From 698f8c2f01ea549d77d7dc3338a12e04c11057b9 Mon Sep 17 00:00:00 2001 From: Daniel Baumann Date: Wed, 17 Apr 2024 14:02:58 +0200 Subject: Adding upstream version 1.64.0+dfsg1. Signed-off-by: Daniel Baumann --- library/core/src/mem/manually_drop.rs | 165 ++++ library/core/src/mem/maybe_uninit.rs | 1292 +++++++++++++++++++++++++++++++ library/core/src/mem/mod.rs | 1180 ++++++++++++++++++++++++++++ library/core/src/mem/transmutability.rs | 43 + library/core/src/mem/valid_align.rs | 247 ++++++ 5 files changed, 2927 insertions(+) create mode 100644 library/core/src/mem/manually_drop.rs create mode 100644 library/core/src/mem/maybe_uninit.rs create mode 100644 library/core/src/mem/mod.rs create mode 100644 library/core/src/mem/transmutability.rs create mode 100644 library/core/src/mem/valid_align.rs (limited to 'library/core/src/mem') diff --git a/library/core/src/mem/manually_drop.rs b/library/core/src/mem/manually_drop.rs new file mode 100644 index 000000000..3d719afe4 --- /dev/null +++ b/library/core/src/mem/manually_drop.rs @@ -0,0 +1,165 @@ +use crate::ops::{Deref, DerefMut}; +use crate::ptr; + +/// A wrapper to inhibit compiler from automatically calling `T`’s destructor. +/// This wrapper is 0-cost. +/// +/// `ManuallyDrop` is guaranteed to have the same layout as `T`, and is subject +/// to the same layout optimizations as `T`. As a consequence, it has *no effect* +/// on the assumptions that the compiler makes about its contents. For example, +/// initializing a `ManuallyDrop<&mut T>` with [`mem::zeroed`] is undefined +/// behavior. If you need to handle uninitialized data, use [`MaybeUninit`] +/// instead. +/// +/// Note that accessing the value inside a `ManuallyDrop` is safe. +/// This means that a `ManuallyDrop` whose content has been dropped must not +/// be exposed through a public safe API. +/// Correspondingly, `ManuallyDrop::drop` is unsafe. +/// +/// # `ManuallyDrop` and drop order. +/// +/// Rust has a well-defined [drop order] of values. To make sure that fields or +/// locals are dropped in a specific order, reorder the declarations such that +/// the implicit drop order is the correct one. +/// +/// It is possible to use `ManuallyDrop` to control the drop order, but this +/// requires unsafe code and is hard to do correctly in the presence of +/// unwinding. +/// +/// For example, if you want to make sure that a specific field is dropped after +/// the others, make it the last field of a struct: +/// +/// ``` +/// struct Context; +/// +/// struct Widget { +/// children: Vec, +/// // `context` will be dropped after `children`. +/// // Rust guarantees that fields are dropped in the order of declaration. +/// context: Context, +/// } +/// ``` +/// +/// [drop order]: https://doc.rust-lang.org/reference/destructors.html +/// [`mem::zeroed`]: crate::mem::zeroed +/// [`MaybeUninit`]: crate::mem::MaybeUninit +#[stable(feature = "manually_drop", since = "1.20.0")] +#[lang = "manually_drop"] +#[derive(Copy, Clone, Debug, Default, PartialEq, Eq, PartialOrd, Ord, Hash)] +#[repr(transparent)] +pub struct ManuallyDrop { + value: T, +} + +impl ManuallyDrop { + /// Wrap a value to be manually dropped. + /// + /// # Examples + /// + /// ```rust + /// use std::mem::ManuallyDrop; + /// let mut x = ManuallyDrop::new(String::from("Hello World!")); + /// x.truncate(5); // You can still safely operate on the value + /// assert_eq!(*x, "Hello"); + /// // But `Drop` will not be run here + /// ``` + #[must_use = "if you don't need the wrapper, you can use `mem::forget` instead"] + #[stable(feature = "manually_drop", since = "1.20.0")] + #[rustc_const_stable(feature = "const_manually_drop", since = "1.32.0")] + #[inline(always)] + pub const fn new(value: T) -> ManuallyDrop { + ManuallyDrop { value } + } + + /// Extracts the value from the `ManuallyDrop` container. + /// + /// This allows the value to be dropped again. + /// + /// # Examples + /// + /// ```rust + /// use std::mem::ManuallyDrop; + /// let x = ManuallyDrop::new(Box::new(())); + /// let _: Box<()> = ManuallyDrop::into_inner(x); // This drops the `Box`. + /// ``` + #[stable(feature = "manually_drop", since = "1.20.0")] + #[rustc_const_stable(feature = "const_manually_drop", since = "1.32.0")] + #[inline(always)] + pub const fn into_inner(slot: ManuallyDrop) -> T { + slot.value + } + + /// Takes the value from the `ManuallyDrop` container out. + /// + /// This method is primarily intended for moving out values in drop. + /// Instead of using [`ManuallyDrop::drop`] to manually drop the value, + /// you can use this method to take the value and use it however desired. + /// + /// Whenever possible, it is preferable to use [`into_inner`][`ManuallyDrop::into_inner`] + /// instead, which prevents duplicating the content of the `ManuallyDrop`. + /// + /// # Safety + /// + /// This function semantically moves out the contained value without preventing further usage, + /// leaving the state of this container unchanged. + /// It is your responsibility to ensure that this `ManuallyDrop` is not used again. + /// + #[must_use = "if you don't need the value, you can use `ManuallyDrop::drop` instead"] + #[stable(feature = "manually_drop_take", since = "1.42.0")] + #[inline] + pub unsafe fn take(slot: &mut ManuallyDrop) -> T { + // SAFETY: we are reading from a reference, which is guaranteed + // to be valid for reads. + unsafe { ptr::read(&slot.value) } + } +} + +impl ManuallyDrop { + /// Manually drops the contained value. This is exactly equivalent to calling + /// [`ptr::drop_in_place`] with a pointer to the contained value. As such, unless + /// the contained value is a packed struct, the destructor will be called in-place + /// without moving the value, and thus can be used to safely drop [pinned] data. + /// + /// If you have ownership of the value, you can use [`ManuallyDrop::into_inner`] instead. + /// + /// # Safety + /// + /// This function runs the destructor of the contained value. Other than changes made by + /// the destructor itself, the memory is left unchanged, and so as far as the compiler is + /// concerned still holds a bit-pattern which is valid for the type `T`. + /// + /// However, this "zombie" value should not be exposed to safe code, and this function + /// should not be called more than once. To use a value after it's been dropped, or drop + /// a value multiple times, can cause Undefined Behavior (depending on what `drop` does). + /// This is normally prevented by the type system, but users of `ManuallyDrop` must + /// uphold those guarantees without assistance from the compiler. + /// + /// [pinned]: crate::pin + #[stable(feature = "manually_drop", since = "1.20.0")] + #[inline] + pub unsafe fn drop(slot: &mut ManuallyDrop) { + // SAFETY: we are dropping the value pointed to by a mutable reference + // which is guaranteed to be valid for writes. + // It is up to the caller to make sure that `slot` isn't dropped again. + unsafe { ptr::drop_in_place(&mut slot.value) } + } +} + +#[stable(feature = "manually_drop", since = "1.20.0")] +#[rustc_const_unstable(feature = "const_deref", issue = "88955")] +impl const Deref for ManuallyDrop { + type Target = T; + #[inline(always)] + fn deref(&self) -> &T { + &self.value + } +} + +#[stable(feature = "manually_drop", since = "1.20.0")] +#[rustc_const_unstable(feature = "const_deref", issue = "88955")] +impl const DerefMut for ManuallyDrop { + #[inline(always)] + fn deref_mut(&mut self) -> &mut T { + &mut self.value + } +} diff --git a/library/core/src/mem/maybe_uninit.rs b/library/core/src/mem/maybe_uninit.rs new file mode 100644 index 000000000..b4ea53608 --- /dev/null +++ b/library/core/src/mem/maybe_uninit.rs @@ -0,0 +1,1292 @@ +use crate::any::type_name; +use crate::fmt; +use crate::intrinsics; +use crate::mem::{self, ManuallyDrop}; +use crate::ptr; +use crate::slice; + +/// A wrapper type to construct uninitialized instances of `T`. +/// +/// # Initialization invariant +/// +/// The compiler, in general, assumes that a variable is properly initialized +/// according to the requirements of the variable's type. For example, a variable of +/// reference type must be aligned and non-null. This is an invariant that must +/// *always* be upheld, even in unsafe code. As a consequence, zero-initializing a +/// variable of reference type causes instantaneous [undefined behavior][ub], +/// no matter whether that reference ever gets used to access memory: +/// +/// ```rust,no_run +/// # #![allow(invalid_value)] +/// use std::mem::{self, MaybeUninit}; +/// +/// let x: &i32 = unsafe { mem::zeroed() }; // undefined behavior! ⚠️ +/// // The equivalent code with `MaybeUninit<&i32>`: +/// let x: &i32 = unsafe { MaybeUninit::zeroed().assume_init() }; // undefined behavior! ⚠️ +/// ``` +/// +/// This is exploited by the compiler for various optimizations, such as eliding +/// run-time checks and optimizing `enum` layout. +/// +/// Similarly, entirely uninitialized memory may have any content, while a `bool` must +/// always be `true` or `false`. Hence, creating an uninitialized `bool` is undefined behavior: +/// +/// ```rust,no_run +/// # #![allow(invalid_value)] +/// use std::mem::{self, MaybeUninit}; +/// +/// let b: bool = unsafe { mem::uninitialized() }; // undefined behavior! ⚠️ +/// // The equivalent code with `MaybeUninit`: +/// let b: bool = unsafe { MaybeUninit::uninit().assume_init() }; // undefined behavior! ⚠️ +/// ``` +/// +/// Moreover, uninitialized memory is special in that it does not have a fixed value ("fixed" +/// meaning "it won't change without being written to"). Reading the same uninitialized byte +/// multiple times can give different results. This makes it undefined behavior to have +/// uninitialized data in a variable even if that variable has an integer type, which otherwise can +/// hold any *fixed* bit pattern: +/// +/// ```rust,no_run +/// # #![allow(invalid_value)] +/// use std::mem::{self, MaybeUninit}; +/// +/// let x: i32 = unsafe { mem::uninitialized() }; // undefined behavior! ⚠️ +/// // The equivalent code with `MaybeUninit`: +/// let x: i32 = unsafe { MaybeUninit::uninit().assume_init() }; // undefined behavior! ⚠️ +/// ``` +/// (Notice that the rules around uninitialized integers are not finalized yet, but +/// until they are, it is advisable to avoid them.) +/// +/// On top of that, remember that most types have additional invariants beyond merely +/// being considered initialized at the type level. For example, a `1`-initialized [`Vec`] +/// is considered initialized (under the current implementation; this does not constitute +/// a stable guarantee) because the only requirement the compiler knows about it +/// is that the data pointer must be non-null. Creating such a `Vec` does not cause +/// *immediate* undefined behavior, but will cause undefined behavior with most +/// safe operations (including dropping it). +/// +/// [`Vec`]: ../../std/vec/struct.Vec.html +/// +/// # Examples +/// +/// `MaybeUninit` serves to enable unsafe code to deal with uninitialized data. +/// It is a signal to the compiler indicating that the data here might *not* +/// be initialized: +/// +/// ```rust +/// use std::mem::MaybeUninit; +/// +/// // Create an explicitly uninitialized reference. The compiler knows that data inside +/// // a `MaybeUninit` may be invalid, and hence this is not UB: +/// let mut x = MaybeUninit::<&i32>::uninit(); +/// // Set it to a valid value. +/// x.write(&0); +/// // Extract the initialized data -- this is only allowed *after* properly +/// // initializing `x`! +/// let x = unsafe { x.assume_init() }; +/// ``` +/// +/// The compiler then knows to not make any incorrect assumptions or optimizations on this code. +/// +/// You can think of `MaybeUninit` as being a bit like `Option` but without +/// any of the run-time tracking and without any of the safety checks. +/// +/// ## out-pointers +/// +/// You can use `MaybeUninit` to implement "out-pointers": instead of returning data +/// from a function, pass it a pointer to some (uninitialized) memory to put the +/// result into. This can be useful when it is important for the caller to control +/// how the memory the result is stored in gets allocated, and you want to avoid +/// unnecessary moves. +/// +/// ``` +/// use std::mem::MaybeUninit; +/// +/// unsafe fn make_vec(out: *mut Vec) { +/// // `write` does not drop the old contents, which is important. +/// out.write(vec![1, 2, 3]); +/// } +/// +/// let mut v = MaybeUninit::uninit(); +/// unsafe { make_vec(v.as_mut_ptr()); } +/// // Now we know `v` is initialized! This also makes sure the vector gets +/// // properly dropped. +/// let v = unsafe { v.assume_init() }; +/// assert_eq!