/o", ProgramKind::Relative), + ("/foo", ProgramKind::Absolute), + ("/dir/../foo", ProgramKind::Absolute), + ]; + + for (program, expected_kind) in vectors { + assert_eq!( + ProgramKind::new(program.as_ref()), + *expected_kind, + "actual != expected program kind for input {program}", + ); + } +} diff --git a/library/std/src/sys/unix/process/process_unix.rs b/library/std/src/sys/unix/process/process_unix.rs index 75bb92437..26ae62817 100644 --- a/library/std/src/sys/unix/process/process_unix.rs +++ b/library/std/src/sys/unix/process/process_unix.rs @@ -453,7 +453,9 @@ impl Command { // successfully launch the program, but erroneously return // ENOENT when used with posix_spawn_file_actions_addchdir_np // which was introduced in macOS 10.15. - return Ok(None); + if self.get_program_kind() == ProgramKind::Relative { + return Ok(None); + } } match posix_spawn_file_actions_addchdir_np.get() { Some(f) => Some((f, cwd)), diff --git a/library/std/src/sys/unix/rand.rs b/library/std/src/sys/unix/rand.rs index bf4920488..a6fe07873 100644 --- a/library/std/src/sys/unix/rand.rs +++ b/library/std/src/sys/unix/rand.rs @@ -1,13 +1,13 @@ -use crate::mem; -use crate::slice; - pub fn hashmap_random_keys() -> (u64, u64) { - let mut v = (0, 0); - unsafe { - let view = slice::from_raw_parts_mut(&mut v as *mut _ as *mut u8, mem::size_of_val(&v)); - imp::fill_bytes(view); - } - v + const KEY_LEN: usize = core::mem::size_of::<u64>(); + + let mut v = [0u8; KEY_LEN * 2]; + imp::fill_bytes(&mut v); + + let key1 = v[0..KEY_LEN].try_into().unwrap(); + let key2 = v[KEY_LEN..].try_into().unwrap(); + + (u64::from_ne_bytes(key1), u64::from_ne_bytes(key2)) } #[cfg(all( diff --git a/library/std/src/sys/unix/thread.rs b/library/std/src/sys/unix/thread.rs index 36a3fa602..f6b627afc 100644 --- a/library/std/src/sys/unix/thread.rs +++ b/library/std/src/sys/unix/thread.rs @@ -116,11 +116,9 @@ impl Thread { debug_assert_eq!(ret, 0); } - #[cfg(any(target_os = "linux", target_os = "android"))] + #[cfg(target_os = "android")] pub fn set_name(name: &CStr) { const PR_SET_NAME: libc::c_int = 15; - // pthread wrapper only appeared in glibc 2.12, so we use syscall - // directly. unsafe { libc::prctl( PR_SET_NAME, @@ -132,6 +130,17 @@ impl Thread { } } + #[cfg(target_os = "linux")] + pub fn set_name(name: &CStr) { + const TASK_COMM_LEN: usize = 16; + + unsafe { + // Available since glibc 2.12, musl 1.1.16, and uClibc 1.0.20. + let name = truncate_cstr(name, TASK_COMM_LEN); + libc::pthread_setname_np(libc::pthread_self(), name.as_ptr()); + } + } + #[cfg(any(target_os = "freebsd", target_os = "dragonfly", target_os = "openbsd"))] pub fn set_name(name: &CStr) { unsafe { @@ -142,6 +151,7 @@ impl Thread { #[cfg(any(target_os = "macos", target_os = "ios", target_os = "watchos"))] pub fn set_name(name: &CStr) { unsafe { + let name = truncate_cstr(name, libc::MAXTHREADNAMESIZE); libc::pthread_setname_np(name.as_ptr()); } } @@ -271,6 +281,20 @@ impl Drop for Thread { } } +#[cfg(any(target_os = "linux", target_os = "macos", target_os = "ios", target_os = "watchos"))] +fn truncate_cstr(cstr: &CStr, max_with_nul: usize) -> crate::borrow::Cow<'_, CStr> { + use crate::{borrow::Cow, ffi::CString}; + + if cstr.to_bytes_with_nul().len() > max_with_nul { + let bytes = cstr.to_bytes()[..max_with_nul - 1].to_vec(); + // SAFETY: the non-nul bytes came straight from a CStr. + // (CString will add the terminating nul.) + Cow::Owned(unsafe { CString::from_vec_unchecked(bytes) }) + } else { + Cow::Borrowed(cstr) + } +} + pub fn available_parallelism() -> io::Result<NonZeroUsize> { cfg_if::cfg_if! { if #[cfg(any( @@ -423,7 +447,7 @@ mod cgroups { Some(b"") => Cgroup::V2, Some(controllers) if from_utf8(controllers) - .is_ok_and(|c| c.split(",").any(|c| c == "cpu")) => + .is_ok_and(|c| c.split(',').any(|c| c == "cpu")) => { Cgroup::V1 } diff --git a/library/std/src/sys/unix/thread_parker/mod.rs b/library/std/src/sys/unix/thread_parker/mod.rs new file mode 100644 index 000000000..e2453580d --- /dev/null +++ b/library/std/src/sys/unix/thread_parker/mod.rs @@ -0,0 +1,21 @@ +//! Thread parking on systems without futex support. + +#![cfg(not(any( + target_os = "linux", + target_os = "android", + all(target_os = "emscripten", target_feature = "atomics"), + target_os = "freebsd", + target_os = "openbsd", + target_os = "dragonfly", + target_os = "fuchsia", +)))] + +cfg_if::cfg_if! { + if #[cfg(target_os = "netbsd")] { + mod netbsd; + pub use netbsd::Parker; + } else { + mod pthread; + pub use pthread::Parker; + } +} diff --git a/library/std/src/sys/unix/thread_parker/netbsd.rs b/library/std/src/sys/unix/thread_parker/netbsd.rs new file mode 100644 index 000000000..7657605b5 --- /dev/null +++ b/library/std/src/sys/unix/thread_parker/netbsd.rs @@ -0,0 +1,113 @@ +use crate::ffi::{c_int, c_void}; +use crate::pin::Pin; +use crate::ptr::{null, null_mut}; +use crate::sync::atomic::{ + AtomicU64, + Ordering::{Acquire, Relaxed, Release}, +}; +use crate::time::Duration; +use libc::{_lwp_self, clockid_t, lwpid_t, time_t, timespec, CLOCK_MONOTONIC}; + +extern "C" { + fn ___lwp_park60( + clock_id: clockid_t, + flags: c_int, + ts: *mut timespec, + unpark: lwpid_t, + hint: *const c_void, + unparkhint: *const c_void, + ) -> c_int; + fn _lwp_unpark(lwp: lwpid_t, hint: *const c_void) -> c_int; +} + +/// The thread is not parked and the token is not available. +/// +/// Zero cannot be a valid LWP id, since it is used as empty value for the unpark +/// argument in _lwp_park. +const EMPTY: u64 = 0; +/// The token is available. Do not park anymore. +const NOTIFIED: u64 = u64::MAX; + +pub struct Parker { + /// The parker state. Contains either one of the two state values above or the LWP + /// id of the parked thread. + state: AtomicU64, +} + +impl Parker { + pub unsafe fn new(parker: *mut Parker) { + parker.write(Parker { state: AtomicU64::new(EMPTY) }) + } + + // Does not actually need `unsafe` or `Pin`, but the pthread implementation does. + pub unsafe fn park(self: Pin<&Self>) { + // If the token has already been made available, we can skip + // a bit of work, so check for it here. + if self.state.load(Acquire) != NOTIFIED { + let parked = _lwp_self() as u64; + let hint = self.state.as_mut_ptr().cast(); + if self.state.compare_exchange(EMPTY, parked, Relaxed, Acquire).is_ok() { + // Loop to guard against spurious wakeups. + loop { + ___lwp_park60(0, 0, null_mut(), 0, hint, null()); + if self.state.load(Acquire) == NOTIFIED { + break; + } + } + } + } + + // At this point, the change to NOTIFIED has always been observed with acquire + // ordering, so we can just use a relaxed store here (instead of a swap). + self.state.store(EMPTY, Relaxed); + } + + // Does not actually need `unsafe` or `Pin`, but the pthread implementation does. + pub unsafe fn park_timeout(self: Pin<&Self>, dur: Duration) { + if self.state.load(Acquire) != NOTIFIED { + let parked = _lwp_self() as u64; + let hint = self.state.as_mut_ptr().cast(); + let mut timeout = timespec { + // Saturate so that the operation will definitely time out + // (even if it is after the heat death of the universe). + tv_sec: dur.as_secs().try_into().ok().unwrap_or(time_t::MAX), + tv_nsec: dur.subsec_nanos().into(), + }; + + if self.state.compare_exchange(EMPTY, parked, Relaxed, Acquire).is_ok() { + // Timeout needs to be mutable since it is modified on NetBSD 9.0 and + // above. + ___lwp_park60(CLOCK_MONOTONIC, 0, &mut timeout, 0, hint, null()); + // Use a swap to get acquire ordering even if the token was set after + // the timeout occurred. + self.state.swap(EMPTY, Acquire); + return; + } + } + + self.state.store(EMPTY, Relaxed); + } + + // Does not actually need `Pin`, but the pthread implementation does. + pub fn unpark(self: Pin<&Self>) { + let state = self.state.swap(NOTIFIED, Release); + if !matches!(state, EMPTY | NOTIFIED) { + let lwp = state as lwpid_t; + let hint = self.state.as_mut_ptr().cast(); + + // If the parking thread terminated and did not actually park, this will + // probably return an error, which is OK. In the worst case, another + // thread has received the same LWP id. It will then receive a spurious + // wakeup, but those are allowable per the API contract. The same reasoning + // applies if a timeout occurred before this call, but the state was not + // yet reset. + + // SAFETY: + // The syscall has no invariants to hold. Only unsafe because it is an + // extern function. + unsafe { + _lwp_unpark(lwp, hint); + } + } + } +} diff --git a/library/std/src/sys/unix/thread_parker.rs b/library/std/src/sys/unix/thread_parker/pthread.rs index ca1a7138f..3dfc0026e 100644 --- a/library/std/src/sys/unix/thread_parker.rs +++ b/library/std/src/sys/unix/thread_parker/pthread.rs @@ -1,15 +1,5 @@ //! Thread parking without `futex` using the `pthread` synchronization primitives. -#![cfg(not(any( - target_os = "linux", - target_os = "android", - all(target_os = "emscripten", target_feature = "atomics"), - target_os = "freebsd", - target_os = "openbsd", - target_os = "dragonfly", - target_os = "fuchsia", -)))] - use crate::cell::UnsafeCell; use crate::marker::PhantomPinned; use crate::pin::Pin; @@ -59,8 +49,8 @@ unsafe fn wait_timeout( target_os = "espidf" ))] let (now, dur) = { - use super::time::SystemTime; use crate::cmp::min; + use crate::sys::time::SystemTime; // OSX implementation of `pthread_cond_timedwait` is buggy // with super long durations. When duration is greater than @@ -85,7 +75,7 @@ unsafe fn wait_timeout( target_os = "espidf" )))] let (now, dur) = { - use super::time::Timespec; + use crate::sys::time::Timespec; (Timespec::now(libc::CLOCK_MONOTONIC), dur) }; diff --git a/library/std/src/sys/unsupported/alloc.rs b/library/std/src/sys/unsupported/alloc.rs index 8d5d0a2f5..d715ae454 100644 --- a/library/std/src/sys/unsupported/alloc.rs +++ b/library/std/src/sys/unsupported/alloc.rs @@ -1,15 +1,16 @@ use crate::alloc::{GlobalAlloc, Layout, System}; +use crate::ptr::null_mut; #[stable(feature = "alloc_system_type", since = "1.28.0")] unsafe impl GlobalAlloc for System { #[inline] unsafe fn alloc(&self, _layout: Layout) -> *mut u8 { - 0 as *mut u8 + null_mut() } #[inline] unsafe fn alloc_zeroed(&self, _layout: Layout) -> *mut u8 { - 0 as *mut u8 + null_mut() } #[inline] @@ -17,6 +18,6 @@ unsafe impl GlobalAlloc for System { #[inline] unsafe fn realloc(&self, _ptr: *mut u8, _layout: Layout, _new_size: usize) -> *mut u8 { - 0 as *mut u8 + null_mut() } } diff --git a/library/std/src/sys/unsupported/common.rs b/library/std/src/sys/unsupported/common.rs index 4c9ade4a8..5cd9e57de 100644 --- a/library/std/src/sys/unsupported/common.rs +++ b/library/std/src/sys/unsupported/common.rs @@ -6,7 +6,7 @@ pub mod memchr { // SAFETY: must be called only once during runtime initialization. // NOTE: this is not guaranteed to run, for example when Rust code is called externally. -pub unsafe fn init(_argc: isize, _argv: *const *const u8) {} +pub unsafe fn init(_argc: isize, _argv: *const *const u8, _sigpipe: u8) {} // SAFETY: must be called only once during runtime cleanup. // NOTE: this is not guaranteed to run, for example when the program aborts. diff --git a/library/std/src/sys/unsupported/fs.rs b/library/std/src/sys/unsupported/fs.rs index 0e1a6257e..6ac1b5d2b 100644 --- a/library/std/src/sys/unsupported/fs.rs +++ b/library/std/src/sys/unsupported/fs.rs @@ -1,7 +1,7 @@ use crate::ffi::OsString; use crate::fmt; use crate::hash::{Hash, Hasher}; -use crate::io::{self, IoSlice, IoSliceMut, ReadBuf, SeekFrom}; +use crate::io::{self, BorrowedCursor, IoSlice, IoSliceMut, SeekFrom}; use crate::path::{Path, PathBuf}; use crate::sys::time::SystemTime; use crate::sys::unsupported; @@ -214,7 +214,7 @@ impl File { self.0 } - pub fn read_buf(&self, _buf: &mut ReadBuf<'_>) -> io::Result<()> { + pub fn read_buf(&self, _cursor: BorrowedCursor<'_>) -> io::Result<()> { self.0 } diff --git a/library/std/src/sys/unsupported/locks/mutex.rs b/library/std/src/sys/unsupported/locks/mutex.rs index d7cb12e0c..2be0b34b9 100644 --- a/library/std/src/sys/unsupported/locks/mutex.rs +++ b/library/std/src/sys/unsupported/locks/mutex.rs @@ -17,9 +17,6 @@ impl Mutex { } #[inline] - pub unsafe fn init(&mut self) {} - - #[inline] pub unsafe fn lock(&self) { assert_eq!