(&v, &[1, 2, 3]); +/// ``` +/// +/// ## Initializing an array element-by-element +/// +/// `MaybeUninit` can be used to initialize a large array element-by-element: +/// +/// ``` +/// use std::mem::{self, MaybeUninit}; +/// +/// let data = { +/// // Create an uninitialized array of `MaybeUninit`. The `assume_init` is +/// // safe because the type we are claiming to have initialized here is a +/// // bunch of `MaybeUninit`s, which do not require initialization. +/// let mut data: [MaybeUninit>; 1000] = unsafe { +/// MaybeUninit::uninit().assume_init() +/// }; +/// +/// // Dropping a `MaybeUninit` does nothing. Thus using raw pointer +/// // assignment instead of `ptr::write` does not cause the old +/// // uninitialized value to be dropped. Also if there is a panic during +/// // this loop, we have a memory leak, but there is no memory safety +/// // issue. +/// for elem in &mut data[..] { +/// elem.write(vec![42]); +/// } +/// +/// // Everything is initialized. Transmute the array to the +/// // initialized type. +/// unsafe { mem::transmute::<_, [Vec; 1000]>(data) } +/// }; +/// +/// assert_eq!(&data[0], &[42]); +/// ``` +/// +/// You can also work with partially initialized arrays, which could +/// be found in low-level datastructures. +/// +/// ``` +/// use std::mem::MaybeUninit; +/// use std::ptr; +/// +/// // Create an uninitialized array of `MaybeUninit`. The `assume_init` is +/// // safe because the type we are claiming to have initialized here is a +/// // bunch of `MaybeUninit`s, which do not require initialization. +/// let mut data: [MaybeUninit; 1000] = unsafe { MaybeUninit::uninit().assume_init() }; +/// // Count the number of elements we have assigned. +/// let mut data_len: usize = 0; +/// +/// for elem in &mut data[0..500] { +/// elem.write(String::from("hello")); +/// data_len += 1; +/// } +/// +/// // For each item in the array, drop if we allocated it. +/// for elem in &mut data[0..data_len] { +/// unsafe { ptr::drop_in_place(elem.as_mut_ptr()); } +/// } +/// ``` +/// +/// ## Initializing a struct field-by-field +/// +/// You can use `MaybeUninit`, and the [`std::ptr::addr_of_mut`] macro, to initialize structs field by field: +/// +/// ```rust +/// use std::mem::MaybeUninit; +/// use std::ptr::addr_of_mut; +/// +/// #[derive(Debug, PartialEq)] +/// pub struct Foo { +/// name: String, +/// list: Vec, +/// } +/// +/// let foo = { +/// let mut uninit: MaybeUninit = MaybeUninit::uninit(); +/// let ptr = uninit.as_mut_ptr(); +/// +/// // Initializing the `name` field +/// // Using `write` instead of assignment via `=` to not call `drop` on the +/// // old, uninitialized value. +/// unsafe { addr_of_mut!((*ptr).name).write("Bob".to_string()); } +/// +/// // Initializing the `list` field +/// // If there is a panic here, then the `String` in the `name` field leaks. +/// unsafe { addr_of_mut!((*ptr).list).write(vec![0, 1, 2]); } +/// +/// // All the fields are initialized, so we call `assume_init` to get an initialized Foo. +/// unsafe { uninit.assume_init() } +/// }; +/// +/// assert_eq!( +/// foo, +/// Foo { +/// name: "Bob".to_string(), +/// list: vec![0, 1, 2] +/// } +/// ); +/// ``` +/// [`std::ptr::addr_of_mut`]: crate::ptr::addr_of_mut +/// [ub]: ../../reference/behavior-considered-undefined.html +/// +/// # Layout +/// +/// `MaybeUninit` is guaranteed to have the same size, alignment, and ABI as `T`: +/// +/// ```rust +/// use std::mem::{MaybeUninit, size_of, align_of}; +/// assert_eq!(size_of::>(), size_of::()); +/// assert_eq!(align_of::>(), align_of::()); +/// ``` +/// +/// However remember that a type *containing* a `MaybeUninit` is not necessarily the same +/// layout; Rust does not in general guarantee that the fields of a `Foo` have the same order as +/// a `Foo` even if `T` and `U` have the same size and alignment. Furthermore because any bit +/// value is valid for a `MaybeUninit` the compiler can't apply non-zero/niche-filling +/// optimizations, potentially resulting in a larger size: +/// +/// ```rust +/// # use std::mem::{MaybeUninit, size_of}; +/// assert_eq!(size_of::>(), 1); +/// assert_eq!(size_of::>>(), 2); +/// ``` +/// +/// If `T` is FFI-safe, then so is `MaybeUninit`. +/// +/// While `MaybeUninit` is `#[repr(transparent)]` (indicating it guarantees the same size, +/// alignment, and ABI as `T`), this does *not* change any of the previous caveats. `Option` and +/// `Option>` may still have different sizes, and types containing a field of type +/// `T` may be laid out (and sized) differently than if that field were `MaybeUninit`. +/// `MaybeUninit` is a union type, and `#[repr(transparent)]` on unions is unstable (see [the +/// tracking issue](https://github.com/rust-lang/rust/issues/60405)). Over time, the exact +/// guarantees of `#[repr(transparent)]` on unions may evolve, and `MaybeUninit` may or may not +/// remain `#[repr(transparent)]`. That said, `MaybeUninit` will *always* guarantee that it has +/// the same size, alignment, and ABI as `T`; it's just that the way `MaybeUninit` implements that +/// guarantee may evolve. +#[stable(feature = "maybe_uninit", since = "1.36.0")] +// Lang item so we can wrap other types in it. This is useful for generators. +#[lang = "maybe_uninit"] +#[derive(Copy)] +#[repr(transparent)] +pub union MaybeUninit { + uninit: (), + value: ManuallyDrop, +} + +#[stable(feature = "maybe_uninit", since = "1.36.0")] +impl Clone for MaybeUninit { + #[inline(always)] + fn clone(&self) -> Self { + // Not calling `T::clone()`, we cannot know if we are initialized enough for that. + *self + } +} + +#[stable(feature = "maybe_uninit_debug", since = "1.41.0")] +impl fmt::Debug for MaybeUninit { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.pad(type_name::()) + } +} + +impl MaybeUninit { + /// Creates a new `MaybeUninit` initialized with the given value. + /// It is safe to call [`assume_init`] on the return value of this function. + /// + /// Note that dropping a `MaybeUninit` will never call `T`'s drop code. + /// It is your responsibility to make sure `T` gets dropped if it got initialized. + /// + /// # Example + /// + /// ``` + /// use std::mem::MaybeUninit; + /// + /// let v: MaybeUninit> = MaybeUninit::new(vec![42]); + /// ``` + /// + /// [`assume_init`]: MaybeUninit::assume_init + #[stable(feature = "maybe_uninit", since = "1.36.0")] + #[rustc_const_stable(feature = "const_maybe_uninit", since = "1.36.0")] + #[must_use = "use `forget` to avoid running Drop code"] + #[inline(always)] + pub const fn new(val: T) -> MaybeUninit { + MaybeUninit { value: ManuallyDrop::new(val) } + } + + /// Creates a new `MaybeUninit` in an uninitialized state. + /// + /// Note that dropping a `MaybeUninit` will never call `T`'s drop code. + /// It is your responsibility to make sure `T` gets dropped if it got initialized. + /// + /// See the [type-level documentation][MaybeUninit] for some examples. + /// + /// # Example + /// + /// ``` + /// use std::mem::MaybeUninit; + /// + /// let v: MaybeUninit = MaybeUninit::uninit(); + /// ``` + #[stable(feature = "maybe_uninit", since = "1.36.0")] + #[rustc_const_stable(feature = "const_maybe_uninit", since = "1.36.0")] + #[must_use] + #[inline(always)] + #[rustc_diagnostic_item = "maybe_uninit_uninit"] + pub const fn uninit() -> MaybeUninit { + MaybeUninit { uninit: () } + } + + /// Create a new array of `MaybeUninit` items, in an uninitialized state. + /// + /// Note: in a future Rust version this method may become unnecessary + /// when Rust allows + /// [inline const expressions](https://github.com/rust-lang/rust/issues/76001). + /// The example below could then use `let mut buf = [const { MaybeUninit::::uninit() }; 32];`. + /// + /// # Examples + /// + /// ```no_run + /// #![feature(maybe_uninit_uninit_array, maybe_uninit_slice)] + /// + /// use std::mem::MaybeUninit; + /// + /// extern "C" { + /// fn read_into_buffer(ptr: *mut u8, max_len: usize) -> usize; + /// } + /// + /// /// Returns a (possibly smaller) slice of data that was actually read + /// fn read(buf: &mut [MaybeUninit]) -> &[u8] { + /// unsafe { + /// let len = read_into_buffer(buf.as_mut_ptr() as *mut u8, buf.len()); + /// MaybeUninit::slice_assume_init_ref(&buf[..len]) + /// } + /// } + /// + /// let mut buf: [MaybeUninit; 32] = MaybeUninit::uninit_array(); + /// let data = read(&mut buf); + /// ``` + #[unstable(feature = "maybe_uninit_uninit_array", issue = "96097")] + #[rustc_const_unstable(feature = "const_maybe_uninit_uninit_array", issue = "96097")] + #[must_use] + #[inline(always)] + pub const fn uninit_array() -> [Self; N] { + // SAFETY: An uninitialized `[MaybeUninit<_>; LEN]` is valid. + unsafe { MaybeUninit::<[MaybeUninit; N]>::uninit().assume_init() } + } + + /// Creates a new `MaybeUninit` in an uninitialized state, with the memory being + /// filled with `0` bytes. It depends on `T` whether that already makes for + /// proper initialization. For example, `MaybeUninit::zeroed()` is initialized, + /// but `MaybeUninit<&'static i32>::zeroed()` is not because references must not + /// be null. + /// + /// Note that dropping a `MaybeUninit` will never call `T`'s drop code. + /// It is your responsibility to make sure `T` gets dropped if it got initialized. + /// + /// # Example + /// + /// Correct usage of this function: initializing a struct with zero, where all + /// fields of the struct can hold the bit-pattern 0 as a valid value. + /// + /// ```rust + /// use std::mem::MaybeUninit; + /// + /// let x = MaybeUninit::<(u8, bool)>::zeroed(); + /// let x = unsafe { x.assume_init() }; + /// assert_eq!(x, (0, false)); + /// ``` + /// + /// *Incorrect* usage of this function: calling `x.zeroed().assume_init()` + /// when `0` is not a valid bit-pattern for the type: + /// + /// ```rust,no_run + /// use std::mem::MaybeUninit; + /// + /// enum NotZero { One = 1, Two = 2 } + /// + /// let x = MaybeUninit::<(u8, NotZero)>::zeroed(); + /// let x = unsafe { x.assume_init() }; + /// // Inside a pair, we create a `NotZero` that does not have a valid discriminant. + /// // This is undefined behavior. ⚠️ + /// ``` + #[stable(feature = "maybe_uninit", since = "1.36.0")] + #[rustc_const_unstable(feature = "const_maybe_uninit_zeroed", issue = "91850")] + #[must_use] + #[inline] + #[rustc_diagnostic_item = "maybe_uninit_zeroed"] + pub const fn zeroed() -> MaybeUninit { + let mut u = MaybeUninit::::uninit(); + // SAFETY: `u.as_mut_ptr()` points to allocated memory. + unsafe { + u.as_mut_ptr().write_bytes(0u8, 1); + } + u + } + + /// Sets the value of the `MaybeUninit`. + /// + /// This overwrites any previous value without dropping it, so be careful + /// not to use this twice unless you want to skip running the destructor. + /// For your convenience, this also returns a mutable reference to the + /// (now safely initialized) contents of `self`. + /// + /// As the content is stored inside a `MaybeUninit`, the destructor is not + /// run for the inner data if the MaybeUninit leaves scope without a call to + /// [`assume_init`], [`assume_init_drop`], or similar. Code that receives + /// the mutable reference returned by this function needs to keep this in + /// mind. The safety model of Rust regards leaks as safe, but they are + /// usually still undesirable. This being said, the mutable reference + /// behaves like any other mutable reference would, so assigning a new value + /// to it will drop the old content. + /// + /// [`assume_init`]: Self::assume_init + /// [`assume_init_drop`]: Self::assume_init_drop + /// + /// # Examples + /// + /// Correct usage of this method: + /// + /// ```rust + /// use std::mem::MaybeUninit; + /// + /// let mut x = MaybeUninit::>::uninit(); + /// + /// { + /// let hello = x.write((&b"Hello, world!").to_vec()); + /// // Setting hello does not leak prior allocations, but drops them + /// *hello = (&b"Hello").to_vec(); + /// hello[0] = 'h' as u8; + /// } + /// // x is initialized now: + /// let s = unsafe { x.assume_init() }; + /// assert_eq!(b"hello", s.as_slice()); + /// ``` + /// + /// This usage of the method causes a leak: + /// + /// ```rust + /// use std::mem::MaybeUninit; + /// + /// let mut x = MaybeUninit::::uninit(); + /// + /// x.write("Hello".to_string()); + /// // This leaks the contained string: + /// x.write("hello".to_string()); + /// // x is initialized now: + /// let s = unsafe { x.assume_init() }; + /// ``` + /// + /// This method can be used to avoid unsafe in some cases. The example below + /// shows a part of an implementation of a fixed sized arena that lends out + /// pinned references. + /// With `write`, we can avoid the need to write through a raw pointer: + /// + /// ```rust + /// use core::pin::Pin; + /// use core::mem::MaybeUninit; + /// + /// struct PinArena { + /// memory: Box<[MaybeUninit]>, + /// len: usize, + /// } + /// + /// impl PinArena { + /// pub fn capacity(&self) -> usize { + /// self.memory.len() + /// } + /// pub fn push(&mut self, val: T) -> Pin<&mut T> { + /// if self.len >= self.capacity() { + /// panic!("Attempted to push to a full pin arena!"); + /// } + /// let ref_ = self.memory[self.len].write(val); + /// self.len += 1; + /// unsafe { Pin::new_unchecked(ref_) } + /// } + /// } + /// ``` + #[stable(feature = "maybe_uninit_write", since = "1.55.0")] + #[rustc_const_unstable(feature = "const_maybe_uninit_write", issue = "63567")] + #[inline(always)] + pub const fn write(&mut self, val: T) -> &mut T { + *self = MaybeUninit::new(val); + // SAFETY: We just initialized this value. + unsafe { self.assume_init_mut() } + } + + /// Gets a pointer to the contained value. Reading from this pointer or turning it + /// into a reference is undefined behavior unless the `MaybeUninit` is initialized. + /// Writing to memory that this pointer (non-transitively) points to is undefined behavior + /// (except inside an `UnsafeCell`). + /// + /// # Examples + /// + /// Correct usage of this method: + /// + /// ```rust + /// use std::mem::MaybeUninit; + /// + /// let mut x = MaybeUninit::>::uninit(); + /// x.write(vec![0, 1, 2]); + /// // Create a reference into the `MaybeUninit`. This is okay because we initialized it. + /// let x_vec = unsafe { &*x.as_ptr() }; + /// assert_eq!(x_vec.len(), 3); + /// ``` + /// + /// *Incorrect* usage of this method: + /// + /// ```rust,no_run + /// use std::mem::MaybeUninit; + /// + /// let x = MaybeUninit::>::uninit(); + /// let x_vec = unsafe { &*x.as_ptr() }; + /// // We have created a reference to an uninitialized vector! This is undefined behavior. ⚠️ + /// ``` + /// + /// (Notice that the rules around references to uninitialized data are not finalized yet, but + /// until they are, it is advisable to avoid them.) + #[stable(feature = "maybe_uninit", since = "1.36.0")] + #[rustc_const_stable(feature = "const_maybe_uninit_as_ptr", since = "1.59.0")] + #[inline(always)] + pub const fn as_ptr(&self) -> *const T { + // `MaybeUninit` and `ManuallyDrop` are both `repr(transparent)` so we can cast the pointer. + self as *const _ as *const T + } + + /// Gets a mutable pointer to the contained value. Reading from this pointer or turning it + /// into a reference is undefined behavior unless the `MaybeUninit` is initialized. + /// + /// # Examples + /// + /// Correct usage of this method: + /// + /// ```rust + /// use std::mem::MaybeUninit; + /// + /// let mut x = MaybeUninit::>::uninit(); + /// x.write(vec![0, 1, 2]); + /// // Create a reference into the `MaybeUninit>`. + /// // This is okay because we initialized it. + /// let x_vec = unsafe { &mut *x.as_mut_ptr() }; + /// x_vec.push(3); + /// assert_eq!(x_vec.len(), 4); + /// ``` + /// + /// *Incorrect* usage of this method: + /// + /// ```rust,no_run + /// use std::mem::MaybeUninit; + /// + /// let mut x = MaybeUninit::>::uninit(); + /// let x_vec = unsafe { &mut *x.as_mut_ptr() }; + /// // We have created a reference to an uninitialized vector! This is undefined behavior. ⚠️ + /// ``` + /// + /// (Notice that the rules around references to uninitialized data are not finalized yet, but + /// until they are, it is advisable to avoid them.) + #[stable(feature = "maybe_uninit", since = "1.36.0")] + #[rustc_const_unstable(feature = "const_maybe_uninit_as_mut_ptr", issue = "75251")] + #[inline(always)] + pub const fn as_mut_ptr(&mut self) -> *mut T { + // `MaybeUninit` and `ManuallyDrop` are both `repr(transparent)` so we can cast the pointer. + self as *mut _ as *mut T + } + + /// Extracts the value from the `MaybeUninit` container. This is a great way + /// to ensure that the data will get dropped, because the resulting `T` is + /// subject to the usual drop handling. + /// + /// # Safety + /// + /// It is up to the caller to guarantee that the `MaybeUninit` really is in an initialized + /// state. Calling this when the content is not yet fully initialized causes immediate undefined + /// behavior. The [type-level documentation][inv] contains more information about + /// this initialization invariant. + /// + /// [inv]: #initialization-invariant + /// + /// On top of that, remember that most types have additional invariants beyond merely + /// being considered initialized at the type level. For example, a `1`-initialized [`Vec`] + /// is considered initialized (under the current implementation; this does not constitute + /// a stable guarantee) because the only requirement the compiler knows about it + /// is that the data pointer must be non-null. Creating such a `Vec` does not cause + /// *immediate* undefined behavior, but will cause undefined behavior with most + /// safe operations (including dropping it). + /// + /// [`Vec`]: ../../std/vec/struct.Vec.html + /// + /// # Examples + /// + /// Correct usage of this method: + /// + /// ```rust + /// use std::mem::MaybeUninit; + /// + /// let mut x = MaybeUninit::::uninit(); + /// x.write(true); + /// let x_init = unsafe { x.assume_init() }; + /// assert_eq!(x_init, true); + /// ``` + /// + /// *Incorrect* usage of this method: + /// + /// ```rust,no_run + /// use std::mem::MaybeUninit; + /// + /// let x = MaybeUninit::>::uninit(); + /// let x_init = unsafe { x.assume_init() }; + /// // `x` had not been initialized yet, so this last line caused undefined behavior. ⚠️ + /// ``` + #[stable(feature = "maybe_uninit", since = "1.36.0")] + #[rustc_const_stable(feature = "const_maybe_uninit_assume_init_by_value", since = "1.59.0")] + #[inline(always)] + #[rustc_diagnostic_item = "assume_init"] + #[track_caller] + pub const unsafe fn assume_init(self) -> T { + // SAFETY: the caller must guarantee that `self` is initialized. + // This also means that `self` must be a `value` variant. + unsafe { + intrinsics::assert_inhabited::(); + ManuallyDrop::into_inner(self.value) + } + } + + /// Reads the value from the `MaybeUninit` container. The resulting `T` is subject + /// to the usual drop handling. + /// + /// Whenever possible, it is preferable to use [`assume_init`] instead, which + /// prevents duplicating the content of the `MaybeUninit`. + /// + /// # Safety + /// + /// It is up to the caller to guarantee that the `MaybeUninit` really is in an initialized + /// state. Calling this when the content is not yet fully initialized causes undefined + /// behavior. The [type-level documentation][inv] contains more information about + /// this initialization invariant. + /// + /// Moreover, similar to the [`ptr::read`] function, this function creates a + /// bitwise copy of the contents, regardless whether the contained type + /// implements the [`Copy`] trait or not. When using multiple copies of the + /// data (by calling `assume_init_read` multiple times, or first calling + /// `assume_init_read` and then [`assume_init`]), it is your responsibility + /// to ensure that that data may indeed be duplicated. + /// + /// [inv]: #initialization-invariant + /// [`assume_init`]: MaybeUninit::assume_init + /// + /// # Examples + /// + /// Correct usage of this method: + /// + /// ```rust + /// use std::mem::MaybeUninit; + /// + /// let mut x = MaybeUninit::::uninit(); + /// x.write(13); + /// let x1 = unsafe { x.assume_init_read() }; + /// // `u32` is `Copy`, so we may read multiple times. + /// let x2 = unsafe { x.assume_init_read() }; + /// assert_eq!(x1, x2); + /// + /// let mut x = MaybeUninit::>>::uninit(); + /// x.write(None); + /// let x1 = unsafe { x.assume_init_read() }; + /// // Duplicating a `None` value is okay, so we may read multiple times. + /// let x2 = unsafe { x.assume_init_read() }; + /// assert_eq!