(self.locked.replace(true), false, "cannot recursively acquire mutex"); } diff --git a/library/std/src/sys/unsupported/process.rs b/library/std/src/sys/unsupported/process.rs index 42a1ff730..633f17c05 100644 --- a/library/std/src/sys/unsupported/process.rs +++ b/library/std/src/sys/unsupported/process.rs @@ -200,6 +200,9 @@ impl<'a> Iterator for CommandArgs<'a> { fn next(&mut self) -> Option<&'a OsStr> { None } + fn size_hint(&self) -> (usize, Option<usize>) { + (0, Some(0)) + } } impl<'a> ExactSizeIterator for CommandArgs<'a> {} diff --git a/library/std/src/sys/wasi/fs.rs b/library/std/src/sys/wasi/fs.rs index 6614ae397..510cf36b1 100644 --- a/library/std/src/sys/wasi/fs.rs +++ b/library/std/src/sys/wasi/fs.rs @@ -3,7 +3,7 @@ use super::fd::WasiFd; use crate::ffi::{CStr, CString, OsStr, OsString}; use crate::fmt; -use crate::io::{self, IoSlice, IoSliceMut, ReadBuf, SeekFrom}; +use crate::io::{self, BorrowedCursor, IoSlice, IoSliceMut, SeekFrom}; use crate::iter; use crate::mem::{self, ManuallyDrop}; use crate::os::raw::c_int; @@ -439,8 +439,8 @@ impl File { true } - pub fn read_buf(&self, buf: &mut ReadBuf<'_>) -> io::Result<()> { - crate::io::default_read_buf(|buf| self.read(buf), buf) + pub fn read_buf(&self, cursor: BorrowedCursor<'_>) -> io::Result<()> { + crate::io::default_read_buf(|buf| self.read(buf), cursor) } pub fn write(&self, buf: &[u8]) -> io::Result<usize> { diff --git a/library/std/src/sys/wasi/stdio.rs b/library/std/src/sys/wasi/stdio.rs index 4cc0e4ed5..d2081771b 100644 --- a/library/std/src/sys/wasi/stdio.rs +++ b/library/std/src/sys/wasi/stdio.rs @@ -4,7 +4,7 @@ use super::fd::WasiFd; use crate::io::{self, IoSlice, IoSliceMut}; use crate::mem::ManuallyDrop; use crate::os::raw; -use crate::os::wasi::io::{AsRawFd, FromRawFd}; +use crate::os::wasi::io::{AsFd, AsRawFd, BorrowedFd, FromRawFd}; pub struct Stdin; pub struct Stdout; @@ -23,6 +23,13 @@ impl AsRawFd for Stdin { } } +impl AsFd for Stdin { + #[inline] + fn as_fd(&self) -> BorrowedFd<'_> { + unsafe { BorrowedFd::borrow_raw(0) } + } +} + impl io::Read for Stdin { fn read(&mut self, data: &mut [u8]) -> io::Result<usize> { self.read_vectored(&mut [IoSliceMut::new(data)]) @@ -51,6 +58,13 @@ impl AsRawFd for Stdout { } } +impl AsFd for Stdout { + #[inline] + fn as_fd(&self) -> BorrowedFd<'_> { + unsafe { BorrowedFd::borrow_raw(1) } + } +} + impl io::Write for Stdout { fn write(&mut self, data: &[u8]) -> io::Result<usize> { self.write_vectored(&[IoSlice::new(data)]) @@ -82,6 +96,13 @@ impl AsRawFd for Stderr { } } +impl AsFd for Stderr { + #[inline] + fn as_fd(&self) -> BorrowedFd<'_> { + unsafe { BorrowedFd::borrow_raw(2) } + } +} + impl io::Write for Stderr { fn write(&mut self, data: &[u8]) -> io::Result<usize> { self.write_vectored(&[IoSlice::new(data)]) diff --git a/library/std/src/sys/windows/alloc.rs b/library/std/src/sys/windows/alloc.rs index fdc81cdea..d53ea1600 100644 --- a/library/std/src/sys/windows/alloc.rs +++ b/library/std/src/sys/windows/alloc.rs @@ -16,6 +16,7 @@ mod tests; // Flag to indicate that the memory returned by `HeapAlloc` should be zeroed. const HEAP_ZERO_MEMORY: c::DWORD = 0x00000008; +#[link(name = "kernel32")] extern "system" { // Get a handle to the default heap of the current process, or null if the operation fails. // @@ -168,7 +169,7 @@ unsafe fn allocate(layout: Layout, zeroed: bool) -> *mut u8 { // SAFETY: Because the size and alignment of a header is <= `MIN_ALIGN` and `aligned` // is aligned to at least `MIN_ALIGN` and has at least `MIN_ALIGN` bytes of padding before // it, it is safe to write a header directly before it. - unsafe { ptr::write((aligned as *mut Header).offset(-1), Header(ptr)) }; + unsafe { ptr::write((aligned as *mut Header).sub(1), Header(ptr)) }; // SAFETY: The returned pointer does not point to the to the start of an allocated block, // but there is a header readable directly before it containing the location of the start @@ -213,7 +214,7 @@ unsafe impl GlobalAlloc for System { // SAFETY: Because of the contract of `System`, `ptr` is guaranteed to be non-null // and have a header readable directly before it. - unsafe { ptr::read((ptr as *mut Header).offset(-1)).0 } + unsafe { ptr::read((ptr as *mut Header).sub(1)).0 } } }; diff --git a/library/std/src/sys/windows/c.rs b/library/std/src/sys/windows/c.rs index 478068c73..89d0ab59b 100644 --- a/library/std/src/sys/windows/c.rs +++ b/library/std/src/sys/windows/c.rs @@ -66,6 +66,7 @@ pub type LPSYSTEM_INFO = *mut SYSTEM_INFO; pub type LPWSABUF = *mut WSABUF; pub type LPWSAOVERLAPPED = *mut c_void; pub type LPWSAOVERLAPPED_COMPLETION_ROUTINE = *mut c_void; +pub type BCRYPT_ALG_HANDLE = LPVOID; pub type PCONDITION_VARIABLE = *mut CONDITION_VARIABLE; pub type PLARGE_INTEGER = *mut c_longlong; @@ -278,6 +279,7 @@ pub const STATUS_INVALID_PARAMETER: NTSTATUS = 0xc000000d_u32 as _; pub const STATUS_PENDING: NTSTATUS = 0x103 as _; pub const STATUS_END_OF_FILE: NTSTATUS = 0xC0000011_u32 as _; pub const STATUS_NOT_IMPLEMENTED: NTSTATUS = 0xC0000002_u32 as _; +pub const STATUS_NOT_SUPPORTED: NTSTATUS = 0xC00000BB_u32 as _; // Equivalent to the `NT_SUCCESS` C preprocessor macro. // See: https://docs.microsoft.com/en-us/windows-hardware/drivers/kernel/using-ntstatus-values @@ -285,7 +287,8 @@ pub fn nt_success(status: NTSTATUS) -> bool { status >= 0 } -pub const BCRYPT_USE_SYSTEM_PREFERRED_RNG: DWORD = 0x00000002; +// "RNG\0" +pub const BCRYPT_RNG_ALGORITHM: &[u16] = &[b'R' as u16, b'N' as u16, b'G' as u16, 0]; #[repr(C)] pub struct UNICODE_STRING { @@ -455,6 +458,12 @@ pub enum FILE_INFO_BY_HANDLE_CLASS { } #[repr(C)] +pub struct FILE_ATTRIBUTE_TAG_INFO { + pub FileAttributes: DWORD, + pub ReparseTag: DWORD, +} + +#[repr(C)] pub struct FILE_DISPOSITION_INFO { pub DeleteFile: BOOLEAN, } @@ -501,6 +510,8 @@ pub struct FILE_END_OF_FILE_INFO { pub EndOfFile: LARGE_INTEGER, } +/// NB: Use carefully! In general using this as a reference is likely to get the +/// provenance wrong for the `rest` field! #[repr(C)] pub struct REPARSE_DATA_BUFFER { pub ReparseTag: c_uint, @@ -509,6 +520,8 @@ pub struct REPARSE_DATA_BUFFER { pub rest: (), } +/// NB: Use carefully! In general using this as a reference is likely to get the +/// provenance wrong for the `PathBuffer` field! #[repr(C)] pub struct SYMBOLIC_LINK_REPARSE_BUFFER { pub SubstituteNameOffset: c_ushort, @@ -519,6 +532,8 @@ pub struct SYMBOLIC_LINK_REPARSE_BUFFER { pub PathBuffer: WCHAR, } +/// NB: Use carefully! In general using this as a reference is likely to get the +/// provenance wrong for the `PathBuffer` field! #[repr(C)] pub struct MOUNT_POINT_REPARSE_BUFFER { pub SubstituteNameOffset: c_ushort, @@ -1217,11 +1232,18 @@ extern "system" { // >= Vista / Server 2008 // https://docs.microsoft.com/en-us/windows/win32/api/bcrypt/nf-bcrypt-bcryptgenrandom pub fn BCryptGenRandom( - hAlgorithm: LPVOID, + hAlgorithm: BCRYPT_ALG_HANDLE, pBuffer: *mut u8, cbBuffer: ULONG, dwFlags: ULONG, ) -> NTSTATUS; + pub fn BCryptOpenAlgorithmProvider( + phalgorithm: *mut BCRYPT_ALG_HANDLE, + pszAlgId: LPCWSTR, + pszimplementation: LPCWSTR, + dwflags: ULONG, + ) -> NTSTATUS; + pub fn BCryptCloseAlgorithmProvider(hAlgorithm: BCRYPT_ALG_HANDLE, dwFlags: ULONG) -> NTSTATUS; } // Functions that aren't available on every version of Windows that we support, @@ -1251,17 +1273,14 @@ compat_fn_with_fallback! { } compat_fn_optional! { - pub static SYNCH_API: &CStr = ansi_str!("api-ms-win-core-synch-l1-2-0"); - - // >= Windows 8 / Server 2012 - // https://docs.microsoft.com/en-us/windows/win32/api/synchapi/nf-synchapi-waitonaddress + crate::sys::compat::load_synch_functions(); pub fn WaitOnAddress( Address: LPVOID, CompareAddress: LPVOID, AddressSize: SIZE_T, dwMilliseconds: DWORD - ) -> BOOL; - pub fn WakeByAddressSingle(Address: LPVOID) -> (); + ); + pub fn WakeByAddressSingle(Address: LPVOID); } compat_fn_with_fallback! { diff --git a/library/std/src/sys/windows/cmath.rs b/library/std/src/sys/windows/cmath.rs index 1a5421fac..43ab8c7ee 100644 --- a/library/std/src/sys/windows/cmath.rs +++ b/library/std/src/sys/windows/cmath.rs @@ -44,7 +44,7 @@ mod shims { } // On 32-bit x86 MSVC these functions aren't defined, so we just define shims -// which promote everything fo f64, perform the calculation, and then demote +// which promote everything to f64, perform the calculation, and then demote // back to f32. While not precisely correct should be "correct enough" for now. #[cfg(all(target_env = "msvc", target_arch = "x86"))] mod shims { diff --git a/library/std/src/sys/windows/compat.rs b/library/std/src/sys/windows/compat.rs index ccc90177a..7dff81ecb 100644 --- a/library/std/src/sys/windows/compat.rs +++ b/library/std/src/sys/windows/compat.rs @@ -7,52 +7,66 @@ //! `GetModuleHandle` and `GetProcAddress` to look up DLL entry points at //! runtime. //! -//! This implementation uses a static initializer to look up the DLL entry -//! points. The CRT (C runtime) executes static initializers before `main` -//! is called (for binaries) and before `DllMain` is called (for DLLs). -//! This is the ideal time to look up DLL imports, because we are guaranteed -//! that no other threads will attempt to call these entry points. Thus, -//! we can look up the imports and store them in `static mut` fields -//! without any synchronization. +//! This is implemented simply by storing a function pointer in an atomic. +//! Loading and calling this function will have little or no overhead +//! compared with calling any other dynamically imported function. //! -//! This has an additional advantage: Because the DLL import lookup happens -//! at module initialization, the cost of these lookups is deterministic, -//! and is removed from the code paths that actually call the DLL imports. -//! That is, there is no unpredictable "cache miss" that occurs when calling -//! a DLL import. For applications that benefit from predictable delays, -//! this is a benefit. This also eliminates the comparison-and-branch -//! from the hot path. -//! -//! Currently, the standard library uses only a small number of dynamic -//! DLL imports. If this number grows substantially, then the cost of -//! performing all of the lookups at initialization time might become -//! substantial. -//! -//! The mechanism of registering a static initializer with the CRT is -//! documented in -//! [CRT Initialization](https://docs.microsoft.com/en-us/cpp/c-runtime-library/crt-initialization?view=msvc-160). -//! It works by contributing a global symbol to the `.CRT$XCU` section. -//! The linker builds a table of all static initializer functions. -//! The CRT startup code then iterates that table, calling each -//! initializer function. -//! -//! # **WARNING!!