(x1, x2); + /// ``` + /// + /// *Incorrect* usage of this method: + /// + /// ```rust,no_run + /// use std::mem::MaybeUninit; + /// + /// let mut x = MaybeUninit::>>::uninit(); + /// x.write(Some(vec![0, 1, 2])); + /// let x1 = unsafe { x.assume_init_read() }; + /// let x2 = unsafe { x.assume_init_read() }; + /// // We now created two copies of the same vector, leading to a double-free ⚠️ when + /// // they both get dropped! + /// ``` + #[stable(feature = "maybe_uninit_extra", since = "1.60.0")] + #[rustc_const_unstable(feature = "const_maybe_uninit_assume_init_read", issue = "63567")] + #[inline(always)] + #[track_caller] + pub const unsafe fn assume_init_read(&self) -> T { + // SAFETY: the caller must guarantee that `self` is initialized. + // Reading from `self.as_ptr()` is safe since `self` should be initialized. + unsafe { + intrinsics::assert_inhabited::(); + self.as_ptr().read() + } + } + + /// Drops the contained value in place. + /// + /// If you have ownership of the `MaybeUninit`, you can also use + /// [`assume_init`] as an alternative. + /// + /// # Safety + /// + /// It is up to the caller to guarantee that the `MaybeUninit` really is + /// in an initialized state. Calling this when the content is not yet fully + /// initialized causes undefined behavior. + /// + /// On top of that, all additional invariants of the type `T` must be + /// satisfied, as the `Drop` implementation of `T` (or its members) may + /// rely on this. For example, setting a [`Vec`] to an invalid but + /// non-null address makes it initialized (under the current implementation; + /// this does not constitute a stable guarantee), because the only + /// requirement the compiler knows about it is that the data pointer must be + /// non-null. Dropping such a `Vec` however will cause undefined + /// behaviour. + /// + /// [`assume_init`]: MaybeUninit::assume_init + /// [`Vec`]: ../../std/vec/struct.Vec.html + #[stable(feature = "maybe_uninit_extra", since = "1.60.0")] + pub unsafe fn assume_init_drop(&mut self) { + // SAFETY: the caller must guarantee that `self` is initialized and + // satisfies all invariants of `T`. + // Dropping the value in place is safe if that is the case. + unsafe { ptr::drop_in_place(self.as_mut_ptr()) } + } + + /// Gets a shared reference to the contained value. + /// + /// This can be useful when we want to access a `MaybeUninit` that has been + /// initialized but don't have ownership of the `MaybeUninit` (preventing the use + /// of `.assume_init()`). + /// + /// # Safety + /// + /// Calling this when the content is not yet fully initialized causes undefined + /// behavior: it is up to the caller to guarantee that the `MaybeUninit` really + /// is in an initialized state. + /// + /// # Examples + /// + /// ### Correct usage of this method: + /// + /// ```rust + /// use std::mem::MaybeUninit; + /// + /// let mut x = MaybeUninit::>::uninit(); + /// // Initialize `x`: + /// x.write(vec![1, 2, 3]); + /// // Now that our `MaybeUninit<_>` is known to be initialized, it is okay to + /// // create a shared reference to it: + /// let x: &Vec = unsafe { + /// // SAFETY: `x` has been initialized. + /// x.assume_init_ref() + /// }; + /// assert_eq!(x, &vec![1, 2, 3]); + /// ``` + /// + /// ### *Incorrect* usages of this method: + /// + /// ```rust,no_run + /// use std::mem::MaybeUninit; + /// + /// let x = MaybeUninit::>::uninit(); + /// let x_vec: &Vec = unsafe { x.assume_init_ref() }; + /// // We have created a reference to an uninitialized vector! This is undefined behavior. ⚠️ + /// ``` + /// + /// ```rust,no_run + /// use std::{cell::Cell, mem::MaybeUninit}; + /// + /// let b = MaybeUninit::>::uninit(); + /// // Initialize the `MaybeUninit` using `Cell::set`: + /// unsafe { + /// b.assume_init_ref().set(true); + /// // ^^^^^^^^^^^^^^^ + /// // Reference to an uninitialized `Cell`: UB! + /// } + /// ``` + #[stable(feature = "maybe_uninit_ref", since = "1.55.0")] + #[rustc_const_stable(feature = "const_maybe_uninit_assume_init_ref", since = "1.59.0")] + #[inline(always)] + pub const unsafe fn assume_init_ref(&self) -> &T { + // SAFETY: the caller must guarantee that `self` is initialized. + // This also means that `self` must be a `value` variant. + unsafe { + intrinsics::assert_inhabited::(); + &*self.as_ptr() + } + } + + /// Gets a mutable (unique) reference to the contained value. + /// + /// This can be useful when we want to access a `MaybeUninit` that has been + /// initialized but don't have ownership of the `MaybeUninit` (preventing the use + /// of `.assume_init()`). + /// + /// # Safety + /// + /// Calling this when the content is not yet fully initialized causes undefined + /// behavior: it is up to the caller to guarantee that the `MaybeUninit` really + /// is in an initialized state. For instance, `.assume_init_mut()` cannot be used to + /// initialize a `MaybeUninit`. + /// + /// # Examples + /// + /// ### Correct usage of this method: + /// + /// ```rust + /// # #![allow(unexpected_cfgs)] + /// use std::mem::MaybeUninit; + /// + /// # unsafe extern "C" fn initialize_buffer(buf: *mut [u8; 1024]) { *buf = [0; 1024] } + /// # #[cfg(FALSE)] + /// extern "C" { + /// /// Initializes *all* the bytes of the input buffer. + /// fn initialize_buffer(buf: *mut [u8; 1024]); + /// } + /// + /// let mut buf = MaybeUninit::<[u8; 1024]>::uninit(); + /// + /// // Initialize `buf`: + /// unsafe { initialize_buffer(buf.as_mut_ptr()); } + /// // Now we know that `buf` has been initialized, so we could `.assume_init()` it. + /// // However, using `.assume_init()` may trigger a `memcpy` of the 1024 bytes. + /// // To assert our buffer has been initialized without copying it, we upgrade + /// // the `&mut MaybeUninit<[u8; 1024]>` to a `&mut [u8; 1024]`: + /// let buf: &mut [u8; 1024] = unsafe { + /// // SAFETY: `buf` has been initialized. + /// buf.assume_init_mut() + /// }; + /// + /// // Now we can use `buf` as a normal slice: + /// buf.sort_unstable(); + /// assert!( + /// buf.windows(2).all(|pair| pair[0] <= pair[1]), + /// "buffer is sorted", + /// ); + /// ``` + /// + /// ### *Incorrect* usages of this method: + /// + /// You cannot use `.assume_init_mut()` to initialize a value: + /// + /// ```rust,no_run + /// use std::mem::MaybeUninit; + /// + /// let mut b = MaybeUninit::::uninit(); + /// unsafe { + /// *b.assume_init_mut() = true; + /// // We have created a (mutable) reference to an uninitialized `bool`! + /// // This is undefined behavior. ⚠️ + /// } + /// ``` + /// + /// For instance, you cannot [`Read`] into an uninitialized buffer: + /// + /// [`Read`]: ../../std/io/trait.Read.html + /// + /// ```rust,no_run + /// use std::{io, mem::MaybeUninit}; + /// + /// fn read_chunk (reader: &'_ mut dyn io::Read) -> io::Result<[u8; 64]> + /// { + /// let mut buffer = MaybeUninit::<[u8; 64]>::uninit(); + /// reader.read_exact(unsafe { buffer.assume_init_mut() })?; + /// // ^^^^^^^^^^^^^^^^^^^^^^^^ + /// // (mutable) reference to uninitialized memory! + /// // This is undefined behavior. + /// Ok(unsafe { buffer.assume_init() }) + /// } + /// ``` + /// + /// Nor can you use direct field access to do field-by-field gradual initialization: + /// + /// ```rust,no_run + /// use std::{mem::MaybeUninit, ptr}; + /// + /// struct Foo { + /// a: u32, + /// b: u8, + /// } + /// + /// let foo: Foo = unsafe { + /// let mut foo = MaybeUninit::::uninit(); + /// ptr::write(&mut foo.assume_init_mut().a as *mut u32, 1337); + /// // ^^^^^^^^^^^^^^^^^^^^^ + /// // (mutable) reference to uninitialized memory! + /// // This is undefined behavior. + /// ptr::write(&mut foo.assume_init_mut().b as *mut u8, 42); + /// // ^^^^^^^^^^^^^^^^^^^^^ + /// // (mutable) reference to uninitialized memory! + /// // This is undefined behavior. + /// foo.assume_init() + /// }; + /// ``` + #[stable(feature = "maybe_uninit_ref", since = "1.55.0")] + #[rustc_const_unstable(feature = "const_maybe_uninit_assume_init", issue = "none")] + #[inline(always)] + pub const unsafe fn assume_init_mut(&mut self) -> &mut T { + // SAFETY: the caller must guarantee that `self` is initialized. + // This also means that `self` must be a `value` variant. + unsafe { + intrinsics::assert_inhabited::(); + &mut *self.as_mut_ptr() + } + } + + /// Extracts the values from an array of `MaybeUninit` containers. + /// + /// # Safety + /// + /// It is up to the caller to guarantee that all elements of the array are + /// in an initialized state. + /// + /// # Examples + /// + /// ``` + /// #![feature(maybe_uninit_uninit_array)] + /// #![feature(maybe_uninit_array_assume_init)] + /// use std::mem::MaybeUninit; + /// + /// let mut array: [MaybeUninit; 3] = MaybeUninit::uninit_array(); + /// array[0].write(0); + /// array[1].write(1); + /// array[2].write(2); + /// + /// // SAFETY: Now safe as we initialised all elements + /// let array = unsafe { + /// MaybeUninit::array_assume_init(array) + /// }; + /// + /// assert_eq!(array, [0, 1, 2]); + /// ``` + #[unstable(feature = "maybe_uninit_array_assume_init", issue = "96097")] + #[rustc_const_unstable(feature = "const_maybe_uninit_array_assume_init", issue = "96097")] + #[inline(always)] + #[track_caller] + pub const unsafe fn array_assume_init(array: [Self; N]) -> [T; N] { + // SAFETY: + // * The caller guarantees that all elements of the array are initialized + // * `MaybeUninit` and T are guaranteed to have the same layout + // * `MaybeUninit` does not drop, so there are no double-frees + // And thus the conversion is safe + let ret = unsafe { + intrinsics::assert_inhabited::<[T; N]>(); + (&array as *const _ as *const [T; N]).read() + }; + + // FIXME: required to avoid `~const Destruct` bound + super::forget(array); + ret + } + + /// Assuming all the elements are initialized, get a slice to them. + /// + /// # Safety + /// + /// It is up to the caller to guarantee that the `MaybeUninit` elements + /// really are in an initialized state. + /// Calling this when the content is not yet fully initialized causes undefined behavior. + /// + /// See [`assume_init_ref`] for more details and examples. + /// + /// [`assume_init_ref`]: MaybeUninit::assume_init_ref + #[unstable(feature = "maybe_uninit_slice", issue = "63569")] + #[rustc_const_unstable(feature = "maybe_uninit_slice", issue = "63569")] + #[inline(always)] + pub const unsafe fn slice_assume_init_ref(slice: &[Self]) -> &[T] { + // SAFETY: casting `slice` to a `*const [T]` is safe since the caller guarantees that + // `slice` is initialized, and `MaybeUninit` is guaranteed to have the same layout as `T`. + // The pointer obtained is valid since it refers to memory owned by `slice` which is a + // reference and thus guaranteed to be valid for reads. + unsafe { &*(slice as *const [Self] as *const [T]) } + } + + /// Assuming all the elements are initialized, get a mutable slice to them. + /// + /// # Safety + /// + /// It is up to the caller to guarantee that the `MaybeUninit` elements + /// really are in an initialized state. + /// Calling this when the content is not yet fully initialized causes undefined