* -//! The environment that a static initializer function runs in is highly -//! constrained. There are **many** restrictions on what static initializers -//! can safely do. Static initializer functions **MUST NOT** do any of the -//! following (this list is not comprehensive): -//! * touch any other static field that is used by a different static -//! initializer, because the order that static initializers run in -//! is not defined. -//! * call `LoadLibrary` or any other function that acquires the DLL -//! loader lock. -//! * call any Rust function or CRT function that touches any static -//! (global) state. +//! The stored function pointer starts out as an importer function which will +//! swap itself with the real function when it's called for the first time. If +//! the real function can't be imported then a fallback function is used in its +//! place. While this is low cost for the happy path (where the function is +//! already loaded) it does mean there's some overhead the first time the +//! function is called. In the worst case, multiple threads may all end up +//! importing the same function unnecessarily. use crate::ffi::{c_void, CStr}; use crate::ptr::NonNull; +use crate::sync::atomic::Ordering; use crate::sys::c; +// This uses a static initializer to preload some imported functions. +// The CRT (C runtime) executes static initializers before `main` +// is called (for binaries) and before `DllMain` is called (for DLLs). +// +// It works by contributing a global symbol to the `.CRT$XCT` section. +// The linker builds a table of all static initializer functions. +// The CRT startup code then iterates that table, calling each +// initializer function. +// +// NOTE: User code should instead use .CRT$XCU to reliably run after std's initializer. +// If you're reading this and would like a guarantee here, please +// file an issue for discussion; currently we don't guarantee any functionality +// before main. +// See https://docs.microsoft.com/en-us/cpp/c-runtime-library/crt-initialization?view=msvc-170 +#[used] +#[link_section = ".CRT$XCT"] +static INIT_TABLE_ENTRY: unsafe extern "C" fn() = init; + +/// Preload some imported functions. +/// +/// Note that any functions included here will be unconditionally loaded in +/// the final binary, regardless of whether or not they're actually used. +/// +/// Therefore, this should be limited to `compat_fn_optional` functions which +/// must be preloaded or any functions where lazier loading demonstrates a +/// negative performance impact in practical situations. +/// +/// Currently we only preload `WaitOnAddress` and `WakeByAddressSingle`. +unsafe extern "C" fn init() { + // In an exe this code is executed before main() so is single threaded. + // In a DLL the system's loader lock will be held thereby synchronizing + // access. So the same best practices apply here as they do to running in DllMain: + // https://docs.microsoft.com/en-us/windows/win32/dlls/dynamic-link-library-best-practices + // + // DO NOT do anything interesting or complicated in this function! DO NOT call + // any Rust functions or CRT functions if those functions touch any global state, + // because this function runs during global initialization. For example, DO NOT + // do any dynamic allocation, don't call LoadLibrary, etc. + + // Attempt to preload the synch functions. + load_synch_functions(); +} + /// Helper macro for creating CStrs from literals and symbol names. macro_rules! ansi_str { (sym $ident:ident) => {{ @@ -85,39 +99,6 @@ pub(crate) const fn const_cstr_from_bytes(bytes: &'static [u8]) -> &'static CStr unsafe { crate::ffi::CStr::from_bytes_with_nul_unchecked(bytes) } } -#[used] -#[link_section = ".CRT$XCU"] -static INIT_TABLE_ENTRY: unsafe extern "C" fn() = init; - -/// This is where the magic preloading of symbols happens. -/// -/// Note that any functions included here will be unconditionally included in -/// the final binary, regardless of whether or not they're actually used. -/// -/// Therefore, this is limited to `compat_fn_optional` functions which must be -/// preloaded and any functions which may be more time sensitive, even for the first call. -unsafe extern "C" fn init() { - // There is no locking here. This code is executed before main() is entered, and - // is guaranteed to be single-threaded. - // - // DO NOT do anything interesting or complicated in this function! DO NOT call - // any Rust functions or CRT functions if those functions touch any global state, - // because this function runs during global initialization. For example, DO NOT - // do any dynamic allocation, don't call LoadLibrary, etc. - - if let Some(synch) = Module::new(c::SYNCH_API) { - // These are optional and so we must manually attempt to load them - // before they can be used. - c::WaitOnAddress::preload(synch); - c::WakeByAddressSingle::preload(synch); - } - - if let Some(kernel32) = Module::new(c::KERNEL32) { - // Preloading this means getting a precise time will be as fast as possible. - c::GetSystemTimePreciseAsFileTime::preload(kernel32); - } -} - /// Represents a loaded module. /// /// Note that the modules std depends on must not be unloaded. @@ -151,7 +132,7 @@ impl Module { macro_rules! compat_fn_with_fallback { (pub static $module:ident: &CStr = $name:expr; $( $(#[$meta:meta])* - pub fn $symbol:ident($($argname:ident: $argtype:ty),*) -> $rettype:ty $fallback_body:block + $vis:vis fn $symbol:ident($($argname:ident: $argtype:ty),*) -> $rettype:ty $fallback_body:block )*) => ( pub static $module: &CStr = $name; $( @@ -196,11 +177,6 @@ macro_rules! compat_fn_with_fallback { $fallback_body } - #[allow(unused)] - pub(in crate::sys) fn preload(module: Module) { - load_from_module(Some(module)); - } - #[inline(always)] pub unsafe fn call($($argname: $argtype),*) -> $rettype { let func: F = mem::transmute(PTR.load(Ordering::Relaxed)); @@ -208,66 +184,60 @@ macro_rules! compat_fn_with_fallback { } } $(#[$meta])* - pub use $symbol::call as $symbol; + $vis use $symbol::call as $symbol; )*) } -/// A function that either exists or doesn't. +/// Optionally loaded functions. /// -/// NOTE: Optional functions must be preloaded in the `init` function above, or they will always be None. +/// Actual loading of the function defers to $load_functions. macro_rules! compat_fn_optional { - (pub static $module:ident: &CStr = $name:expr; $( - $(#[$meta:meta])* - pub fn $symbol:ident($($argname:ident: $argtype:ty),*) -> $rettype:ty; - )*) => ( - pub static $module: &CStr = $name; + ($load_functions:expr; $( - $(#[$meta])* - pub mod $symbol { - #[allow(unused_imports)] - use super::*; - use crate::mem; - use crate::sync::atomic::{AtomicPtr, Ordering}; - use crate::sys::compat::Module; - use crate::ptr::{self, NonNull}; - - type F = unsafe extern "system" fn($($argtype),*) -> $rettype; - - /// `PTR` will either be `null()` or set to the loaded function. - static PTR: AtomicPtr<c_void> = AtomicPtr::new(ptr::null_mut()); - - /// Only allow access to the function if it has loaded successfully. - #[inline(always)] - #[cfg(not(miri))] - pub fn option() -> Option<F> { - unsafe { - NonNull::new(PTR.load(Ordering::Relaxed)).map(|f| mem::transmute(f)) + $(#[$meta:meta])* + $vis:vis fn $symbol:ident($($argname:ident: $argtype:ty),*) $(-> $rettype:ty)?; + )+) => ( + $( + pub mod $symbol { + use super::*; + use crate::ffi::c_void; + use crate::mem; + use crate::ptr::{self, NonNull}; + use crate::sync::atomic::{AtomicPtr, Ordering}; + + pub(in crate::sys) static PTR: AtomicPtr<c_void> = AtomicPtr::new(ptr::null_mut()); + + type F = unsafe extern "system" fn($($argtype),*) $(-> $rettype)?; + + #[inline(always)] + pub fn option() -> Option<F> { + // Miri does not understand the way we do preloading + // therefore load the function here instead. + #[cfg(miri)] $load_functions; + NonNull::new(PTR.load(Ordering::Relaxed)).map(|f| unsafe { mem::transmute(f) }) } } + )+ + ) +} - // Miri does not understand the way we do preloading - // therefore load the function here instead. - #[cfg(miri)] - pub fn option() -> Option<F> { - let mut func = NonNull::new(PTR.load(Ordering::Relaxed)); - if func.is_none() { - unsafe { Module::new($module).map(preload) }; - func = NonNull::new(PTR.load(Ordering::Relaxed)); - } - unsafe { - func.map(|f| mem::transmute(f)) - } - } +/// Load all needed functions from "api-ms-win-core-synch-l1-2-0". +pub(super) fn load_synch_functions() { + fn try_load() -> Option<()> { + const MODULE_NAME: &CStr = ansi_str!("api-ms-win-core-synch-l1-2-0"); + const WAIT_ON_ADDRESS: &CStr = ansi_str!("WaitOnAddress"); + const WAKE_BY_ADDRESS_SINGLE: &CStr = ansi_str!("WakeByAddressSingle"); + + // Try loading the library and all the required functions. + // If any step fails, then they all fail. + let library = unsafe { Module::new(MODULE_NAME) }?; + let wait_on_address = library.proc_address(WAIT_ON_ADDRESS)?; + let wake_by_address_single = library.proc_address(WAKE_BY_ADDRESS_SINGLE)?; + + c::WaitOnAddress::PTR.store(wait_on_address.as_ptr(), Ordering::Relaxed); + c::WakeByAddressSingle::PTR.store(wake_by_address_single.as_ptr(), Ordering::Relaxed); + Some(()) + } - #[allow(unused)] - pub(in crate::sys) fn preload(module: Module) { - unsafe { - static SYMBOL_NAME: &CStr = ansi_str!(sym $symbol); - if let Some(f) = module.proc_address(SYMBOL_NAME) { - PTR.store(f.as_ptr(), Ordering::Relaxed); - } - } - } - } - )*) + try_load(); } diff --git a/library/std/src/sys/windows/fs.rs b/library/std/src/sys/windows/fs.rs index aed082b3e..155d0297e 100644 --- a/library/std/src/sys/windows/fs.rs +++ b/library/std/src/sys/windows/fs.rs @@ -2,8 +2,8 @@ use crate::os::windows::prelude::*; use crate::ffi::OsString; use crate::fmt; -use crate::io::{self, Error, IoSlice, IoSliceMut, ReadBuf, SeekFrom}; -use crate::mem; +use crate::io::{self, BorrowedCursor, Error, IoSlice, IoSliceMut, SeekFrom}; +use crate::mem::{self, MaybeUninit}; use crate::os::windows::io::{AsHandle, BorrowedHandle}; use crate::path::{Path, PathBuf}; use crate::ptr; @@ -11,7 +11,7 @@ use crate::slice; use crate::sync::Arc; use crate::sys::handle::Handle; use crate::sys::time::SystemTime; -use crate::sys::{c, cvt}; +use crate::sys::{c, cvt, Align8}; use crate::sys_common::{AsInner, FromInner, IntoInner}; use crate::thread; @@ -326,9 +326,15 @@ impl File { cvt(c::GetFileInformationByHandle(self.handle.as_raw_handle(), &mut info))?; let mut reparse_tag = 0; if info.dwFileAttributes & c::FILE_ATTRIBUTE_REPARSE_POINT != 0 { - let mut b = [0; c::MAXIMUM_REPARSE_DATA_BUFFER_SIZE]; - if let Ok((_, buf)) = self.reparse_point(&mut b) { - reparse_tag = buf.ReparseTag; + let mut attr_tag: c::FILE_ATTRIBUTE_TAG_INFO = mem::zeroed(); + cvt(c::GetFileInformationByHandleEx( + self.handle.as_raw_handle(), + c::FileAttributeTagInfo, + ptr::addr_of_mut!(attr_tag).cast(), + mem::size_of::<c::FILE_ATTRIBUTE_TAG_INFO>().try_into().unwrap(), + ))?; + if attr_tag.FileAttributes & c::FILE_ATTRIBUTE_REPARSE_POINT != 0 { + reparse_tag = attr_tag.ReparseTag; } } Ok(FileAttr { @@ -389,9 +395,15 @@ impl File { attr.file_size = info.AllocationSize as u64; attr.number_of_links = Some(info.NumberOfLinks); if attr.file_type().is_reparse_point() { - let mut b = [0; c::MAXIMUM_REPARSE_DATA_BUFFER_SIZE]; - if let Ok((_, buf)) = self.reparse_point(&mut b) { - attr.reparse_tag = buf.ReparseTag; + let mut attr_tag: c::FILE_ATTRIBUTE_TAG_INFO = mem::zeroed(); + cvt(c::GetFileInformationByHandleEx( + self.handle.as_raw_handle(), + c::FileAttributeTagInfo, + ptr::addr_of_mut!(attr_tag).cast(), + mem::size_of::<c::FILE_ATTRIBUTE_TAG_INFO>().try_into().unwrap(), + ))?; + if attr_tag.FileAttributes & c::FILE_ATTRIBUTE_REPARSE_POINT != 0 { + attr.reparse_tag = attr_tag.ReparseTag; } } Ok(attr) @@ -415,8 +427,8 @@ impl File { self.handle.read_at(buf, offset) } - pub fn read_buf(&self, buf: &mut ReadBuf<'_>) -> io::Result<()> { - self.handle.read_buf(buf) + pub fn read_buf(&self, cursor: BorrowedCursor<'_>) -> io::Result<()> { + self.handle.read_buf(cursor) } pub fn write(&self, buf: &[u8]) -> io::Result<usize> { @@ -458,38 +470,46 @@ impl File { Ok(Self { handle: self.handle.try_clone()? }) } - fn reparse_point<'a>( + // NB: returned pointer is derived from `space`, and has provenance to + // match. A raw pointer is returned rather than a reference in order to + // avoid narrowing provenance to the actual `REPARSE_DATA_BUFFER`. + fn reparse_point( &self, - space: &'a mut [u8; c::MAXIMUM_REPARSE_DATA_BUFFER_SIZE], - ) -> io::Result<(c::DWORD, &'a c::REPARSE_DATA_BUFFER)> { + space: &mut Align8<[MaybeUninit<u8>]>, + ) -> io::Result<(c::DWORD, *const c::REPARSE_DATA_BUFFER)> { unsafe { let mut bytes = 0; cvt({ + // Grab this in advance to avoid it invalidating the pointer + // we get from `space.0.as_mut_ptr()`. + let len = space.0.len(); c::DeviceIoControl( self.handle.as_raw_handle(), c::FSCTL_GET_REPARSE_POINT, ptr::null_mut(), 0, - space.as_mut_ptr() as *mut _, - space.len() as c::DWORD, + space.0.as_mut_ptr().cast(), + len as c::DWORD, &mut bytes, ptr::null_mut(), ) })?; - Ok((bytes, &*(space.as_ptr() as *const c::REPARSE_DATA_BUFFER))) + const _: () = assert!(core::mem::align_of::<c::REPARSE_DATA_BUFFER>() <= 8); + Ok((bytes, space.0.as_ptr().cast::<c::REPARSE_DATA_BUFFER>())) } } fn readlink(&self) -> io::Result<PathBuf> { - let mut space = [0u8; c::MAXIMUM_REPARSE_DATA_BUFFER_SIZE]; + let mut space = Align8([MaybeUninit::<u8>::uninit(); c::MAXIMUM_REPARSE_DATA_BUFFER_SIZE]); let (_bytes, buf) = self.reparse_point(&mut space)?; unsafe { - let (path_buffer, subst_off, subst_len, relative) = match buf.ReparseTag { + let (path_buffer, subst_off, subst_len, relative) = match (*buf).ReparseTag { c::IO_REPARSE_TAG_SYMLINK => { let info: *const c::SYMBOLIC_LINK_REPARSE_BUFFER = - &buf.rest as *const _ as *const _; + ptr::addr_of!((*buf).rest).cast(); + assert!(info.is_aligned()); ( - &(*info).PathBuffer as *const _ as *const u16, + ptr::addr_of!((*info).PathBuffer).cast::<u16>(), (*info).SubstituteNameOffset / 2, (*info).SubstituteNameLength / 2, (*info).Flags & c::SYMLINK_FLAG_RELATIVE != 0, @@ -497,9 +517,10 @@ impl File { } c::IO_REPARSE_TAG_MOUNT_POINT => { let info: *const c::MOUNT_POINT_REPARSE_BUFFER = - &buf.rest as *const _ as *const _; + ptr::addr_of!((*buf).rest).cast(); + assert!(info.is_aligned()); ( - &(*info).PathBuffer as *const _ as *const u16, + ptr::addr_of!((*info).PathBuffer).cast::<u16>(), (*info).SubstituteNameOffset / 2, (*info).SubstituteNameLength / 2, false, @@ -512,7 +533,7 @@ impl File { )); } }; - let subst_ptr = path_buffer.offset(subst_off as isize); + let subst_ptr = path_buffer.add(subst_off.into()); let mut subst = slice::from_raw_parts(subst_ptr, subst_len as usize); // Absolute paths start with an NT internal namespace prefix `\??\` // We should not let it leak through. @@ -649,27 +670,31 @@ impl File { /// A buffer for holding directory entries. struct DirBuff { - buffer: Vec<u8>, + buffer: Box<Align8<[MaybeUninit<u8>; Self::BUFFER_SIZE]>>, } impl DirBuff { + const BUFFER_SIZE: usize = 1024; fn new() -> Self { - const BUFFER_SIZE: usize = 1024; - Self { buffer: vec![0_u8; BUFFER_SIZE] } + Self { + // Safety: `Align8<[MaybeUninit<u8>; N]>` does not need + // initialization. + buffer: unsafe { Box::new_uninit().assume_init() }, + } } fn capacity(&self) -> usize { - self.buffer.len() + self.buffer.0.len() } fn as_mut_ptr(&mut self) -> *mut u8 { - self.buffer.as_mut_ptr().cast() + self.buffer.0.as_mut_ptr().cast() } /// Returns a `DirBuffIter`. fn iter(&self) -> DirBuffIter<'_> { DirBuffIter::new(self) } } -impl AsRef<[u8]> for DirBuff { - fn as_ref(&self) -> &[u8] { - &self.buffer +impl AsRef<[MaybeUninit<u8>]> for DirBuff { + fn as_ref(&self) -> &[MaybeUninit<u8>] { + &self.buffer.0 } } @@ -677,7 +702,7 @@ impl AsRef<[u8]> for DirBuff { /// /// Currently only returns file names (UTF-16 encoded). struct DirBuffIter<'a> { - buffer: Option<&'a [u8]>, + buffer: Option<&'a [MaybeUninit<u8>]>, cursor: usize, } impl<'a> DirBuffIter<'a> { @@ -692,14 +717,21 @@ impl<'a> Iterator for DirBuffIter<'a> { let buffer = &self.buffer?[self.cursor..]; // Get the name and next entry from the buffer. - // SAFETY: The buffer contains a `FILE_ID_BOTH_DIR_INFO` struct but the - // last field (the file name) is unsized. So an offset has to be - // used to get the file name slice. + // SAFETY: + // - The buffer contains a `FILE_ID_BOTH_DIR_INFO` struct but the last + // field (the file name) is unsized. So an offset has to be used to + // get the file name slice. + // - The OS has guaranteed initialization of the fields of + // `FILE_ID_BOTH_DIR_INFO` and the trailing filename (for at least + // `FileNameLength` bytes) let (name, is_directory, next_entry) = unsafe { let info = buffer.as_ptr().cast::<c::FILE_ID_BOTH_DIR_INFO>(); + // Guaranteed to be aligned in documentation for + // https://docs.microsoft.com/en-us/windows/win32/api/winbase/ns-winbase-file_id_both_dir_info + assert!(info.is_aligned()); let next_entry = (*info).NextEntryOffset as usize; let name = crate::slice::from_raw_parts( - (*info).FileName.as_ptr().cast::<u16>(), + ptr::addr_of!((*info).FileName).cast::<u16>(), (*info).FileNameLength as usize / size_of::<u16>(), ); let is_directory = ((*info).FileAttributes & c::FILE_ATTRIBUTE_DIRECTORY) != 0; @@ -1337,18 +1369,19 @@ fn symlink_junction_inner(original: &Path, junction: &Path) -> io::Result<()> { let h = f.as_inner().as_raw_handle(); unsafe { - let mut data = [0u8; c::MAXIMUM_REPARSE_DATA_BUFFER_SIZE]; - let db = data.as_mut_ptr() as *mut c::REPARSE_MOUNTPOINT_DATA_BUFFER; - let buf = &mut (*db).ReparseTarget as *mut c::WCHAR; + let mut data = Align8([MaybeUninit::<u8>::uninit(); c::MAXIMUM_REPARSE_DATA_BUFFER_SIZE]); + let data_ptr = data.0.as_mut_ptr(); + let db = data_ptr.cast::<c::REPARSE_MOUNTPOINT_DATA_BUFFER>(); + let buf = ptr::addr_of_mut!((*db).ReparseTarget).cast::<c::WCHAR>(); let mut i = 0; // FIXME: this conversion is very hacky let v = br"\??\"; let v = v.iter().map(|x| *x as u16); for c in v.chain(original.as_os_str().encode_wide()) { - *buf.offset(i) = c; + *buf.add(i) = c; i += 1; } - *buf.offset(i) = 0; + *buf.add(i) = 0; i += 1; (*db).ReparseTag = c::IO_REPARSE_TAG_MOUNT_POINT; (*db).ReparseTargetMaximumLength = (i * 2) as c::WORD; @@ -1359,7 +1392,7 @@ fn symlink_junction_inner(original: &Path, junction: &Path) -> io::Result<()> { cvt(c::DeviceIoControl( h as *mut _, c::FSCTL_SET_REPARSE_POINT, - data.as_ptr() as *mut _, + data_ptr.cast(), (*db).ReparseDataLength + 8, ptr::null_mut(), 0, diff --git a/library/std/src/sys/windows/handle.rs b/library/std/src/sys/windows/handle.rs index e24b09cc9..ae33d48c6 100644 --- a/library/std/src/sys/windows/handle.rs +++ b/library/std/src/sys/windows/handle.rs @@ -4,7 +4,7 @@ mod tests; use crate::cmp; -use crate::io::{self, ErrorKind, IoSlice, IoSliceMut, Read, ReadBuf}; +use crate::io::{self, BorrowedCursor, ErrorKind, IoSlice, IoSliceMut, Read}; use crate::mem; use crate::os::windows::io::{ AsHandle, AsRawHandle, BorrowedHandle, FromRawHandle, IntoRawHandle, OwnedHandle, RawHandle, @@ -112,18 +112,16 @@ impl Handle { } } - pub fn read_buf(&self, buf: &mut ReadBuf<'_>) -> io::Result<()> { - let res = unsafe { - self.synchronous_read(buf.unfilled_mut().as_mut_ptr(), buf.remaining(), None) - }; + pub fn read_buf(&self, mut cursor: BorrowedCursor<'_>) -> io::Result<()> { + let res = + unsafe { self.synchronous_read(cursor.as_mut().as_mut_ptr(), cursor.capacity(), None) }; match res { Ok(read) => { // Safety: `read` bytes were written to the initialized portion of the buffer unsafe { - buf.assume_init(read as usize); + cursor.advance(read as usize); } - buf.add_filled(read as usize); Ok(()) } diff --git a/library/std/src/sys/windows/locks/mutex.rs b/library/std/src/sys/windows/locks/mutex.rs index f91e8f9f5..91207f5f4 100644 --- a/library/std/src/sys/windows/locks/mutex.rs +++ b/library/std/src/sys/windows/locks/mutex.rs @@ -37,8 +37,6 @@ impl Mutex { pub const fn new() -> Mutex { Mutex { srwlock: UnsafeCell::new(c::SRWLOCK_INIT) } } - #[inline] - pub unsafe fn init(&mut self) {} #[inline] pub unsafe fn lock(&self) { diff --git a/library/std/src/sys/windows/mod.rs b/library/std/src/sys/windows/mod.rs index b3f6d2d0a..eab9b9612 100644 --- a/library/std/src/sys/windows/mod.rs +++ b/library/std/src/sys/windows/mod.rs @@ -2,6 +2,7 @@ use crate::ffi::{CStr, OsStr, OsString}; use crate::io::ErrorKind; +use crate::mem::MaybeUninit; use crate::os::windows::ffi::{OsStrExt, OsStringExt}; use crate::path::PathBuf; use crate::time::Duration; @@ -47,7 +48,7 @@ cfg_if::cfg_if! { // SAFETY: must be called only once during runtime initialization. // NOTE: this is not guaranteed to run, for example when Rust code is called externally. -pub unsafe fn init(_argc: isize, _argv: *const *const u8) { +pub unsafe fn init(_argc: isize, _argv: *const *const u8, _sigpipe: u8) { stack_overflow::init(); // Normally, `thread::spawn` will call `Thread::set_name` but since this thread already @@ -204,8 +205,8 @@ where // This initial size also works around `GetFullPathNameW` returning // incorrect size hints for some short paths: // https://github.com/dylni/normpath/issues/5 - let mut stack_buf = [0u16; 512]; - let mut heap_buf = Vec::new(); + let mut stack_buf: [MaybeUninit<u16>; 512] = MaybeUninit::uninit_array(); + let mut heap_buf: Vec<MaybeUninit<u16>> = Vec::new(); unsafe { let mut n = stack_buf.len(); loop { @@ -214,6 +215,11 @@ where } else { let extra = n - heap_buf.len(); heap_buf.reserve(extra); + // We used `reserve` and not `reserve_exact`, so in theory we + // may have gotten more than requested. If so, we'd like to use + // it... so long as we won't cause overflow. + n = heap_buf.capacity().min(c::DWORD::MAX as usize); + // Safety: MaybeUninit<u16> does not need initialization heap_buf.set_len(n); &mut heap_buf[..] }; @@ -228,13 +234,13 @@ where // error" is still 0 then we interpret it as a 0 length buffer and // not an actual error. c::SetLastError(0); - let k = match f1(buf.as_mut_ptr(), n as c::DWORD) { + let k = match f1(buf.as_mut_ptr().cast::<u16>(), n as c::DWORD) { 0 if c::GetLastError() == 0 => 0, 0 => return Err(crate::io::Error::last_os_error()), n => n, } as usize; if k == n && c::GetLastError() == c::ERROR_INSUFFICIENT_BUFFER { - n *= 2; + n = n.saturating_mul(2).min(c::DWORD::MAX as usize); } else if k > n { n = k; } else if k == n { @@ -244,7 +250,9 @@ where // Therefore k never equals n. unreachable!(); } else { - return Ok(f2(&buf[..k])); + // Safety: First `k` values are initialized. + let slice: &[u16] = MaybeUninit::slice_assume_init_ref(&buf[..k]); + return Ok(f2(slice)); } } } @@ -321,3 +329,11 @@ pub fn abort_internal() -> ! { } crate::intrinsics::abort(); } + +/// Align the inner value to 8 bytes. +/// +/// This is enough for almost all of the buffers we're likely to work with in +/// the Windows APIs we use. +#[repr(C, align(8))] +#[derive(Copy, Clone)] +pub(crate) struct Align8<T: ?Sized>(pub T); diff --git a/library/std/src/sys/windows/os.rs b/library/std/src/sys/windows/os.rs index bcac996c0..352337ba3 100644 --- a/library/std/src/sys/windows/os.rs +++ b/library/std/src/sys/windows/os.rs @@ -99,11 +99,11 @@ impl Iterator for Env { } let p = self.cur as *const u16; let mut len = 0; - while *p.offset(len) != 0 { + while *p.add(len) != 0 { len += 1; } - let s = slice::from_raw_parts(p, len as usize); - self.cur = self.cur.offset(len + 1); + let s = slice::from_raw_parts(p, len); + self.cur = self.cur.add(len + 1); // Windows allows environment variables to start with an equals // symbol (in any other position, this is the separator between diff --git a/library/std/src/sys/windows/os_str.rs b/library/std/src/sys/windows/os_str.rs index 11883f150..4bdd8c505 100644 --- a/library/std/src/sys/windows/os_str.rs +++ b/library/std/src/sys/windows/os_str.rs @@ -164,9 +164,7 @@ impl Slice { } pub fn to_owned(&self) -> Buf { - let mut buf = Wtf8Buf::with_capacity(self.inner.len()); - buf.push_wtf8(&self.inner); - Buf { inner: buf } + Buf { inner: self.inner.to_owned() } } pub fn clone_into(&self, buf: &mut Buf) { diff --git a/library/std/src/sys/windows/path/tests.rs b/library/std/src/sys/windows/path/tests.rs index 6eab38cab..623c62361 100644 --- a/library/std/src/sys/windows/path/tests.rs +++ b/library/std/src/sys/windows/path/tests.rs @@ -105,7 +105,7 @@ fn test_parse_prefix_verbatim_device() { assert_eq!(prefix, parse_prefix(r"\\?/C:\windows\system32\notepad.exe")); } -// See #93586 for more infomation. +// See #93586 for more information. #[test] fn test_windows_prefix_components() { use crate::path::Path; diff --git a/library/std/src/sys/windows/rand.rs b/library/std/src/sys/windows/rand.rs index f8fd93a73..d6cd8f802 100644 --- a/library/std/src/sys/windows/rand.rs +++ b/library/std/src/sys/windows/rand.rs @@ -1,35 +1,126 @@ -use crate::io; +//! # Random key generation +//! +//! This module wraps the RNG provided by the OS. There are a few different +//! ways to interface with the OS RNG so it's worth exploring each of the options. +//! Note that at the time of writing these all go through the (undocumented) +//! `bcryptPrimitives.dll` but they use different route to get there. +//! +//! Originally we were using [`RtlGenRandom`], however that function is +//! deprecated and warns it "may be altered or unavailable in subsequent versions". +//! +//! So we switched to [`BCryptGenRandom`] with the `BCRYPT_USE_SYSTEM_PREFERRED_RNG` +//! flag to query and find the system configured RNG. However, this change caused a small +//! but significant number of users to experience panics caused by a failure of +//! this function. See [#94098]. +//! +//! The current version changes this to use the `BCRYPT_RNG_ALG_HANDLE` +//! [Pseudo-handle], which gets the default RNG algorithm without querying the +//! system preference thus hopefully avoiding the previous issue. +//! This is only supported on Windows 10+ so a fallback is used for older versions. +//! +//! [#94098]: https://github.com/rust-lang/rust/issues/94098 +//! [`RtlGenRandom`]: https://docs.microsoft.com/en-us/windows/win32/api/ntsecapi/nf-ntsecapi-rtlgenrandom +//! [`BCryptGenRandom`]: https://docs.microsoft.com/en-us/windows/win32/api/bcrypt/nf-bcrypt-bcryptgenrandom +//! [Pseudo-handle]: https://docs.microsoft.com/en-us/windows/win32/seccng/cng-algorithm-pseudo-handles use crate::mem; use crate::ptr; use crate::sys::c; +/// Generates high quality secure random keys for use by [`HashMap`]. +/// +/// This is used to seed the default [`RandomState`]. +/// +/// [`HashMap`]: crate::collections::HashMap +/// [`RandomState`]: crate::collections::hash_map::RandomState pub fn hashmap_random_keys() -> (u64, u64) { - let mut v = (0, 0); - let ret = unsafe { - c::BCryptGenRandom( - ptr::null_mut(), - &mut v as *mut _ as *mut u8, - mem::size_of_val(&v) as c::ULONG, - c::BCRYPT_USE_SYSTEM_PREFERRED_RNG, - ) - }; - if ret != 0 { fallback_rng() } else { v } + Rng::open().and_then(|rng| rng.gen_random_keys()).unwrap_or_else(fallback_rng) +} + +struct Rng(c::BCRYPT_ALG_HANDLE); +impl Rng { + #[cfg(miri)] + fn open() -> Result<Self, c::NTSTATUS> { + const BCRYPT_RNG_ALG_HANDLE: c::BCRYPT_ALG_HANDLE = ptr::invalid_mut(0x81); + let _ = ( + c::BCryptOpenAlgorithmProvider, + c::BCryptCloseAlgorithmProvider, + c::BCRYPT_RNG_ALGORITHM, + c::STATUS_NOT_SUPPORTED, + ); + Ok(Self(BCRYPT_RNG_ALG_HANDLE)) + } + #[cfg(not(miri))] + // Open a handle to the RNG algorithm. + fn open() -> Result<Self, c::NTSTATUS> { + use crate::sync::atomic::AtomicPtr; + use crate::sync::atomic::Ordering::{Acquire, Release}; + const ERROR_VALUE: c::LPVOID = ptr::invalid_mut(usize::MAX); + + // An atomic is used so we don't need to reopen the handle every time. + static HANDLE: AtomicPtr<crate::ffi::c_void> = AtomicPtr::new(ptr::null_mut()); + + let mut handle = HANDLE.load(Acquire); + // We use a sentinel value to designate an error occurred last time. + if handle == ERROR_VALUE { + Err(c::STATUS_NOT_SUPPORTED) + } else if handle.is_null() { + let status = unsafe { + c::BCryptOpenAlgorithmProvider( + &mut handle, + c::BCRYPT_RNG_ALGORITHM.as_ptr(), + ptr::null(), + 0, + ) + }; + if c::nt_success(status) { + // If another thread opens a handle first then use that handle instead. + let result = HANDLE.compare_exchange(ptr::null_mut(), handle, Release, Acquire); + if let Err(previous_handle) = result { + // Close our handle and return the previous one. + unsafe { c::BCryptCloseAlgorithmProvider(handle, 0) }; + handle = previous_handle; + } + Ok(Self(handle)) + } else { + HANDLE.store(ERROR_VALUE, Release); + Err(status) + } + } else { + Ok(Self(handle)) + } + } + + fn gen_random_keys(self) -> Result<(u64, u64), c::NTSTATUS> { + let mut v = (0, 0); + let status = unsafe { + let size = mem::size_of_val(&v).try_into().unwrap(); + c::BCryptGenRandom(self.0, ptr::addr_of_mut!(v).cast(), size, 0) + }; + if c::nt_success(status) { Ok(v) } else { Err(status) } + } } /// Generate random numbers using the fallback RNG function (RtlGenRandom) #[cfg(not(target_vendor = "uwp"))] #[inline(never)] -fn fallback_rng() -> (u64, u64) { +fn fallback_rng(rng_status: c::NTSTATUS) -> (u64, u64) { let mut v = (0, 0); let ret = unsafe { c::RtlGenRandom(&mut v as *mut _ as *mut u8, mem::size_of_val(&v) as c::ULONG) }; - if ret != 0 { v } else { panic!("fallback RNG broken: {}", io::Error::last_os_error()) } + if ret != 0 { + v + } else { + panic!( + "RNG broken: {rng_status:#x}, fallback RNG broken: {}", + crate::io::Error::last_os_error() + ) + } } /// We can't use RtlGenRandom with UWP, so there is no fallback #[cfg(target_vendor = "uwp")] #[inline(never)] -fn fallback_rng() -> (u64, u64) { - panic!("fallback RNG broken: RtlGenRandom() not supported on UWP"); +fn fallback_rng(rng_status: c::NTSTATUS) -> (u64, u64) { + panic!("RNG broken: {rng_status:#x} fallback RNG broken: RtlGenRandom() not supported on UWP"); } diff --git a/library/std/src/sys/windows/stdio.rs b/library/std/src/sys/windows/stdio.rs index a001d6b98..70c9b14a0 100644 --- a/library/std/src/sys/windows/stdio.rs +++ b/library/std/src/sys/windows/stdio.rs @@ -3,6 +3,7 @@ use crate::char::decode_utf16; use crate::cmp; use crate::io; +use crate::mem::MaybeUninit; use crate::os::windows::io::{FromRawHandle, IntoRawHandle}; use crate::ptr; use crate::str; @@ -169,13 +170,14 @@ fn write( } fn write_valid_utf8_to_console(handle: c::HANDLE, utf8: &str) -> io::Result<usize> { - let mut utf16 = [0u16; MAX_BUFFER_SIZE / 2]; + let mut utf16 = [MaybeUninit::<u16>::uninit(); MAX_BUFFER_SIZE / 2]; let mut len_utf16 = 0; for (chr, dest) in utf8.encode_utf16().zip(utf16.iter_mut()) { - *dest = chr; + *dest = MaybeUninit::new(chr); len_utf16 += 1; } - let utf16 = &utf16[..len_utf16]; + // Safety: We've initialized `len_utf16` values. + let utf16: &[u16] = unsafe { MaybeUninit::slice_assume_init_ref(&utf16[..len_utf16]) }; let mut written = write_u16s(handle, &utf16)?; @@ -250,11 +252,14 @@ impl io::Read for Stdin { return Ok(bytes_copied); } else if buf.len() - bytes_copied < 4 { // Not enough space to get a UTF-8 byte. We will use the incomplete UTF8. - let mut utf16_buf = [0u16; 1]; + let mut utf16_buf = [MaybeUninit::new(0); 1]; // Read one u16 character. let read = read_u16s_fixup_surrogates(handle, &mut utf16_buf, 1, &mut self.surrogate)?; // Read bytes, using the (now-empty) self.incomplete_utf8 as extra space. - let read_bytes = utf16_to_utf8(&utf16_buf[..read], &mut self.incomplete_utf8.bytes)?; + let read_bytes = utf16_to_utf8( + unsafe { MaybeUninit::slice_assume_init_ref(&utf16_buf[..read]) }, + &mut self.incomplete_utf8.bytes, + )?; // Read in the bytes from incomplete_utf8 until the buffer is full. self.incomplete_utf8.len = read_bytes as u8; @@ -262,15 +267,18 @@ impl io::Read for Stdin { bytes_copied += self.incomplete_utf8.read(&mut buf[bytes_copied..]); Ok(bytes_copied) } else { - let mut utf16_buf = [0u16; MAX_BUFFER_SIZE / 2]; + let mut utf16_buf = [MaybeUninit::<u16>::uninit(); MAX_BUFFER_SIZE / 2]; + // In the worst case, a UTF-8 string can take 3 bytes for every `u16` of a UTF-16. So // we can read at most a third of `buf.len()` chars and uphold the guarantee no data gets // lost. let amount = cmp::min(buf.len() / 3, utf16_buf.len()); let read = read_u16s_fixup_surrogates(handle, &mut utf16_buf, amount, &mut self.surrogate)?; - - match utf16_to_utf8(&utf16_buf[..read], buf) { + // Safety `read_u16s_fixup_surrogates` returns the number of items + // initialized. + let utf16s = unsafe { MaybeUninit::slice_assume_init_ref(&utf16_buf[..read]) }; + match utf16_to_utf8(utf16s, buf) { Ok(value) => return Ok(bytes_copied + value), Err(e) => return Err(e), } @@ -283,14 +291,14 @@ impl io::Read for Stdin { // This is a best effort, and might not work if we are not the only reader on Stdin. fn read_u16s_fixup_surrogates( handle: c::HANDLE, - buf: &mut [u16], + buf: &mut [MaybeUninit<u16>], mut amount: usize, surrogate: &mut u16, ) -> io::Result<usize> { // Insert possibly remaining unpaired surrogate from last read. let mut start = 0; if *surrogate != 0 { - buf[0] = *surrogate; + buf[0] = MaybeUninit::new(*surrogate); *surrogate = 0; start = 1; if amount == 1 { @@ -303,7 +311,10 @@ fn read_u16s_fixup_surrogates( let mut amount = read_u16s(handle, &mut buf[start..amount])? + start; if amount > 0 { - let last_char = buf[amount - 1]; + // Safety: The returned `amount` is the number of values initialized, + // and it is not 0, so we know that `buf[amount - 1]` have been + // initialized. + let last_char = unsafe { buf[amount - 1].assume_init() }; if last_char >= 0xD800 && last_char <= 0xDBFF { // high surrogate *surrogate = last_char; @@ -313,7 +324,8 @@ fn read_u16s_fixup_surrogates( Ok(amount) } -fn read_u16s(handle: c::HANDLE, buf: &mut [u16]) -> io::Result<usize> { +// Returns `Ok(n)` if it initialized `n` values in `buf`. +fn read_u16s(handle: c::HANDLE, buf: &mut [MaybeUninit<u16>]) -> io::Result<usize> { // Configure the `pInputControl` parameter to not only return on `\r\n` but also Ctrl-Z, the // traditional DOS method to indicate end of character stream / user input (SUB). // See #38274 and https://stackoverflow.com/questions/43836040/win-api-readconsole. @@ -346,8 +358,9 @@ fn read_u16s(handle: c::HANDLE, buf: &mut [u16]) -> io::Result<usize> { } break; } - - if amount > 0 && buf[amount as usize - 1] == CTRL_Z { + // Safety: if `amount > 0`, then that many bytes were written, so + // `buf[amount as usize - 1]` has been initialized. + if amount > 0 && unsafe { buf[amount as usize - 1].assume_init() } == CTRL_Z { amount -= 1; } Ok(amount as usize) diff --git a/library/std/src/sys/windows/thread_local_dtor.rs b/library/std/src/sys/windows/thread_local_dtor.rs index 25d1c6e8e..9707a95df 100644 --- a/library/std/src/sys/windows/thread_local_dtor.rs +++ b/library/std/src/sys/windows/thread_local_dtor.rs @@ -8,10 +8,14 @@ #[thread_local] static mut DESTRUCTORS: Vec<(*mut u8, unsafe extern "C" fn(*mut u8))> = Vec::new(); +// Ensure this can never be inlined because otherwise this may break in dylibs. +// See #44391. +#[inline(never)] pub unsafe fn register_dtor(t: *mut u8, dtor: unsafe extern "C" fn(*mut u8)) { DESTRUCTORS.push((t, dtor)); } +#[inline(never)] // See comment above /// Runs destructors. This should not be called until thread exit. pub unsafe fn run_keyless_dtors() { // Drop all the destructors. diff --git a/library/std/src/sys/windows/thread_parker.rs b/library/std/src/sys/windows/thread_parker.rs index d876e0f6f..2f7ae863b 100644 --- a/library/std/src/sys/windows/thread_parker.rs +++ b/library/std/src/sys/windows/thread_parker.rs @@ -197,19 +197,17 @@ impl Parker { // purpose, to make sure every unpark() has a release-acquire ordering // with park(). if self.state.swap(NOTIFIED, Release) == PARKED { - if let Some(wake_by_address_single) = c::WakeByAddressSingle::option() { - unsafe { + unsafe { + if let Some(wake_by_address_single) = c::WakeByAddressSingle::option() { wake_by_address_single(self.ptr()); - } - } else { - // If we run NtReleaseKeyedEvent before the waiting thread runs - // NtWaitForKeyedEvent, this (shortly) blocks until we can wake it up. - // If the waiting thread wakes up before we run NtReleaseKeyedEvent - // (e.g. due to a timeout), this blocks until we do wake up a thread. - // To prevent this thread from blocking indefinitely in that case, - // park_impl() will, after seeing the state set to NOTIFIED after - // waking up, call NtWaitForKeyedEvent again to unblock us. - unsafe { + } else { + // If we run NtReleaseKeyedEvent before the waiting thread runs + // NtWaitForKeyedEvent, this (shortly) blocks until we can wake it up. + // If the waiting thread wakes up before we run NtReleaseKeyedEvent + // (e.g. due to a timeout), this blocks until we do wake up a thread. + // To prevent this thread from blocking indefinitely in that case, + // park_impl() will, after seeing the state set to NOTIFIED after + // waking up, call NtWaitForKeyedEvent again to unblock us. c::NtReleaseKeyedEvent(keyed_event_handle(), self.ptr(), 0, ptr::null_mut()); } } diff --git a/library/std/src/sys_common/net.rs b/library/std/src/sys_common/net.rs index 33d336c43..3ad802afa 100644 --- a/library/std/src/sys_common/net.rs +++ b/library/std/src/sys_common/net.rs @@ -10,7 +10,7 @@ use crate::net::{Ipv4Addr, Ipv6Addr, Shutdown, SocketAddr}; use crate::ptr; use crate::sys::net::netc as c; use crate::sys::net::{cvt, cvt_gai, cvt_r, init, wrlen_t, Socket}; -use crate::sys_common::{FromInner, IntoInner}; +use crate::sys_common::{AsInner, FromInner, IntoInner}; use crate::time::Duration; use libc::{c_int, c_void}; @@ -345,6 +345,12 @@ impl TcpStream { } } +impl AsInner<Socket> for TcpStream { + fn as_inner(&self) -> &Socket { + &self.inner + } +} + impl FromInner<Socket> for TcpStream { fn from_inner(socket: Socket) -> TcpStream { TcpStream { inner: socket } diff --git a/library/std/src/sys_common/remutex.rs b/library/std/src/sys_common/remutex.rs index 8921af311..b448ae3a9 100644 --- a/library/std/src/sys_common/remutex.rs +++ b/library/std/src/sys_common/remutex.rs @@ -1,13 +1,11 @@ #[cfg(all(test, not(target_os = "emscripten")))] mod tests; +use super::mutex as sys; use crate::cell::UnsafeCell; -use crate::marker::PhantomPinned; use crate::ops::Deref; use crate::panic::{RefUnwindSafe, UnwindSafe}; -use crate::pin::Pin; use crate::sync::atomic::{AtomicUsize, Ordering::Relaxed}; -use crate::sys::locks as sys; /// A re-entrant mutual exclusion /// @@ -41,11 +39,10 @@ use crate::sys::locks as sys; /// synchronization is left to the mutex, making relaxed memory ordering for /// the `owner` field fine in all cases. pub struct ReentrantMutex<T> { - mutex: sys::Mutex, + mutex: sys::MovableMutex, owner: AtomicUsize, lock_count: UnsafeCell<u32>, data: T, - _pinned: PhantomPinned, } unsafe impl<T: Send> Send for ReentrantMutex<T> {} @@ -68,39 +65,22 @@ impl<T> RefUnwindSafe for ReentrantMutex<T> {} /// guarded data. #[must_use = "if unused the ReentrantMutex will immediately unlock"] pub