behavior. + /// + /// See [`assume_init_mut`] for more details and examples. + /// + /// [`assume_init_mut`]: MaybeUninit::assume_init_mut + #[unstable(feature = "maybe_uninit_slice", issue = "63569")] + #[rustc_const_unstable(feature = "const_maybe_uninit_assume_init", issue = "none")] + #[inline(always)] + pub const unsafe fn slice_assume_init_mut(slice: &mut [Self]) -> &mut [T] { + // SAFETY: similar to safety notes for `slice_get_ref`, but we have a + // mutable reference which is also guaranteed to be valid for writes. + unsafe { &mut *(slice as *mut [Self] as *mut [T]) } + } + + /// Gets a pointer to the first element of the array. + #[unstable(feature = "maybe_uninit_slice", issue = "63569")] + #[rustc_const_unstable(feature = "maybe_uninit_slice", issue = "63569")] + #[inline(always)] + pub const fn slice_as_ptr(this: &[MaybeUninit]) -> *const T { + this.as_ptr() as *const T + } + + /// Gets a mutable pointer to the first element of the array. + #[unstable(feature = "maybe_uninit_slice", issue = "63569")] + #[rustc_const_unstable(feature = "maybe_uninit_slice", issue = "63569")] + #[inline(always)] + pub const fn slice_as_mut_ptr(this: &mut [MaybeUninit]) -> *mut T { + this.as_mut_ptr() as *mut T + } + + /// Copies the elements from `src` to `this`, returning a mutable reference to the now initialized contents of `this`. + /// + /// If `T` does not implement `Copy`, use [`write_slice_cloned`] + /// + /// This is similar to [`slice::copy_from_slice`]. + /// + /// # Panics + /// + /// This function will panic if the two slices have different lengths. + /// + /// # Examples + /// + /// ``` + /// #![feature(maybe_uninit_write_slice)] + /// use std::mem::MaybeUninit; + /// + /// let mut dst = [MaybeUninit::uninit(); 32]; + /// let src = [0; 32]; + /// + /// let init = MaybeUninit::write_slice(&mut dst, &src); + /// + /// assert_eq!(init, src); + /// ``` + /// + /// ``` + /// #![feature(maybe_uninit_write_slice)] + /// use std::mem::MaybeUninit; + /// + /// let mut vec = Vec::with_capacity(32); + /// let src = [0; 16]; + /// + /// MaybeUninit::write_slice(&mut vec.spare_capacity_mut()[..src.len()], &src); + /// + /// // SAFETY: we have just copied all the elements of len into the spare capacity + /// // the first src.len() elements of the vec are valid now. + /// unsafe { + /// vec.set_len(src.len()); + /// } + /// + /// assert_eq!(vec, src); + /// ``` + /// + /// [`write_slice_cloned`]: MaybeUninit::write_slice_cloned + #[unstable(feature = "maybe_uninit_write_slice", issue = "79995")] + pub fn write_slice<'a>(this: &'a mut [MaybeUninit], src: &[T]) -> &'a mut [T] + where + T: Copy, + { + // SAFETY: &[T] and &[MaybeUninit] have the same layout + let uninit_src: &[MaybeUninit] = unsafe { super::transmute(src) }; + + this.copy_from_slice(uninit_src); + + // SAFETY: Valid elements have just been copied into `this` so it is initialized + unsafe { MaybeUninit::slice_assume_init_mut(this) } + } + + /// Clones the elements from `src` to `this`, returning a mutable reference to the now initialized contents of `this`. + /// Any already initialized elements will not be dropped. + /// + /// If `T` implements `Copy`, use [`write_slice`] + /// + /// This is similar to [`slice::clone_from_slice`] but does not drop existing elements. + /// + /// # Panics + /// + /// This function will panic if the two slices have different lengths, or if the implementation of `Clone` panics. + /// + /// If there is a panic, the already cloned elements will be dropped. + /// + /// # Examples + /// + /// ``` + /// #![feature(maybe_uninit_write_slice)] + /// use std::mem::MaybeUninit; + /// + /// let mut dst = [MaybeUninit::uninit(), MaybeUninit::uninit(), MaybeUninit::uninit(), MaybeUninit::uninit(), MaybeUninit::uninit()]; + /// let src = ["wibbly".to_string(), "wobbly".to_string(), "timey".to_string(), "wimey".to_string(), "stuff".to_string()]; + /// + /// let init = MaybeUninit::write_slice_cloned(&mut dst, &src); + /// + /// assert_eq!(init, src); + /// ``` + /// + /// ``` + /// #![feature(maybe_uninit_write_slice)] + /// use std::mem::MaybeUninit; + /// + /// let mut vec = Vec::with_capacity(32); + /// let src = ["rust", "is", "a", "pretty", "cool", "language"]; + /// + /// MaybeUninit::write_slice_cloned(&mut vec.spare_capacity_mut()[..src.len()], &src); + /// + /// // SAFETY: we have just cloned all the elements of len into the spare capacity + /// // the first src.len() elements of the vec are valid now. + /// unsafe { + /// vec.set_len(src.len()); + /// } + /// + /// assert_eq!(vec, src); + /// ``` + /// + /// [`write_slice`]: MaybeUninit::write_slice + #[unstable(feature = "maybe_uninit_write_slice", issue = "79995")] + pub fn write_slice_cloned<'a>(this: &'a mut [MaybeUninit], src: &[T]) -> &'a mut [T] + where + T: Clone, + { + // unlike copy_from_slice this does not call clone_from_slice on the slice + // this is because `MaybeUninit` does not implement Clone. + + struct Guard<'a, T> { + slice: &'a mut [MaybeUninit], + initialized: usize, + } + + impl<'a, T> Drop for Guard<'a, T> { + fn drop(&mut self) { + let initialized_part = &mut self.slice[..self.initialized]; + // SAFETY: this raw slice will contain only initialized objects + // that's why, it is allowed to drop it. + unsafe { + crate::ptr::drop_in_place(MaybeUninit::slice_assume_init_mut(initialized_part)); + } + } + } + + assert_eq!(this.len(), src.len(), "destination and source slices have different lengths"); + // NOTE: We need to explicitly slice them to the same length + // for bounds checking to be elided, and the optimizer will + // generate memcpy for simple cases (for example T = u8). + let len = this.len(); + let src = &src[..len]; + + // guard is needed b/c panic might happen during a clone + let mut guard = Guard { slice: this, initialized: 0 }; + + for i in 0..len { + guard.slice[i].write(src[i].clone()); + guard.initialized += 1; + } + + super::forget(guard); + + // SAFETY: Valid elements have just been written into `this` so it is initialized + unsafe { MaybeUninit::slice_assume_init_mut(this) } + } + + /// Returns the contents of this `MaybeUninit` as a slice of potentially uninitialized bytes. + /// + /// Note that even if the contents of a `MaybeUninit` have been initialized, the value may still + /// contain padding bytes which are left uninitialized. + /// + /// # Examples + /// + /// ``` + /// #![feature(maybe_uninit_as_bytes, maybe_uninit_slice)] + /// use std::mem::MaybeUninit; + /// + /// let val = 0x12345678i32; + /// let uninit = MaybeUninit::new(val); + /// let uninit_bytes = uninit.as_bytes(); + /// let bytes = unsafe { MaybeUninit::slice_assume_init_ref(uninit_bytes) }; + /// assert_eq!(bytes, val.to_ne_bytes()); + /// ``` + #[unstable(feature = "maybe_uninit_as_bytes", issue = "93092")] + pub fn as_bytes(&self) -> &[MaybeUninit] { + // SAFETY: MaybeUninit is always valid, even for padding bytes + unsafe { + slice::from_raw_parts(self.as_ptr() as *const MaybeUninit, mem::size_of::()) + } + } + + /// Returns the contents of this `MaybeUninit` as a mutable slice of potentially uninitialized + /// bytes. + /// + /// Note that even if the contents of a `MaybeUninit` have been initialized, the value may still + /// contain padding bytes which are left uninitialized. + /// + /// # Examples + /// + /// ``` + /// #![feature(maybe_uninit_as_bytes)] + /// use std::mem::MaybeUninit; + /// + /// let val = 0x12345678i32; + /// let mut uninit = MaybeUninit::new(val); + /// let uninit_bytes = uninit.as_bytes_mut(); + /// if cfg!(target_endian = "little") { + /// uninit_bytes[0].write(0xcd); + /// } else { + /// uninit_bytes[3].write(0xcd); + /// } + /// let val2 = unsafe { uninit.assume_init() }; + /// assert_eq!(val2, 0x123456cd); + /// ``` + #[unstable(feature = "maybe_uninit_as_bytes", issue = "93092")] + pub fn as_bytes_mut(&mut self) -> &mut [MaybeUninit] { + // SAFETY: MaybeUninit is always valid, even for padding bytes + unsafe { + slice::from_raw_parts_mut( + self.as_mut_ptr() as *mut MaybeUninit, + mem::size_of::(), + ) + } + } + + /// Returns the contents of this slice of `MaybeUninit` as a slice of potentially uninitialized + /// bytes. + /// + /// Note that even if the contents of a `MaybeUninit` have been initialized, the value may still + /// contain padding bytes which are left uninitialized. + /// + /// # Examples + /// + /// ``` + /// #![feature(maybe_uninit_as_bytes, maybe_uninit_write_slice, maybe_uninit_slice)] + /// use std::mem::MaybeUninit; + /// + /// let uninit = [MaybeUninit::new(0x1234u16), MaybeUninit::new(0x5678u16)]; + /// let uninit_bytes = MaybeUninit::slice_as_bytes(&uninit); + /// let bytes = unsafe { MaybeUninit::slice_assume_init_ref(&uninit_bytes) }; + /// let val1 = u16::from_ne_bytes(bytes[0..2].try_into().unwrap()); + /// let val2 = u16::from_ne_bytes(bytes[2..4].try_into().unwrap()); + /// assert_eq!(&[val1, val2], &[0x1234u16, 0x5678u16]); + /// ``` + #[unstable(feature = "maybe_uninit_as_bytes", issue = "93092")] + pub fn slice_as_bytes(this: &[MaybeUninit]) -> &[MaybeUninit] { + // SAFETY: MaybeUninit is always valid, even for padding bytes + unsafe { + slice::from_raw_parts( + this.as_ptr() as *const MaybeUninit, + this.len() * mem::size_of::(), + ) + } + } + + /// Returns the contents of this mutable slice of `MaybeUninit` as a mutable slice of + /// potentially uninitialized bytes. + /// + /// Note that even if the contents of a `MaybeUninit` have been initialized, the value may still + /// contain padding bytes which are left uninitialized. + /// + /// # Examples + /// + /// ``` + /// #![feature(maybe_uninit_as_bytes, maybe_uninit_write_slice, maybe_uninit_slice)] + /// use std::mem::MaybeUninit; + /// + /// let mut uninit = [MaybeUninit::::uninit(), MaybeUninit::::uninit()]; + /// let uninit_bytes = MaybeUninit::slice_as_bytes_mut(&mut uninit); + /// MaybeUninit::write_slice(uninit_bytes, &[0x12, 0x34, 0x56, 0x78]); + /// let vals = unsafe { MaybeUninit::slice_assume_init_ref(&uninit) }; + /// if cfg!(target_endian = "little") { + /// assert_eq!(vals, &[0x3412u16, 0x7856u16]); + /// } else { + /// assert_eq!