struct ReentrantMutexGuard<'a, T: 'a> { - lock: Pin<&'a ReentrantMutex<T>>, + lock: &'a ReentrantMutex<T>, } impl<T> !Send for ReentrantMutexGuard<'_, T> {} impl<T> ReentrantMutex<T> { /// Creates a new reentrant mutex in an unlocked state. - /// - /// # Unsafety - /// - /// This function is unsafe because it is required that `init` is called - /// once this mutex is in its final resting place, and only then are the - /// lock/unlock methods safe. - pub const unsafe fn new(t: T) -> ReentrantMutex<T> { + pub const fn new(t: T) -> ReentrantMutex<T> { ReentrantMutex { - mutex: sys::Mutex::new(), + mutex: sys::MovableMutex::new(), owner: AtomicUsize::new(0), lock_count: UnsafeCell::new(0), data: t, - _pinned: PhantomPinned, } } - /// Initializes this mutex so it's ready for use. - /// - /// # Unsafety - /// - /// Unsafe to call more than once, and must be called after this will no - /// longer move in memory. - pub unsafe fn init(self: Pin<&mut Self>) { - self.get_unchecked_mut().mutex.init() - } - /// Acquires a mutex, blocking the current thread until it is able to do so. /// /// This function will block the caller until it is available to acquire the mutex. @@ -113,15 +93,14 @@ impl<T> ReentrantMutex<T> { /// If another user of this mutex panicked while holding the mutex, then /// this call will return failure if the mutex would otherwise be /// acquired. - pub fn lock(self: Pin<&Self>) -> ReentrantMutexGuard<'_, T> { + pub fn lock(&self) -> ReentrantMutexGuard<'_, T> { let this_thread = current_thread_unique_ptr(); - // Safety: We only touch lock_count when we own the lock, - // and since self is pinned we can safely call the lock() on the mutex. + // Safety: We only touch lock_count when we own the lock. unsafe { if self.owner.load(Relaxed) == this_thread { self.increment_lock_count(); } else { - self.mutex.lock(); + self.mutex.raw_lock(); self.owner.store(this_thread, Relaxed); debug_assert_eq!(*self.lock_count.get(), 0); *self.lock_count.get() = 1; @@ -142,10 +121,9 @@ impl<T> ReentrantMutex<T> { /// If another user of this mutex panicked while holding the mutex, then /// this call will return failure if the mutex would otherwise be /// acquired. - pub fn try_lock(self: Pin<&Self>) -> Option<ReentrantMutexGuard<'_, T>> { + pub fn try_lock(&self) -> Option<ReentrantMutexGuard<'_, T>> { let this_thread = current_thread_unique_ptr(); - // Safety: We only touch lock_count when we own the lock, - // and since self is pinned we can safely call the try_lock on the mutex. + // Safety: We only touch lock_count when we own the lock. unsafe { if self.owner.load(Relaxed) == this_thread { self.increment_lock_count(); @@ -179,12 +157,12 @@ impl<T> Deref for ReentrantMutexGuard<'_, T> { impl<T> Drop for ReentrantMutexGuard<'_, T> { #[inline] fn drop(&mut self) { - // Safety: We own the lock, and the lock is pinned. + // Safety: We own the lock. unsafe { *self.lock.lock_count.get() -= 1; if *self.lock.lock_count.get() == 0 { self.lock.owner.store(0, Relaxed); - self.lock.mutex.unlock(); + self.lock.mutex.raw_unlock(); } } } diff --git a/library/std/src/sys_common/remutex/tests.rs b/library/std/src/sys_common/remutex/tests.rs index 64873b850..8e97ce11c 100644 --- a/library/std/src/sys_common/remutex/tests.rs +++ b/library/std/src/sys_common/remutex/tests.rs @@ -1,18 +1,11 @@ -use crate::boxed::Box; use crate::cell::RefCell; -use crate::pin::Pin; use crate::sync::Arc; use crate::sys_common::remutex::{ReentrantMutex, ReentrantMutexGuard}; use crate::thread; #[test] fn smoke() { - let m = unsafe { - let mut m = Box::pin(ReentrantMutex::new(())); - m.as_mut().init(); - m - }; - let m = m.as_ref(); + let m = ReentrantMutex::new(()); { let a = m.lock(); { @@ -29,20 +22,15 @@ fn smoke() { #[test] fn is_mutex() { - let m = unsafe { - // FIXME: Simplify this if Arc gets an Arc::get_pin_mut. - let mut m = Arc::new(ReentrantMutex::new(RefCell::new(0))); - Pin::new_unchecked(Arc::get_mut_unchecked(&mut m)).init(); - Pin::new_unchecked(m) - }; + let m = Arc::new(ReentrantMutex::new(RefCell::new(0))); let m2 = m.clone(); - let lock = m.as_ref().lock(); + let lock = m.lock(); let child = thread::spawn(move || { - let lock = m2.as_ref().lock(); + let lock = m2.lock(); assert_eq!(*lock.borrow(), 4950); }); for i in 0..100 { - let lock = m.as_ref().lock(); + let lock = m.lock(); *lock.borrow_mut() += i; } drop(lock); @@ -51,22 +39,17 @@ fn is_mutex() { #[test] fn trylock_works() { - let m = unsafe { - // FIXME: Simplify this if Arc gets an Arc::get_pin_mut. - let mut m = Arc::new(ReentrantMutex::new(())); - Pin::new_unchecked(Arc::get_mut_unchecked(&mut m)).init(); - Pin::new_unchecked(m) - }; + let m = Arc::new(ReentrantMutex::new(())); let m2 = m.clone(); - let _lock = m.as_ref().try_lock(); - let _lock2 = m.as_ref().try_lock(); + let _lock = m.try_lock(); + let _lock2 = m.try_lock(); thread::spawn(move || { - let lock = m2.as_ref().try_lock(); + let lock = m2.try_lock(); assert!(lock.is_none()); }) .join() .unwrap(); - let _lock3 = m.as_ref().try_lock(); + let _lock3 = m.try_lock(); } pub struct Answer<'a>(pub ReentrantMutexGuard<'a, RefCell<u32>>); diff --git a/library/std/src/sys_common/thread_local_key.rs b/library/std/src/sys_common/thread_local_key.rs index 70beebe86..032bf604d 100644 --- a/library/std/src/sys_common/thread_local_key.rs +++ b/library/std/src/sys_common/thread_local_key.rs @@ -69,8 +69,10 @@ use crate::sys_common::mutex::StaticMutex; /// ```ignore (cannot-doctest-private-modules) /// use tls::os::{StaticKey, INIT}; /// +/// // Use a regular global static to store the key. /// static KEY: StaticKey = INIT; /// +/// // The state provided via `get` and `set` is thread-local. /// unsafe { /// assert!(KEY.get().is_null()); /// KEY.set(1 as *mut u8); diff --git a/library/std/src/sys_common/thread_local_key/tests.rs b/library/std/src/sys_common/thread_local_key/tests.rs index 968738a41..6f32b858f 100644 --- a/library/std/src/sys_common/thread_local_key/tests.rs +++ b/library/std/src/sys_common/thread_local_key/tests.rs @@ -1,4 +1,5 @@ use super::{Key, StaticKey}; +use core::ptr; fn assert_sync<T: Sync>() {} fn assert_send<T: Send>() {} @@ -12,8 +13,8 @@ fn smoke() { let k2 = Key::new(None); assert!(k1.get().is_null()); assert!(k2.get().is_null()); - k1.set(1 as *mut _); - k2.set(2 as *mut _); + k1.set(ptr::invalid_mut(1)); + k2.set(ptr::invalid_mut(2)); assert_eq!(k1.get() as usize, 1); assert_eq!(k2.get() as usize, 2); } @@ -26,8 +27,8 @@ fn statik() { unsafe { assert!(K1.get().is_null()); assert!(K2.get().is_null()); - K1.set(1 as *mut _); - K2.set(2 as *mut _); + K1.set(ptr::invalid_mut(1)); + K2.set(ptr::invalid_mut(2)); assert_eq!(K1.get() as usize, 1); assert_eq!(K2.get() as usize, 2); } diff --git a/library/std/src/sys_common/thread_parker/mod.rs b/library/std/src/sys_common/thread_parker/mod.rs index cbd7832eb..f86a9a555 100644 --- a/library/std/src/sys_common/thread_parker/mod.rs +++ b/library/std/src/sys_common/thread_parker/mod.rs @@ -7,6 +7,7 @@ cfg_if::cfg_if! { target_os = "openbsd", target_os = "dragonfly", target_os = "fuchsia", + target_os = "hermit", ))] { mod futex; pub use futex::Parker; diff --git a/library/std/src/sys_common/wtf8.rs b/library/std/src/sys_common/wtf8.rs index 57fa49893..dd53767d4 100644 --- a/library/std/src/sys_common/wtf8.rs +++ b/library/std/src/sys_common/wtf8.rs @@ -89,6 +89,24 @@ impl CodePoint { self.value } + /// Returns the numeric value of the code point if it is a leading surrogate. + #[inline] + pub fn to_lead_surrogate(&self) -> Option<u16> { + match self.value { + lead @ 0xD800..=0xDBFF => Some(lead as u16), + _ => None, + } + } + + /// Returns the numeric value of the code point if it is a trailing surrogate. + #[inline] + pub fn to_trail_surrogate(&self) -> Option<u16> { + match self.value { + trail @ 0xDC00..=0xDFFF => Some(trail as u16), + _ => None, + } + } + /// Optionally returns a Unicode scalar value for the code point. /// /// Returns `None` if the code point is a surrogate (from U+D800 to U+DFFF). @@ -117,6 +135,14 @@ impl CodePoint { #[derive(Eq, PartialEq, Ord, PartialOrd, Clone)] pub struct Wtf8Buf { bytes: Vec<u8>, + + /// Do we know that `bytes` holds a valid UTF-8 encoding? We can easily + /// know this if we're constructed from a `String` or `&str`. + /// + /// It is possible for `bytes` to have valid UTF-8 without this being + /// set, such as when we're concatenating `&Wtf8`'s and surrogates become + /// paired, as we don't bother to rescan the entire string. + is_known_utf8: bool, } impl ops::Deref for Wtf8Buf { @@ -147,13 +173,13 @@ impl Wtf8Buf { /// Creates a new, empty WTF-8 string. #[inline] pub fn new() -> Wtf8Buf { - Wtf8Buf { bytes: Vec::new() } + Wtf8Buf { bytes: Vec::new(), is_known_utf8: true } } /// Creates a new, empty WTF-8 string with pre-allocated capacity for `capacity` bytes. #[inline] pub fn with_capacity(capacity: usize) -> Wtf8Buf { - Wtf8Buf { bytes: Vec::with_capacity(capacity) } + Wtf8Buf { bytes: Vec::with_capacity(capacity), is_known_utf8: true } } /// Creates a WTF-8 string from a UTF-8 `String`. @@ -163,7 +189,7 @@ impl Wtf8Buf { /// Since WTF-8 is a superset of UTF-8, this always succeeds. #[inline] pub fn from_string(string: String) -> Wtf8Buf { - Wtf8Buf { bytes: string.into_bytes() } + Wtf8Buf { bytes: string.into_bytes(), is_known_utf8: true } } /// Creates a WTF-8 string from a UTF-8 `&str` slice. @@ -173,11 +199,12 @@ impl Wtf8Buf { /// Since WTF-8 is a superset of UTF-8, this always succeeds. #[inline] pub fn from_str(str: &str) -> Wtf8Buf { - Wtf8Buf { bytes: <[_]>::to_vec(str.as_bytes()) } + Wtf8Buf { bytes: <[_]>::to_vec(str.as_bytes()), is_known_utf8: true } } pub fn clear(&mut self) { - self.bytes.clear() + self.bytes.clear(); + self.is_known_utf8 = true; } /// Creates a WTF-8 string from a potentially ill-formed UTF-16 slice of 16-bit code units. @@ -193,9 +220,11 @@ impl Wtf8Buf { let surrogate = surrogate.unpaired_surrogate(); // Surrogates are known to be in the code point range. let code_point = unsafe { CodePoint::from_u32_unchecked(surrogate as u32) }; + // The string will now contain an unpaired surrogate. + string.is_known_utf8 = false; // Skip the WTF-8 concatenation check, // surrogate pairs are already decoded by decode_utf16 - string.push_code_point_unchecked(code_point) + string.push_code_point_unchecked(code_point); } } } @@ -203,7 +232,7 @@ impl Wtf8Buf { } /// Copied from String::push - /// This does **not** include the WTF-8 concatenation check. + /// This does **not** include the WTF-8 concatenation check or `is_known_utf8` check. fn push_code_point_unchecked(&mut self, code_point: CodePoint) { let mut bytes = [0; 4]; let bytes = char::encode_utf8_raw(code_point.value, &mut bytes); @@ -217,6 +246,9 @@ impl Wtf8Buf { #[inline] pub fn as_mut_slice(&mut self) -> &mut Wtf8 { + // Safety: `Wtf8` doesn't expose any way to mutate the bytes that would + // cause them to change from well-formed UTF-8 to ill-formed UTF-8, + // which would break the assumptions of the `is_known_utf8` field. unsafe { Wtf8::from_mut_bytes_unchecked(&mut self.bytes) } } @@ -236,7 +268,8 @@ impl Wtf8Buf { /// in the given `Wtf8Buf`. The `Wtf8Buf` may reserve more space to avoid /// frequent reallocations. After calling `try_reserve`, capacity will be /// greater than or equal to `self.len() + additional`. Does nothing if - /// capacity is already sufficient. + /// capacity is already sufficient. This method preserves the contents even + /// if an error occurs. /// /// # Errors /// @@ -313,7 +346,15 @@ impl Wtf8Buf { self.push_char(decode_surrogate_pair(lead, trail)); self.bytes.extend_from_slice(other_without_trail_surrogate); } - _ => self.bytes.extend_from_slice(&other.bytes), + _ => { + // If we'll be pushing a string containing a surrogate, we may + // no longer have UTF-8. + if other.next_surrogate(0).is_some() { + self.is_known_utf8 = false; + } + + self.bytes.extend_from_slice(&other.bytes); + } } } @@ -330,13 +371,19 @@ impl Wtf8Buf { /// like concatenating ill-formed UTF-16 strings effectively would. #[inline] pub fn push(&mut self, code_point: CodePoint) { - if let trail @ 0xDC00..