(vals, &[0x1234u16, 0x5678u16]); + /// } + /// ``` + #[unstable(feature = "maybe_uninit_as_bytes", issue = "93092")] + pub fn slice_as_bytes_mut(this: &mut [MaybeUninit]) -> &mut [MaybeUninit] { + // SAFETY: MaybeUninit is always valid, even for padding bytes + unsafe { + slice::from_raw_parts_mut( + this.as_mut_ptr() as *mut MaybeUninit, + this.len() * mem::size_of::(), + ) + } + } +} diff --git a/library/core/src/mem/mod.rs b/library/core/src/mem/mod.rs new file mode 100644 index 000000000..20b2d5e26 --- /dev/null +++ b/library/core/src/mem/mod.rs @@ -0,0 +1,1180 @@ +//! Basic functions for dealing with memory. +//! +//! This module contains functions for querying the size and alignment of +//! types, initializing and manipulating memory. + +#![stable(feature = "rust1", since = "1.0.0")] + +use crate::clone; +use crate::cmp; +use crate::fmt; +use crate::hash; +use crate::intrinsics; +use crate::marker::{Copy, DiscriminantKind, Sized}; +use crate::ptr; + +mod manually_drop; +#[stable(feature = "manually_drop", since = "1.20.0")] +pub use manually_drop::ManuallyDrop; + +mod maybe_uninit; +#[stable(feature = "maybe_uninit", since = "1.36.0")] +pub use maybe_uninit::MaybeUninit; + +mod valid_align; +// For now this type is left crate-local. It could potentially make sense to expose +// it publicly, as it would be a nice parameter type for methods which need to take +// alignment as a parameter, such as `Layout::padding_needed_for`. +pub(crate) use valid_align::ValidAlign; + +mod transmutability; +#[unstable(feature = "transmutability", issue = "99571")] +pub use transmutability::{Assume, BikeshedIntrinsicFrom}; + +#[stable(feature = "rust1", since = "1.0.0")] +#[doc(inline)] +pub use crate::intrinsics::transmute; + +/// Takes ownership and "forgets" about the value **without running its destructor**. +/// +/// Any resources the value manages, such as heap memory or a file handle, will linger +/// forever in an unreachable state. However, it does not guarantee that pointers +/// to this memory will remain valid. +/// +/// * If you want to leak memory, see [`Box::leak`]. +/// * If you want to obtain a raw pointer to the memory, see [`Box::into_raw`]. +/// * If you want to dispose of a value properly, running its destructor, see +/// [`mem::drop`]. +/// +/// # Safety +/// +/// `forget` is not marked as `unsafe`, because Rust's safety guarantees +/// do not include a guarantee that destructors will always run. For example, +/// a program can create a reference cycle using [`Rc`][rc], or call +/// [`process::exit`][exit] to exit without running destructors. Thus, allowing +/// `mem::forget` from safe code does not fundamentally change Rust's safety +/// guarantees. +/// +/// That said, leaking resources such as memory or I/O objects is usually undesirable. +/// The need comes up in some specialized use cases for FFI or unsafe code, but even +/// then, [`ManuallyDrop`] is typically preferred. +/// +/// Because forgetting a value is allowed, any `unsafe` code you write must +/// allow for this possibility. You cannot return a value and expect that the +/// caller will necessarily run the value's destructor. +/// +/// [rc]: ../../std/rc/struct.Rc.html +/// [exit]: ../../std/process/fn.exit.html +/// +/// # Examples +/// +/// The canonical safe use of `mem::forget` is to circumvent a value's destructor +/// implemented by the `Drop` trait. For example, this will leak a `File`, i.e. reclaim +/// the space taken by the variable but never close the underlying system resource: +/// +/// ```no_run +/// use std::mem; +/// use std::fs::File; +/// +/// let file = File::open("foo.txt").unwrap(); +/// mem::forget(file); +/// ``` +/// +/// This is useful when the ownership of the underlying resource was previously +/// transferred to code outside of Rust, for example by transmitting the raw +/// file descriptor to C code. +/// +/// # Relationship with `ManuallyDrop` +/// +/// While `mem::forget` can also be used to transfer *memory* ownership, doing so is error-prone. +/// [`ManuallyDrop`] should be used instead. Consider, for example, this code: +/// +/// ``` +/// use std::mem; +/// +/// let mut v = vec![65, 122]; +/// // Build a `String` using the contents of `v` +/// let s = unsafe { String::from_raw_parts(v.as_mut_ptr(), v.len(), v.capacity()) }; +/// // leak `v` because its memory is now managed by `s` +/// mem::forget(v); // ERROR - v is invalid and must not be passed to a function +/// assert_eq!(s, "Az"); +/// // `s` is implicitly dropped and its memory deallocated. +/// ``` +/// +/// There are two issues with the above example: +/// +/// * If more code were added between the construction of `String` and the invocation of +/// `mem::forget()`, a panic within it would cause a double free because the same memory +/// is handled by both `v` and `s`. +/// * After calling `v.as_mut_ptr()` and transmitting the ownership of the data to `s`, +/// the `v` value is invalid. Even when a value is just moved to `mem::forget` (which won't +/// inspect it), some types have strict requirements on their values that +/// make them invalid when dangling or no longer owned. Using invalid values in any +/// way, including passing them to or returning them from functions, constitutes +/// undefined behavior and may break the assumptions made by the compiler. +/// +/// Switching to `ManuallyDrop` avoids both issues: +/// +/// ``` +/// use std::mem::ManuallyDrop; +/// +/// let v = vec![65, 122]; +/// // Before we disassemble `v` into its raw parts, make sure it +/// // does not get dropped! +/// let mut v = ManuallyDrop::new(v); +/// // Now disassemble `v`. These operations cannot panic, so there cannot be a leak. +/// let (ptr, len, cap) = (v.as_mut_ptr(), v.len(), v.capacity()); +/// // Finally, build a `String`. +/// let s = unsafe { String::from_raw_parts(ptr, len, cap) }; +/// assert_eq!(s, "Az"); +/// // `s` is implicitly dropped and its memory deallocated. +/// ``` +/// +/// `ManuallyDrop` robustly prevents double-free because we disable `v`'s destructor +/// before doing anything else. `mem::forget()` doesn't allow this because it consumes its +/// argument, forcing us to call it only after extracting anything we need from `v`. Even +/// if a panic were introduced between construction of `ManuallyDrop` and building the +/// string (which cannot happen in the code as shown), it would result in a leak and not a +/// double free. In other words, `ManuallyDrop` errs on the side of leaking instead of +/// erring on the side of (double-)dropping. +/// +/// Also, `ManuallyDrop` prevents us from having to "touch" `v` after transferring the +/// ownership to `s` — the final step of interacting with `v` to dispose of it without +/// running its destructor is entirely avoided. +/// +/// [`Box`]: ../../std/boxed/struct.Box.html +/// [`Box::leak`]: ../../std/boxed/struct.Box.html#method.leak +/// [`Box::into_raw`]: ../../std/boxed/struct.Box.html#method.into_raw +/// [`mem::drop`]: drop +/// [ub]: ../../reference/behavior-considered-undefined.html +#[inline] +#[rustc_const_stable(feature = "const_forget", since = "1.46.0")] +#[stable(feature = "rust1", since = "1.0.0")] +#[cfg_attr(not(test), rustc_diagnostic_item = "mem_forget")] +pub const fn forget(t: T) { + let _ = ManuallyDrop::new(t); +} + +/// Like [`forget`], but also accepts unsized values. +/// +/// This function is just a shim intended to be removed when the `unsized_locals` feature gets +/// stabilized. +#[inline] +#[unstable(feature = "forget_unsized", issue = "none")] +pub fn forget_unsized(t: T) { + intrinsics::forget(t) +} + +/// Returns the size of a type in bytes. +/// +/// More specifically, this is the offset in bytes between successive elements +/// in an array with that item type including alignment padding. Thus, for any +/// type `T` and length `n`, `[T; n]` has a size of `n * size_of::()`. +/// +/// In general, the size of a type is not stable across compilations, but +/// specific types such as primitives are. +/// +/// The following table gives the size for primitives. +/// +/// Type | size_of::\() +/// ---- | --------------- +/// () | 0 +/// bool | 1 +/// u8 | 1 +/// u16 | 2 +/// u32 | 4 +/// u64 | 8 +/// u128 | 16 +/// i8 | 1 +/// i16 | 2 +/// i32 | 4 +/// i64 | 8 +/// i128 | 16 +/// f32 | 4 +/// f64 | 8 +/// char | 4 +/// +/// Furthermore, `usize` and `isize` have the same size. +/// +/// The types `*const T`, `&T`, `Box`, `Option<&T>`, and `Option>` all have +/// the same size. If `T` is Sized, all of those types have the same size as `usize`. +/// +/// The mutability of a pointer does not change its size. As such, `&T` and `&mut T` +/// have the same size. Likewise for `*const T` and `*mut T`. +/// +/// # Size of `#[repr(C)]` items +/// +/// The `C` representation for items has a defined layout. With this layout, +/// the size of items is also stable as long as all fields have a stable size. +/// +/// ## Size of Structs +/// +/// For `structs`, the size is determined by the following algorithm. +/// +/// For each field in the struct ordered by declaration order: +/// +/// 1. Add the size of the field. +/// 2. Round up the current size to the nearest multiple of the next field's [alignment]. +/// +/// Finally, round the size of the struct to the nearest multiple of its [alignment]. +/// The alignment of the struct is usually the largest alignment of all its +/// fields; this can be changed with the use of `repr(align(N))`. +/// +/// Unlike `C`, zero sized structs are not rounded up to one byte in size. +/// +/// ## Size of Enums +/// +/// Enums that carry no data other than the discriminant have the same size as C enums +/// on the platform they are compiled for. +/// +/// ## Size of Unions +/// +/// The size of a union is the size of its largest field. +/// +/// Unlike `C`, zero sized unions are not rounded up to one byte in size. +/// +/// # Examples +/// +/// ``` +/// use std::mem; +/// +/// // Some primitives +/// assert_eq!(4, mem::size_of::()); +/// assert_eq!(8, mem::size_of::()); +/// assert_eq!(0, mem::size_of::<()>()); +/// +/// // Some arrays +/// assert_eq!(8, mem::size_of::<[i32; 2]>()); +/// assert_eq!(12, mem::size_of::<[i32; 3]>()); +/// assert_eq!(0, mem::size_of::<[i32; 0]>()); +/// +/// +/// // Pointer size equality +/// assert_eq!(mem::size_of::<&i32>(), mem::size_of::<*const i32>()); +/// assert_eq!(mem::size_of::<&i32>(), mem::size_of::>()); +/// assert_eq!(mem::size_of::<&i32>(), mem::size_of::>()); +/// assert_eq!(mem::size_of::>(), mem::size_of::>>()); +/// ``` +/// +/// Using `#[repr(C)]`. +/// +/// ``` +/// use std::mem; +/// +/// #[repr(C)] +/// struct FieldStruct { +/// first: u8, +/// second: u16, +/// third: u8 +/// } +/// +/// // The size of the first field is 1, so add 1 to the size. Size is 1. +/// // The alignment of the second field is 2, so add 1 to the size for padding. Size is 2. +/// // The size of the second field is 2, so add 2 to the size. Size is 4. +/// // The alignment of the third field is 1, so add 0 to the size for padding. Size is 4. +/// // The size of the third field is 1, so add 1 to the size. Size is 5. +/// // Finally, the alignment of the struct is 2 (because the largest alignment amongst its +/// // fields is 2), so add 1 to the size for padding. Size is 6. +/// assert_eq!(6, mem::size_of::()); +/// +/// #[repr(C)] +/// struct TupleStruct(u8, u16, u8); +/// +/// // Tuple structs follow the same rules. +/// assert_eq!(6, mem::size_of::()); +/// +/// // Note that reordering the fields can lower the size. We can remove both padding bytes +/// // by putting `third` before `second`. +/// #[repr(C)] +/// struct FieldStructOptimized { +/// first: u8, +/// third: u8, +/// second: u16 +/// } +/// +/// assert_eq!(4, mem::size_of::()); +/// +/// // Union size is the size of the largest field. +/// #[repr(C)] +/// union ExampleUnion { +/// smaller: u8, +/// larger: u16 +/// } +/// +/// assert_eq!