=0xDFFF = code_point.to_u32() { + if let Some(trail) = code_point.to_trail_surrogate() { if let Some(lead) = (&*self).final_lead_surrogate() { let len_without_lead_surrogate = self.len() - 3; self.bytes.truncate(len_without_lead_surrogate); - self.push_char(decode_surrogate_pair(lead, trail as u16)); + self.push_char(decode_surrogate_pair(lead, trail)); return; } + + // We're pushing a trailing surrogate. + self.is_known_utf8 = false; + } else if code_point.to_lead_surrogate().is_some() { + // We're pushing a leading surrogate. + self.is_known_utf8 = false; } // No newly paired surrogates at the boundary. @@ -363,9 +410,10 @@ impl Wtf8Buf { /// (that is, if the string contains surrogates), /// the original WTF-8 string is returned instead. pub fn into_string(self) -> Result<String, Wtf8Buf> { - match self.next_surrogate(0) { - None => Ok(unsafe { String::from_utf8_unchecked(self.bytes) }), - Some(_) => Err(self), + if self.is_known_utf8 || self.next_surrogate(0).is_none() { + Ok(unsafe { String::from_utf8_unchecked(self.bytes) }) + } else { + Err(self) } } @@ -375,6 +423,11 @@ impl Wtf8Buf { /// /// Surrogates are replaced with `"\u{FFFD}"` (the replacement character āļæ½ā) pub fn into_string_lossy(mut self) -> String { + // Fast path: If we already have UTF-8, we can return it immediately. + if self.is_known_utf8 { + return unsafe { String::from_utf8_unchecked(self.bytes) }; + } + let mut pos = 0; loop { match self.next_surrogate(pos) { @@ -397,7 +450,7 @@ impl Wtf8Buf { /// Converts a `Box<Wtf8>` into a `Wtf8Buf`. pub fn from_box(boxed: Box<Wtf8>) -> Wtf8Buf { let bytes: Box<[u8]> = unsafe { mem::transmute(boxed) }; - Wtf8Buf { bytes: bytes.into_vec() } + Wtf8Buf { bytes: bytes.into_vec(), is_known_utf8: false } } } @@ -575,6 +628,11 @@ impl Wtf8 { } } + /// Creates an owned `Wtf8Buf` from a borrowed `Wtf8`. + pub fn to_owned(&self) -> Wtf8Buf { + Wtf8Buf { bytes: self.bytes.to_vec(), is_known_utf8: false } + } + /// Lossily converts the string to UTF-8. /// Returns a UTF-8 `&str` slice if the contents are well-formed in UTF-8. /// @@ -664,7 +722,8 @@ impl Wtf8 { } pub fn clone_into(&self, buf: &mut Wtf8Buf) { - self.bytes.clone_into(&mut buf.bytes) + buf.is_known_utf8 = false; + self.bytes.clone_into(&mut buf.bytes); } /// Boxes this `Wtf8`. @@ -704,12 +763,12 @@ impl Wtf8 { #[inline] pub fn to_ascii_lowercase(&self) -> Wtf8Buf { - Wtf8Buf { bytes: self.bytes.to_ascii_lowercase() } + Wtf8Buf { bytes: self.bytes.to_ascii_lowercase(), is_known_utf8: false } } #[inline] pub fn to_ascii_uppercase(&self) -> Wtf8Buf { - Wtf8Buf { bytes: self.bytes.to_ascii_uppercase() } + Wtf8Buf { bytes: self.bytes.to_ascii_uppercase(), is_known_utf8: false } } #[inline] diff --git a/library/std/src/sys_common/wtf8/tests.rs b/library/std/src/sys_common/wtf8/tests.rs index 931996791..1a302d646 100644 --- a/library/std/src/sys_common/wtf8/tests.rs +++ b/library/std/src/sys_common/wtf8/tests.rs @@ -20,6 +20,36 @@ fn code_point_to_u32() { } #[test] +fn code_point_to_lead_surrogate() { + fn c(value: u32) -> CodePoint { + CodePoint::from_u32(value).unwrap() + } + assert_eq!(c(0).to_lead_surrogate(), None); + assert_eq!(c(0xE9).to_lead_surrogate(), None); + assert_eq!(c(0xD800).to_lead_surrogate(), Some(0xD800)); + assert_eq!(c(0xDBFF).to_lead_surrogate(), Some(0xDBFF)); + assert_eq!(c(0xDC00).to_lead_surrogate(), None); + assert_eq!(c(0xDFFF).to_lead_surrogate(), None); + assert_eq!(c(0x1F4A9).to_lead_surrogate(), None); + assert_eq!(c(0x10FFFF).to_lead_surrogate(), None); +} + +#[test] +fn code_point_to_trail_surrogate() { + fn c(value: u32) -> CodePoint { + CodePoint::from_u32(value).unwrap() + } + assert_eq!(c(0).to_trail_surrogate(), None); + assert_eq!(c(0xE9).to_trail_surrogate(), None); + assert_eq!(c(0xD800).to_trail_surrogate(), None); + assert_eq!(c(0xDBFF).to_trail_surrogate(), None); + assert_eq!(c(0xDC00).to_trail_surrogate(), Some(0xDC00)); + assert_eq!(c(0xDFFF).to_trail_surrogate(), Some(0xDFFF)); + assert_eq!(c(0x1F4A9).to_trail_surrogate(), None); + assert_eq!(c(0x10FFFF).to_trail_surrogate(), None); +} + +#[test] fn code_point_from_char() { assert_eq!(CodePoint::from_char('a').to_u32(), 0x61); assert_eq!(CodePoint::from_char('š©').to_u32(), 0x1F4A9); @@ -70,35 +100,66 @@ fn wtf8buf_from_string() { #[test] fn wtf8buf_from_wide() { - assert_eq!(Wtf8Buf::from_wide(&[]).bytes, b""); - assert_eq!( - Wtf8Buf::from_wide(&[0x61, 0xE9, 0x20, 0xD83D, 0xD83D, 0xDCA9]).bytes, - b"a\xC3\xA9 \xED\xA0\xBD\xF0\x9F\x92\xA9" - ); + let buf = Wtf8Buf::from_wide(&[]); + assert_eq!(buf.bytes, b""); + assert!(buf.is_known_utf8); + + let buf = Wtf8Buf::from_wide(&[0x61, 0xE9, 0x20, 0xD83D, 0xDCA9]); + assert_eq!(buf.bytes, b"a\xC3\xA9 \xF0\x9F\x92\xA9"); + assert!(buf.is_known_utf8); + + let buf = Wtf8Buf::from_wide(&[0x61, 0xE9, 0x20, 0xD83D, 0xD83D, 0xDCA9]); + assert_eq!(buf.bytes, b"a\xC3\xA9 \xED\xA0\xBD\xF0\x9F\x92\xA9"); + assert!(!buf.is_known_utf8); + + let buf = Wtf8Buf::from_wide(&[0xD800]); + assert_eq!(buf.bytes, b"\xED\xA0\x80"); + assert!(!buf.is_known_utf8); + + let buf = Wtf8Buf::from_wide(&[0xDBFF]); + assert_eq!(buf.bytes, b"\xED\xAF\xBF"); + assert!(!buf.is_known_utf8); + + let buf = Wtf8Buf::from_wide(&[0xDC00]); + assert_eq!(buf.bytes, b"\xED\xB0\x80"); + assert!(!buf.is_known_utf8); + + let buf = Wtf8Buf::from_wide(&[0xDFFF]); + assert_eq!(buf.bytes, b"\xED\xBF\xBF"); + assert!(!buf.is_known_utf8); } #[test] fn wtf8buf_push_str() { let mut string = Wtf8Buf::new(); assert_eq!(string.bytes, b""); + assert!(string.is_known_utf8); + string.push_str("aĆ© š©"); assert_eq!(string.bytes, b"a\xC3\xA9 \xF0\x9F\x92\xA9"); + assert!(string.is_known_utf8); } #[test] fn wtf8buf_push_char() { let mut string = Wtf8Buf::from_str("aĆ© "); assert_eq!(string.bytes, b"a\xC3\xA9 "); + assert!(string.is_known_utf8); + string.push_char('š©'); assert_eq!(string.bytes, b"a\xC3\xA9 \xF0\x9F\x92\xA9"); + assert!(string.is_known_utf8); } #[test] fn wtf8buf_push() { let mut string = Wtf8Buf::from_str("aĆ© "); assert_eq!(string.bytes, b"a\xC3\xA9 "); + assert!(string.is_known_utf8); + string.push(CodePoint::from_char('š©')); assert_eq!(string.bytes, b"a\xC3\xA9 \xF0\x9F\x92\xA9"); + assert!(string.is_known_utf8); fn c(value: u32) -> CodePoint { CodePoint::from_u32(value).unwrap() @@ -106,37 +167,46 @@ fn wtf8buf_push() { let mut string = Wtf8Buf::new(); string.push(c(0xD83D)); // lead + assert!(!string.is_known_utf8); string.push(c(0xDCA9)); // trail assert_eq!(string.bytes, b"\xF0\x9F\x92\xA9"); // Magic! let mut string = Wtf8Buf::new(); string.push(c(0xD83D)); // lead + assert!(!string.is_known_utf8); string.push(c(0x20)); // not surrogate string.push(c(0xDCA9)); // trail assert_eq!(string.bytes, b"\xED\xA0\xBD \xED\xB2\xA9"); let mut string = Wtf8Buf::new(); string.push(c(0xD800)); // lead + assert!(!string.is_known_utf8); string.push(c(0xDBFF)); // lead assert_eq!(string.bytes, b"\xED\xA0\x80\xED\xAF\xBF"); let mut string = Wtf8Buf::new(); string.push(c(0xD800)); // lead + assert!(!string.is_known_utf8); string.push(c(0xE000)); // not surrogate assert_eq!(string.bytes, b"\xED\xA0\x80\xEE\x80\x80"); let mut string = Wtf8Buf::new(); string.push(c(0xD7FF)); // not surrogate + assert!(string.is_known_utf8); string.push(c(0xDC00)); // trail + assert!(!string.is_known_utf8); assert_eq!(string.bytes, b"\xED\x9F\xBF\xED\xB0\x80"); let mut string = Wtf8Buf::new(); string.push(c(0x61)); // not surrogate, < 3 bytes + assert!(string.is_known_utf8); string.push(c(0xDC00)); // trail + assert!(!string.is_known_utf8); assert_eq!(string.bytes, b"\x61\xED\xB0\x80"); let mut string = Wtf8Buf::new(); string.push(c(0xDC00)); // trail + assert!(!string.is_known_utf8); assert_eq!(string.bytes, b"\xED\xB0\x80"); } @@ -146,6 +216,7 @@ fn wtf8buf_push_wtf8() { assert_eq!(string.bytes, b"a\xC3\xA9"); string.push_wtf8(Wtf8::from_str(" š©")); assert_eq!(string.bytes, b"a\xC3\xA9 \xF0\x9F\x92\xA9"); + assert!(string.is_known_utf8); fn w(v: &[u8]) -> &Wtf8 { unsafe { Wtf8::from_bytes_unchecked(v) } @@ -161,37 +232,68 @@ fn wtf8buf_push_wtf8() { string.push_wtf8(w(b" ")); // not surrogate string.push_wtf8(w(b"\xED\xB2\xA9")); // trail assert_eq!(string.bytes, b"\xED\xA0\xBD \xED\xB2\xA9"); + assert!(!string.is_known_utf8); let mut string = Wtf8Buf::new(); string.push_wtf8(w(b"\xED\xA0\x80")); // lead string.push_wtf8(w(b"\xED\xAF\xBF")); // lead assert_eq!(string.bytes, b"\xED\xA0\x80\xED\xAF\xBF"); + assert!(!string.is_known_utf8); let mut string = Wtf8Buf::new(); string.push_wtf8(w(b"\xED\xA0\x80")); // lead string.push_wtf8(w(b"\xEE\x80\x80")); // not surrogate assert_eq!(string.bytes, b"\xED\xA0\x80\xEE\x80\x80"); + assert!(!string.is_known_utf8); let mut string = Wtf8Buf::new(); string.push_wtf8(w(b"\xED\x9F\xBF")); // not surrogate string.push_wtf8(w(b"\xED\xB0\x80")); // trail assert_eq!(string.bytes, b"\xED\x9F\xBF\xED\xB0\x80"); + assert!(!string.is_known_utf8); let mut string = Wtf8Buf::new(); string.push_wtf8(w(b"a")); // not surrogate, < 3 bytes string.push_wtf8(w(b"\xED\xB0\x80")); // trail assert_eq!(string.bytes, b"\x61\xED\xB0\x80"); + assert!(!string.is_known_utf8); let mut string = Wtf8Buf::new(); string.push_wtf8(w(b"\xED\xB0\x80")); // trail assert_eq!(string.bytes, b"\xED\xB0\x80"); + assert!(!string.is_known_utf8); } #[test] fn wtf8buf_truncate() { let mut string = Wtf8Buf::from_str("aĆ©"); + assert!(string.is_known_utf8); + + string.truncate(3); + assert_eq!(string.bytes, b"a\xC3\xA9"); + assert!(string.is_known_utf8); + string.truncate(1); assert_eq!(string.bytes, b"a"); + assert!(string.is_known_utf8); + + string.truncate(0); + assert_eq!(string.bytes, b""); + assert!(string.is_known_utf8); +} + +#[test] +fn wtf8buf_truncate_around_non_bmp() { + let mut string = Wtf8Buf::from_str("š©"); + assert!(string.is_known_utf8); + + string.truncate(4); + assert_eq!(string.bytes, b"\xF0\x9F\x92\xA9"); + assert!(string.is_known_utf8); + + string.truncate(0); + assert_eq!(string.bytes, b""); + assert!(string.is_known_utf8); } #[test] @@ -209,10 +311,36 @@ fn wtf8buf_truncate_fail_longer() { } #[test] +#[should_panic] +fn wtf8buf_truncate_splitting_non_bmp3() { + let mut string = Wtf8Buf::from_str("š©"); + assert!(string.is_known_utf8); + string.truncate(3); +} + +#[test] +#[should_panic] +fn wtf8buf_truncate_splitting_non_bmp2() { + let mut string = Wtf8Buf::from_str("š©"); + assert!(string.is_known_utf8); + string.truncate(2); +} + +#[test] +#[should_panic] +fn wtf8buf_truncate_splitting_non_bmp1() { + let mut string = Wtf8Buf::from_str("š©"); + assert!(string.is_known_utf8); + string.truncate(1); +} + +#[test] fn wtf8buf_into_string() { let mut string = Wtf8Buf::from_str("aĆ© š©"); + assert!(string.is_known_utf8); assert_eq!(string.clone().into_string(), Ok(String::from("aĆ© š©"))); string.push(CodePoint::from_u32(0xD800).unwrap()); + assert!(!string.is_known_utf8); assert_eq!(string.clone().into_string(), Err(string)); } @@ -229,15 +357,33 @@ fn wtf8buf_from_iterator() { fn f(values: &[u32]) -> Wtf8Buf { values.iter().map(|&c| CodePoint::from_u32(c).unwrap()).collect::<Wtf8Buf>() } - assert_eq!(f(&[0x61, 0xE9, 0x20, 0x1F4A9]).bytes, b"a\xC3\xA9 \xF0\x9F\x92\xA9"); + assert_eq!( + f(&[0x61, 0xE9, 0x20, 0x1F4A9]), + Wtf8Buf { bytes: b"a\xC3\xA9 \xF0\x9F\x92\xA9".to_vec(), is_known_utf8: true } + ); assert_eq!(f(&[0xD83D, 0xDCA9]).bytes, b"\xF0\x9F\x92\xA9"); // Magic! - assert_eq!(f(&[0xD83D, 0x20, 0xDCA9]).bytes, b"\xED\xA0\xBD \xED\xB2\xA9"); - assert_eq!(f(&[0xD800, 0xDBFF]).bytes, b"\xED\xA0\x80\xED\xAF\xBF"); - assert_eq!(f(&[0xD800, 0xE000]).bytes, b"\xED\xA0\x80\xEE\x80\x80"); - assert_eq!(f(&[0xD7FF, 0xDC00]).bytes, b"\xED\x9F\xBF\xED\xB0\x80"); - assert_eq!(f(&[0x61, 0xDC00]).bytes, b"\x61\xED\xB0\x80"); - assert_eq!(f(&[0xDC00]).bytes, b"\xED\xB0\x80"); + assert_eq!( + f(&[0xD83D, 0x20, 0xDCA9]), + Wtf8Buf { bytes: b"\xED\xA0\xBD \xED\xB2\xA9".to_vec(), is_known_utf8: false } + ); + assert_eq!( + f(&[0xD800, 0xDBFF]), + Wtf8Buf { bytes: b"\xED\xA0\x80\xED\xAF\xBF".to_vec(), is_known_utf8: false } + ); + assert_eq!( + f(&[0xD800, 0xE000]), + Wtf8Buf { bytes: b"\xED\xA0\x80\xEE\x80\x80".to_vec(), is_known_utf8: false } + ); + assert_eq!( + f(&[0xD7FF, 0xDC00]), + Wtf8Buf { bytes: b"\xED\x9F\xBF\xED\xB0\x80".to_vec(), is_known_utf8: false } + ); + assert_eq!( + f(&[0x61, 0xDC00]), + Wtf8Buf { bytes: b"\x61\xED\xB0\x80".to_vec(), is_known_utf8: false } + ); + assert_eq!(f(&[0xDC00]), Wtf8Buf { bytes: b"\xED\xB0\x80".to_vec(), is_known_utf8: false }); } #[test] @@ -251,15 +397,36 @@ fn wtf8buf_extend() { string } - assert_eq!(e(&[0x61, 0xE9], &[0x20, 0x1F4A9]).bytes, b"a\xC3\xA9 \xF0\x9F\x92\xA9"); + assert_eq!( + e(&[0x61, 0xE9], &[0x20, 0x1F4A9]), + Wtf8Buf { bytes: b"a\xC3\xA9 \xF0\x9F\x92\xA9".to_vec(), is_known_utf8: true } + ); assert_eq!(e(&[0xD83D], &[0xDCA9]).bytes, b"\xF0\x9F\x92\xA9"); // Magic! - assert_eq!(e(&[0xD83D, 0x20], &[0xDCA9]).bytes, b"\xED\xA0\xBD \xED\xB2\xA9"); - assert_eq!(e(&[0xD800], &[0xDBFF]).bytes, b"\xED\xA0\x80\xED\xAF\xBF"); - assert_eq!(e(&[0xD800], &[0xE000]).bytes, b"\xED\xA0\x80\xEE\x80\x80"); - assert_eq!(e(&[0xD7FF], &[0xDC00]).bytes, b"\xED\x9F\xBF\xED\xB0\x80"); - assert_eq!(e(&[0x61], &[0xDC00]).bytes, b"\x61\xED\xB0\x80"); - assert_eq!(e(&[], &[0xDC00]).bytes, b"\xED\xB0\x80"); + assert_eq!( + e(&[0xD83D, 0x20], &[0xDCA9]), + Wtf8Buf { bytes: b"\xED\xA0\xBD \xED\xB2\xA9".to_vec(), is_known_utf8: false } + ); + assert_eq!( + e(&[0xD800], &[0xDBFF]), + Wtf8Buf { bytes: b"\xED\xA0\x80\xED\xAF\xBF".to_vec(), is_known_utf8: false } + ); + assert_eq!( + e(&[0xD800], &[0xE000]), + Wtf8Buf { bytes: b"\xED\xA0\x80\xEE\x80\x80".to_vec(), is_known_utf8: false } + ); + assert_eq!( + e(&[0xD7FF], &[0xDC00]), + Wtf8Buf { bytes: b"\xED\x9F\xBF\xED\xB0\x80".to_vec(), is_known_utf8: false } + ); + assert_eq!( + e(&[0x61], &[0xDC00]), + Wtf8Buf { bytes: b"\x61\xED\xB0\x80".to_vec(), is_known_utf8: false } + ); + assert_eq!( + e(&[], &[0xDC00]), + Wtf8Buf { bytes: b"\xED\xB0\x80".to_vec(), is_known_utf8: false } + ); } #[test] @@ -407,3 +574,93 @@ fn wtf8_encode_wide_size_hint() { assert_eq!((0, Some(0)), iter.size_hint()); assert!(iter.next().is_none()); } + +#[test] +fn wtf8_clone_into() { + let mut string = Wtf8Buf::new(); + Wtf8::from_str("green").clone_into(&mut string); + assert_eq!(string.bytes, b"green"); + + let mut string = Wtf8Buf::from_str("green"); + Wtf8::from_str("").clone_into(&mut string); + assert_eq!(string.bytes, b""); + + let mut string = Wtf8Buf::from_str("red"); + Wtf8::from_str("green").clone_into(&mut string); + assert_eq!(string.bytes, b"green"); + + let mut string = Wtf8Buf::from_str("green"); + Wtf8::from_str("red").clone_into(&mut string); + assert_eq!(string.bytes, b"red"); + + let mut string = Wtf8Buf::from_str("green"); + assert!(string.is_known_utf8); + unsafe { Wtf8::from_bytes_unchecked(b"\xED\xA0\x80").clone_into(&mut string) }; + assert_eq!(string.bytes, b"\xED\xA0\x80"); + assert!(!string.is_known_utf8); +} + +#[test] +fn wtf8_to_ascii_lowercase() { + let lowercase = Wtf8::from_str("").to_ascii_lowercase(); + assert_eq!(lowercase.bytes, b""); + + let lowercase = Wtf8::from_str("GrEeN gRaPeS! š").to_ascii_lowercase(); + assert_eq!(lowercase.bytes, b"green grapes! \xf0\x9f\x8d\x87"); + + let lowercase = unsafe { Wtf8::from_bytes_unchecked(b"\xED\xA0\x80").to_ascii_lowercase() }; + assert_eq!(lowercase.bytes, b"\xED\xA0\x80"); + assert!(!lowercase.is_known_utf8); +} + +#[test] +fn wtf8_to_ascii_uppercase() { + let uppercase = Wtf8::from_str("").to_ascii_uppercase(); + assert_eq!(uppercase.bytes, b""); + + let uppercase = Wtf8::from_str("GrEeN gRaPeS! š").to_ascii_uppercase(); + assert_eq!(uppercase.bytes, b"GREEN GRAPES! \xf0\x9f\x8d\x87"); + + let uppercase = unsafe { Wtf8::from_bytes_unchecked(b"\xED\xA0\x80").to_ascii_uppercase() }; + assert_eq!(uppercase.bytes, b"\xED\xA0\x80"); + assert!(!uppercase.is_known_utf8); +} + +#[test] +fn wtf8_make_ascii_lowercase() { + let mut lowercase = Wtf8Buf::from_str(""); + lowercase.make_ascii_lowercase(); + assert_eq!(lowercase.bytes, b""); + + let mut lowercase = Wtf8Buf::from_str("GrEeN gRaPeS! š"); + lowercase.make_ascii_lowercase(); + assert_eq!(lowercase.bytes, b"green grapes! \xf0\x9f\x8d\x87"); + + let mut lowercase = unsafe { Wtf8::from_bytes_unchecked(b"\xED\xA0\x80").to_owned() }; + lowercase.make_ascii_lowercase(); + assert_eq!(lowercase.bytes, b"\xED\xA0\x80"); + assert!(!lowercase.is_known_utf8); +} + +#[test] +fn wtf8_make_ascii_uppercase() { + let mut uppercase = Wtf8Buf::from_str(""); + uppercase.make_ascii_uppercase(); + assert_eq!(uppercase.bytes, b""); + + let mut uppercase = Wtf8Buf::from_str("GrEeN gRaPeS! š"); + uppercase.make_ascii_uppercase(); + assert_eq!(uppercase.bytes, b"GREEN GRAPES! \xf0\x9f\x8d\x87"); + + let mut uppercase = unsafe { Wtf8::from_bytes_unchecked(b"\xED\xA0\x80").to_owned() }; + uppercase.make_ascii_uppercase(); + assert_eq!(uppercase.bytes, b"\xED\xA0\x80"); + assert!(!uppercase.is_known_utf8); +} + +#[test] +fn wtf8_to_owned() { + let string = unsafe { Wtf8::from_bytes_unchecked(b"\xED\xA0\x80").to_owned() }; + assert_eq!(string.bytes, b"\xED\xA0\x80"); + assert!(!string.is_known_utf8); +} diff --git a/library/std/src/thread/local.rs b/library/std/src/thread/local.rs index f4750cdf7..8aedfc4a6 100644 --- a/library/std/src/thread/local.rs +++ b/library/std/src/thread/local.rs @@ -1036,6 +1036,7 @@ pub mod fast { } #[doc(hidden)] +#[cfg(not(target_thread_local))] pub mod os { use super::lazy::LazyKeyInner; use crate::cell::Cell; @@ -1044,6 +1045,8 @@ pub mod os { use crate::ptr; use crate::sys_common::thread_local_key::StaticKey as OsStaticKey; + /// Use a regular global static to store this key; the state provided will then be + /// thread-local. pub struct Key<T> { // OS-TLS key that we'll use to key off. os: OsStaticKey, diff --git a/library/std/src/thread/mod.rs b/library/std/src/thread/mod.rs index 44c8a50fd..ceea6986e 100644 --- a/library/std/src/thread/mod.rs +++ b/library/std/src/thread/mod.rs @@ -116,7 +116,7 @@ //! Threads are able to have associated names for identification purposes. By default, spawned //! threads are unnamed. To specify a name for a thread, build the thread with [`Builder`] and pass //! the desired thread name to [`Builder::name`]. To retrieve the thread name from within the -//! thread, use [`Thread::name`]. A couple examples of where the name of a thread gets used: +//! thread, use [`Thread::name`]. A couple of examples where the name of a thread gets used: //! //! * If a panic occurs in a named thread, the thread name will be printed in the panic message. //! * The thread name is provided to the OS where applicable (e.g., `pthread_setname_np` in @@ -170,7 +170,6 @@ use crate::ptr::addr_of_mut; use crate::str; use crate::sync::Arc; use crate::sys::thread as imp; -use crate::sys_common::mutex; use crate::sys_common::thread; use crate::sys_common::thread_info; use crate::sys_common::thread_parker::Parker; @@ -193,21 +192,31 @@ pub use scoped::{scope, Scope, ScopedJoinHandle}; #[stable(feature = "rust1", since = "1.0.0")] pub use self::local::{AccessError, LocalKey}; -// The types used by the thread_local! macro to access TLS keys. Note that there -// are two types, the "OS" type and the "fast" type. The OS thread local key -// type is accessed via platform-specific API calls and is slow, while the fast +// Select the type used by the thread_local! macro to access TLS keys. There +// are three types: "static", "fast", "OS". The "OS" thread local key +// type is accessed via platform-specific API calls and is slow, while the "fast" // key type is accessed via code generated via LLVM, where TLS keys are set up -// by the elf linker. Note that the OS TLS type is always available: on macOS -// the standard library is compiled with support for older platform versions -// where fast TLS was not available; end-user code is compiled with fast TLS -// where available, but both are needed. +// by the elf linker. "static" is for single-threaded platforms where a global +// static is sufficient. #[unstable(feature = "libstd_thread_internals", issue = "none")] #[cfg(target_thread_local)] +#[cfg(not(test))] #[doc(hidden)] pub use self::local::fast::Key as __FastLocalKeyInner; #[unstable(feature = "libstd_thread_internals", issue = "none")] +#[cfg(target_thread_local)] +#[cfg(test)] // when building for tests, use real std's key +pub use realstd::thread::__FastLocalKeyInner; + +#[unstable(feature = "libstd_thread_internals", issue = "none")] +#[cfg(target_thread_local)] +#[cfg(test)] +pub use self::local::fast::Key as __FastLocalKeyInnerUnused; // we import this anyway to silence 'unused' warnings + +#[unstable(feature = "libstd_thread_internals", issue = "none")] #[doc(hidden)] +#[cfg(not(target_thread_local))] pub use self::local::os::Key as __OsLocalKeyInner; #[unstable(feature = "libstd_thread_internals", issue = "none")] #[cfg(all(target_family = "wasm", not(target_feature = "atomics")))] @@ -1033,24 +1042,48 @@ pub struct ThreadId(NonZeroU64); impl ThreadId { // Generate a new unique thread ID. fn new() -> ThreadId { - // It is UB to attempt to acquire this mutex reentrantly! - static GUARD: mutex::StaticMutex = mutex::StaticMutex::new(); - static mut COUNTER: u64 = 1; - - unsafe { - let guard = GUARD.lock(); - - // If we somehow use up all our bits, panic so that we're not - // covering up subtle bugs of IDs being reused. - if COUNTER == u64::MAX { - drop(guard); // in case the panic handler ends up calling `ThreadId::new()`, avoid reentrant lock acquire. - panic!("failed to generate unique thread ID: bitspace exhausted"); - } - - let id = COUNTER; - COUNTER += 1; + #[cold] + fn exhausted() -> ! { + panic!("failed to generate unique thread ID: bitspace exhausted") + } - ThreadId(NonZeroU64::new(id).unwrap()) + cfg_if::cfg_if! { + if #[cfg(target_has_atomic = "64")] { + use crate::sync::atomic::{AtomicU64, Ordering::Relaxed}; + + static COUNTER: AtomicU64 = AtomicU64::new(0); + + let mut last = COUNTER.load(Relaxed); + loop { + let Some(id) = last.checked_add(1) else { + exhausted(); + }; + + match COUNTER.compare_exchange_weak(last, id, Relaxed, Relaxed) { + Ok(_) => return ThreadId(NonZeroU64::new(id).unwrap()), + Err(id) => last = id, + } + } + } else { + use crate::sys_common::mutex::StaticMutex; + + // It is UB to attempt to acquire this mutex reentrantly! + static GUARD: StaticMutex = StaticMutex::new(); + static mut COUNTER: u64 = 0; + + unsafe { + let guard = GUARD.lock(); + + let Some(id) = COUNTER.checked_add(1) else { + drop(guard); // in case the panic handler ends up calling `ThreadId::new()`, avoid reentrant lock acquire. + exhausted(); + }; + + COUNTER = id; + drop(guard); + ThreadId(NonZeroU64::new(id).unwrap()) + } + } } } diff --git a/library/std/src/thread/tests.rs b/library/std/src/thread/tests.rs index ec68b5291..777964f04 100644 --- a/library/std/src/thread/tests.rs +++ b/library/std/src/thread/tests.rs @@ -37,6 +37,37 @@ fn test_named_thread() { .unwrap(); } +#[cfg(any( + // Note: musl didn't add pthread_getname_np until 1.2.3 + all(target_os = "linux", target_env = "gnu"), + target_os = "macos", + target_os = "ios", + target_os = "watchos" +))] +#[test] +fn test_named_thread_truncation() { + use crate::ffi::CStr; + + let long_name = crate::iter::once("test_named_thread_truncation") + .chain(crate::iter::repeat(" yada").take(100)) + .collect::<String>(); + + let result = Builder::new().name(long_name.clone()).spawn(move || { + // Rust remembers the full thread name itself. + assert_eq!(thread::current().name(), Some(long_name.as_str())); + + // But the system is limited -- make sure we successfully set a truncation. + let mut buf = vec![0u8; long_name.len() + 1]; + unsafe { + libc::pthread_getname_np(libc::pthread_self(), buf.as_mut_ptr().cast(), buf.len()); + } + let cstr = CStr::from_bytes_until_nul(&buf).unwrap(); + assert!(cstr.to_bytes().len() > 0); + assert!(long_name.as_bytes().starts_with(cstr.to_bytes())); + }); + result.unwrap().join().unwrap(); +} + #[test] #[should_panic] fn test_invalid_named_thread() { @@ -329,3 +360,22 @@ fn test_scoped_threads_nll() { let x = 42_u8; foo(&x); } + +// Regression test for https://github.com/rust-lang/rust/issues/98498. +#[test] +#[cfg(miri)] // relies on Miri's data race detector +fn scope_join_race() { + for _ in 0..100 { + let a_bool = AtomicBool::new(false); + + thread::scope(|s| { + for _ in 0..5 { + s.spawn(|| a_bool.load(Ordering::Relaxed)); + } + + for _ in 0..5 { + s.spawn(|| a_bool.load(Ordering::Relaxed)); + } + }); + } +} diff --git a/library/std/src/time/tests.rs b/library/std/src/time/tests.rs index d710a5744..6229556c8 100644 --- a/library/std/src/time/tests.rs +++ b/library/std/src/time/tests.rs @@ -31,7 +31,8 @@ fn instant_monotonic_concurrent() -> crate::thread::Result<()> { .map(|_| { crate::thread::spawn(|| { let mut old = Instant::now(); - for _ in 0..5_000_000 { + let count = if cfg!(miri) { 1_000 } else { 5_000_000 }; + for _ in 0..count { let new = Instant::now(); assert!(new >= old); old = new; |