(2, mem::size_of::()); +/// ``` +/// +/// [alignment]: align_of +#[inline(always)] +#[must_use] +#[stable(feature = "rust1", since = "1.0.0")] +#[rustc_promotable] +#[rustc_const_stable(feature = "const_mem_size_of", since = "1.24.0")] +#[cfg_attr(not(test), rustc_diagnostic_item = "mem_size_of")] +pub const fn size_of() -> usize { + intrinsics::size_of::() +} + +/// Returns the size of the pointed-to value in bytes. +/// +/// This is usually the same as `size_of::()`. However, when `T` *has* no +/// statically-known size, e.g., a slice [`[T]`][slice] or a [trait object], +/// then `size_of_val` can be used to get the dynamically-known size. +/// +/// [trait object]: ../../book/ch17-02-trait-objects.html +/// +/// # Examples +/// +/// ``` +/// use std::mem; +/// +/// assert_eq!(4, mem::size_of_val(&5i32)); +/// +/// let x: [u8; 13] = [0; 13]; +/// let y: &[u8] = &x; +/// assert_eq!(13, mem::size_of_val(y)); +/// ``` +#[inline] +#[must_use] +#[stable(feature = "rust1", since = "1.0.0")] +#[rustc_const_unstable(feature = "const_size_of_val", issue = "46571")] +#[cfg_attr(not(test), rustc_diagnostic_item = "mem_size_of_val")] +pub const fn size_of_val(val: &T) -> usize { + // SAFETY: `val` is a reference, so it's a valid raw pointer + unsafe { intrinsics::size_of_val(val) } +} + +/// Returns the size of the pointed-to value in bytes. +/// +/// This is usually the same as `size_of::()`. However, when `T` *has* no +/// statically-known size, e.g., a slice [`[T]`][slice] or a [trait object], +/// then `size_of_val_raw` can be used to get the dynamically-known size. +/// +/// # Safety +/// +/// This function is only safe to call if the following conditions hold: +/// +/// - If `T` is `Sized`, this function is always safe to call. +/// - If the unsized tail of `T` is: +/// - a [slice], then the length of the slice tail must be an initialized +/// integer, and the size of the *entire value* +/// (dynamic tail length + statically sized prefix) must fit in `isize`. +/// - a [trait object], then the vtable part of the pointer must point +/// to a valid vtable acquired by an unsizing coercion, and the size +/// of the *entire value* (dynamic tail length + statically sized prefix) +/// must fit in `isize`. +/// - an (unstable) [extern type], then this function is always safe to +/// call, but may panic or otherwise return the wrong value, as the +/// extern type's layout is not known. This is the same behavior as +/// [`size_of_val`] on a reference to a type with an extern type tail. +/// - otherwise, it is conservatively not allowed to call this function. +/// +/// [trait object]: ../../book/ch17-02-trait-objects.html +/// [extern type]: ../../unstable-book/language-features/extern-types.html +/// +/// # Examples +/// +/// ``` +/// #![feature(layout_for_ptr)] +/// use std::mem; +/// +/// assert_eq!(4, mem::size_of_val(&5i32)); +/// +/// let x: [u8; 13] = [0; 13]; +/// let y: &[u8] = &x; +/// assert_eq!(13, unsafe { mem::size_of_val_raw(y) }); +/// ``` +#[inline] +#[must_use] +#[unstable(feature = "layout_for_ptr", issue = "69835")] +#[rustc_const_unstable(feature = "const_size_of_val_raw", issue = "46571")] +pub const unsafe fn size_of_val_raw(val: *const T) -> usize { + // SAFETY: the caller must provide a valid raw pointer + unsafe { intrinsics::size_of_val(val) } +} + +/// Returns the [ABI]-required minimum alignment of a type in bytes. +/// +/// Every reference to a value of the type `T` must be a multiple of this number. +/// +/// This is the alignment used for struct fields. It may be smaller than the preferred alignment. +/// +/// [ABI]: https://en.wikipedia.org/wiki/Application_binary_interface +/// +/// # Examples +/// +/// ``` +/// # #![allow(deprecated)] +/// use std::mem; +/// +/// assert_eq!(4, mem::min_align_of::()); +/// ``` +#[inline] +#[must_use] +#[stable(feature = "rust1", since = "1.0.0")] +#[deprecated(note = "use `align_of` instead", since = "1.2.0")] +pub fn min_align_of() -> usize { + intrinsics::min_align_of::() +} + +/// Returns the [ABI]-required minimum alignment of the type of the value that `val` points to in +/// bytes. +/// +/// Every reference to a value of the type `T` must be a multiple of this number. +/// +/// [ABI]: https://en.wikipedia.org/wiki/Application_binary_interface +/// +/// # Examples +/// +/// ``` +/// # #![allow(deprecated)] +/// use std::mem; +/// +/// assert_eq!(4, mem::min_align_of_val(&5i32)); +/// ``` +#[inline] +#[must_use] +#[stable(feature = "rust1", since = "1.0.0")] +#[deprecated(note = "use `align_of_val` instead", since = "1.2.0")] +pub fn min_align_of_val(val: &T) -> usize { + // SAFETY: val is a reference, so it's a valid raw pointer + unsafe { intrinsics::min_align_of_val(val) } +} + +/// Returns the [ABI]-required minimum alignment of a type in bytes. +/// +/// Every reference to a value of the type `T` must be a multiple of this number. +/// +/// This is the alignment used for struct fields. It may be smaller than the preferred alignment. +/// +/// [ABI]: https://en.wikipedia.org/wiki/Application_binary_interface +/// +/// # Examples +/// +/// ``` +/// use std::mem; +/// +/// assert_eq!(4, mem::align_of::()); +/// ``` +#[inline(always)] +#[must_use] +#[stable(feature = "rust1", since = "1.0.0")] +#[rustc_promotable] +#[rustc_const_stable(feature = "const_align_of", since = "1.24.0")] +pub const fn align_of() -> usize { + intrinsics::min_align_of::() +} + +/// Returns the [ABI]-required minimum alignment of the type of the value that `val` points to in +/// bytes. +/// +/// Every reference to a value of the type `T` must be a multiple of this number. +/// +/// [ABI]: https://en.wikipedia.org/wiki/Application_binary_interface +/// +/// # Examples +/// +/// ``` +/// use std::mem; +/// +/// assert_eq!(4, mem::align_of_val(&5i32)); +/// ``` +#[inline] +#[must_use] +#[stable(feature = "rust1", since = "1.0.0")] +#[rustc_const_unstable(feature = "const_align_of_val", issue = "46571")] +#[allow(deprecated)] +pub const fn align_of_val(val: &T) -> usize { + // SAFETY: val is a reference, so it's a valid raw pointer + unsafe { intrinsics::min_align_of_val(val) } +} + +/// Returns the [ABI]-required minimum alignment of the type of the value that `val` points to in +/// bytes. +/// +/// Every reference to a value of the type `T` must be a multiple of this number. +/// +/// [ABI]: https://en.wikipedia.org/wiki/Application_binary_interface +/// +/// # Safety +/// +/// This function is only safe to call if the following conditions hold: +/// +/// - If `T` is `Sized`, this function is always safe to call. +/// - If the unsized tail of `T` is: +/// - a [slice], then the length of the slice tail must be an initialized +/// integer, and the size of the *entire value* +/// (dynamic tail length + statically sized prefix) must fit in `isize`. +/// - a [trait object], then the vtable part of the pointer must point +/// to a valid vtable acquired by an unsizing coercion, and the size +/// of the *entire value* (dynamic tail length + statically sized prefix) +/// must fit in `isize`. +/// - an (unstable) [extern type], then this function is always safe to +/// call, but may panic or otherwise return the wrong value, as the +/// extern type's layout is not known. This is the same behavior as +/// [`align_of_val`] on a reference to a type with an extern type tail. +/// - otherwise, it is conservatively not allowed to call this function. +/// +/// [trait object]: ../../book/ch17-02-trait-objects.html +/// [extern type]: ../../unstable-book/language-features/extern-types.html +/// +/// # Examples +/// +/// ``` +/// #![feature(layout_for_ptr)] +/// use std::mem; +/// +/// assert_eq!(4, unsafe { mem::align_of_val_raw(&5i32) }); +/// ``` +#[inline] +#[must_use] +#[unstable(feature = "layout_for_ptr", issue = "69835")] +#[rustc_const_unstable(feature = "const_align_of_val_raw", issue = "46571")] +pub const unsafe fn align_of_val_raw(val: *const T) -> usize { + // SAFETY: the caller must provide a valid raw pointer + unsafe { intrinsics::min_align_of_val(val) } +} + +/// Returns `true` if dropping values of type `T` matters. +/// +/// This is purely an optimization hint, and may be implemented conservatively: +/// it may return `true` for types that don't actually need to be dropped. +/// As such always returning `true` would be a valid implementation of +/// this function. However if this function actually returns `false`, then you +/// can be certain dropping `T` has no side effect. +/// +/// Low level implementations of things like collections, which need to manually +/// drop their data, should use this function to avoid unnecessarily +/// trying to drop all their contents when they are destroyed. This might not +/// make a difference in release builds (where a loop that has no side-effects +/// is easily detected and eliminated), but is often a big win for debug builds. +/// +/// Note that [`drop_in_place`] already performs this check, so if your workload +/// can be reduced to some small number of [`drop_in_place`] calls, using this is +/// unnecessary. In particular note that you can [`drop_in_place`] a slice, and that +/// will do a single needs_drop check for all the values. +/// +/// Types like Vec therefore just `drop_in_place(&mut self[..])` without using +/// `needs_drop` explicitly. Types like [`HashMap`], on the other hand, have to drop +/// values one at a time and should use this API. +/// +/// [`drop_in_place`]: crate::ptr::drop_in_place +/// [`HashMap`]: ../../std/collections/struct.HashMap.html +/// +/// # Examples +/// +/// Here's an example of how a collection might make use of `needs_drop`: +/// +/// ``` +/// use std::{mem, ptr}; +/// +/// pub struct MyCollection { +/// # data: [T; 1], +/// /* ... */ +/// } +/// # impl MyCollection { +/// # fn iter_mut(&mut self) -> &mut [T] { &mut self.data } +/// # fn free_buffer(&mut self) {} +/// # } +/// +/// impl Drop for MyCollection { +/// fn drop(&mut self) { +/// unsafe { +/// // drop the data +/// if mem::needs_drop::() { +/// for x in self.iter_mut() { +/// ptr::drop_in_place(x); +/// } +/// } +/// self.free_buffer(); +/// } +/// } +/// } +/// ``` +#[inline] +#[must_use] +#[stable(feature = "needs_drop", since = "1.21.0")] +#[rustc_const_stable(feature = "const_mem_needs_drop", since = "1.36.0")] +#[rustc_diagnostic_item = "needs_drop"] +pub const fn needs_drop() -> bool { + intrinsics::needs_drop::() +} + +/// Returns the value of type `T` represented by the all-zero byte-pattern. +/// +/// This means that, for example, the padding byte in `(u8, u16)` is not +/// necessarily zeroed. +/// +/// There is no guarantee that an all-zero byte-pattern represents a valid value +/// of some type `T`. For example, the all-zero byte-pattern is not a valid value +/// for reference types (`&T`, `&mut T`) and functions pointers. Using `zeroed` +/// on such types causes immediate [undefined behavior][ub] because [the Rust +/// compiler assumes][inv] that there always is a valid value in a variable it +/// considers initialized. +/// +/// This has the same effect as [`MaybeUninit::zeroed().assume_init()`][zeroed]. +/// It is useful for FFI sometimes, but should generally be avoided. +/// +/// [zeroed]: MaybeUninit::zeroed +/// [ub]: ../../reference/behavior-considered-undefined.html +/// [inv]: MaybeUninit#initialization-invariant +/// +/// # Examples +/// +/// Correct usage of this function: initializing an integer with zero. +/// +/// ``` +/// use std::mem; +/// +/// let x: i32 = unsafe { mem::zeroed() }; +/// assert_eq!(0, x); +/// ``` +/// +/// *Incorrect* usage of this function: initializing a reference with zero. +/// +/// ```rust,no_run +/// # #![allow(invalid_value)] +/// use std::mem; +/// +/// let _x: &i32 = unsafe { mem::zeroed() }; // Undefined behavior! +/// let _y: fn() = unsafe { mem::zeroed() }; // And again! +/// ``` +#[inline(always)] +#[must_use] +#[stable(feature = "rust1", since = "1.0.0")] +#[allow(deprecated_in_future)] +#[allow(deprecated)] +#[rustc_diagnostic_item = "mem_zeroed"] +#[track_caller] +pub unsafe fn zeroed() -> T { + // SAFETY: the caller must guarantee that an all-zero value is valid for `T`. + unsafe { + intrinsics::assert_zero_valid::(); + MaybeUninit::zeroed().assume_init() + } +} + +/// Bypasses Rust's normal memory-initialization checks by pretending to +/// produce a value of type `T`, while doing nothing at all. +/// +/// **This function is deprecated.** Use [`MaybeUninit`] instead. +/// It also might be slower than using `MaybeUninit` due to mitigations that were put in place to +/// limit the potential harm caused by incorrect use of this function in legacy code. +/// +/// The reason for deprecation is that the function basically cannot be used +/// correctly: it has the same effect as [`MaybeUninit::uninit().assume_init()`][uninit]. +/// As the [`assume_init` documentation][assume_init] explains, +/// [the Rust compiler assumes][inv] that values are properly initialized. +/// As a consequence, calling e.g. `mem::uninitialized::()` causes immediate +/// undefined behavior for returning a `bool` that is not definitely either `true` +/// or `false`. Worse, truly uninitialized memory like what gets returned here +/// is special in that the compiler knows that it does not have a fixed value. +/// This makes it undefined behavior to have uninitialized data in a variable even +/// if that variable has an integer type. +/// (Notice that the rules around uninitialized integers are not finalized yet, but +/// until they are, it is advisable to avoid them.) +/// +/// [uninit]: MaybeUninit::uninit +/// [assume_init]: MaybeUninit::assume_init +/// [inv]: MaybeUninit#initialization-invariant +#[inline(always)] +#[must_use] +#[deprecated(since = "1.39.0", note = "use `mem::MaybeUninit` instead")] +#[stable(feature = "rust1", since = "1.0.0")] +#[allow(deprecated_in_future)] +#[allow(deprecated)] +#[rustc_diagnostic_item = "mem_uninitialized"] +#[track_caller] +pub unsafe fn uninitialized() -> T { + // SAFETY: the caller must guarantee that an uninitialized value is valid for `T`. + unsafe { + intrinsics::assert_uninit_valid::(); + let mut val = MaybeUninit::::uninit(); + + // Fill memory with 0x01, as an imperfect mitigation for old code that uses this function on + // bool, nonnull, and noundef types. But don't do this if we actively want to detect UB. + if !cfg!(any(miri, sanitize = "memory")) { + val.as_mut_ptr().write_bytes(0x01, 1); + } + + val.assume_init() + } +} + +/// Swaps the values at two mutable locations, without deinitializing either one. +/// +/// * If you want to swap with a default or dummy value, see [`take`]. +/// * If you want to swap with a passed value, returning the old value, see [`replace`]. +/// +/// # Examples +/// +/// ``` +/// use std::mem; +/// +/// let mut x = 5; +/// let mut y = 42; +/// +/// mem::swap(&mut x, &mut y); +/// +/// assert_eq!(42, x); +/// assert_eq!(5, y); +/// ``` +#[inline] +#[stable(feature = "rust1", since = "1.0.0")] +#[rustc_const_unstable(feature = "const_swap", issue = "83163")] +pub const fn swap(x: &mut T, y: &mut T) { + // NOTE(eddyb) SPIR-V's Logical addressing model doesn't allow for arbitrary + // reinterpretation of values as (chunkable) byte arrays, and the loop in the + // block optimization in `swap_slice` is hard to rewrite back + // into the (unoptimized) direct swapping implementation, so we disable it. + // FIXME(eddyb) the block optimization also prevents MIR optimizations from + // understanding `mem::replace`, `Option::take`, etc. - a better overall + // solution might be to make `ptr::swap_nonoverlapping` into an intrinsic, which + // a backend can choose to implement using the block optimization, or not. + // NOTE(scottmcm) MIRI is disabled here as reading in smaller units is a + // pessimization for it. Also, if the type contains any unaligned pointers, + // copying those over multiple reads is difficult to support. + #[cfg(not(any(target_arch = "spirv", miri)))] + { + // For types that are larger multiples of their alignment, the simple way + // tends to copy the whole thing to stack rather than doing it one part + // at a time, so instead treat them as one-element slices and piggy-back + // the slice optimizations that will split up the swaps. + if size_of::() / align_of::() > 4 { + // SAFETY: exclusive references always point to one non-overlapping + // element and are non-null and properly aligned. + return unsafe { ptr::swap_nonoverlapping(x, y, 1) }; + } + } + + // If a scalar consists of just a small number of alignment units, let + // the codegen just swap those pieces directly, as it's likely just a + // few instructions and anything else is probably overcomplicated. + // + // Most importantly, this covers primitives and simd types that tend to + // have size=align where doing anything else can be a pessimization. + // (This will also be used for ZSTs, though any solution works for them.) + swap_simple(x, y); +} + +/// Same as [`swap`] semantically, but always uses the simple implementation. +/// +/// Used elsewhere in `mem` and `ptr` at the bottom layer of calls. +#[rustc_const_unstable(feature = "const_swap", issue = "83163")] +#[inline] +pub(crate) const fn swap_simple(x: &mut T, y: &mut T) { + // We arrange for this to typically be called with small types, + // so this reads-and-writes approach is actually better than using + // copy_nonoverlapping as it easily puts things in LLVM registers + // directly and doesn't end up inlining allocas. + // And LLVM actually optimizes it to 3×memcpy if called with + // a type larger than it's willing to keep in a register. + // Having typed reads and writes in MIR here is also good as + // it lets MIRI and CTFE understand them better, including things + // like enforcing type validity for them. + // Importantly, read+copy_nonoverlapping+write introduces confusing + // asymmetry to the behaviour where one value went through read+write + // whereas the other was copied over by the intrinsic (see #94371). + + // SAFETY: exclusive references are always valid to read/write, + // including being aligned, and nothing here panics so it's drop-safe. + unsafe { + let a = ptr::read(x); + let b = ptr::read(y); + ptr::write(x, b); + ptr::write(y, a); + } +} + +/// Replaces `dest` with the default value of `T`, returning the previous `dest` value. +/// +/// * If you want to replace the values of two variables, see [`swap`]. +/// * If you want to replace with a passed value instead of the default value, see [`replace`]. +/// +/// # Examples +/// +/// A simple example: +/// +/// ``` +/// use std::mem; +/// +/// let mut v: Vec = vec![1, 2]; +/// +/// let old_v = mem::take(&mut v); +/// assert_eq!(vec![1, 2], old_v); +/// assert!(v.is_empty()); +/// ``` +/// +/// `take` allows taking ownership of a struct field by replacing it with an "empty" value. +/// Without `take` you can run into issues like these: +/// +/// ```compile_fail,E0507 +/// struct Buffer { buf: Vec } +/// +/// impl Buffer { +/// fn get_and_reset(&mut self) -> Vec { +/// // error: cannot move out of dereference of `&mut`-pointer +/// let buf = self.buf; +/// self.buf = Vec::new(); +/// buf +/// } +/// } +/// ``` +/// +/// Note that `T` does not necessarily implement [`Clone`], so it can't even clone and reset +/// `self.buf`. But `take` can be used to disassociate the original value of `self.buf` from +/// `self`, allowing it to be returned: +/// +/// ``` +/// use std::mem; +/// +/// # struct Buffer { buf: Vec } +/// impl Buffer { +/// fn get_and_reset(&mut self) -> Vec { +/// mem::take(&mut self.buf) +/// } +/// } +/// +/// let mut buffer = Buffer { buf: vec![0, 1] }; +/// assert_eq!(buffer.buf.len(), 2); +/// +/// assert_eq!(buffer.get_and_reset(), vec![0, 1]); +/// assert_eq!(buffer.buf.len(), 0); +/// ``` +#[inline] +#[stable(feature = "mem_take", since = "1.40.0")] +pub fn take(dest: &mut T) -> T { + replace(dest, T::default()) +} + +/// Moves `src` into the referenced `dest`, returning the previous `dest` value. +/// +/// Neither value is dropped. +/// +/// * If you want to replace the values of two variables, see [`swap`]. +/// * If you want to replace with a default value, see [`take`]. +/// +/// # Examples +/// +/// A simple example: +/// +/// ``` +/// use std::mem; +/// +/// let mut v: Vec = vec![1, 2]; +/// +/// let old_v = mem::replace(&mut v, vec![3, 4, 5]); +/// assert_eq!(vec![1, 2], old_v); +/// assert_eq!(vec![3, 4, 5], v); +/// ``` +/// +/// `replace` allows consumption of a struct field by replacing it with another value. +/// Without `replace` you can run into issues like these: +/// +/// ```compile_fail,E0507 +/// struct Buffer { buf: Vec } +/// +/// impl Buffer { +/// fn replace_index(&mut self, i: usize, v: T) -> T { +/// // error: cannot move out of dereference of `&mut`-pointer +/// let t = self.buf[i]; +/// self.buf[i] = v; +/// t +/// } +/// } +/// ``` +/// +/// Note that `T` does not necessarily implement [`Clone`], so we can't even clone `self.buf[i]` to +/// avoid the move. But `replace` can be used to disassociate the original value at that index from +/// `self`, allowing it to be returned: +/// +/// ``` +/// # #![allow(dead_code)] +/// use std::mem; +/// +/// # struct Buffer { buf: Vec } +/// impl Buffer { +/// fn replace_index(&mut self, i: usize, v: T) -> T { +/// mem::replace(&mut self.buf[i], v) +/// } +/// } +/// +/// let mut buffer = Buffer { buf: vec![0, 1] }; +/// assert_eq!(buffer.buf[0], 0); +/// +/// assert_eq!(buffer.replace_index(0, 2), 0); +/// assert_eq!(buffer.buf[0], 2); +/// ``` +#[inline] +#[stable(feature = "rust1", since = "1.0.0")] +#[must_use = "if you don't need the old value, you can just assign the new value directly"] +#[rustc_const_unstable(feature = "const_replace", issue = "83164")] +#[cfg_attr(not(test), rustc_diagnostic_item = "mem_replace")] +pub const fn replace(dest: &mut T, src: T) -> T { + // SAFETY: We read from `dest` but directly write `src` into it afterwards, + // such that the old value is not duplicated. Nothing is dropped and + // nothing here can panic. + unsafe { + let result = ptr::read(dest); + ptr::write(dest, src); + result + } +} + +/// Disposes of a value. +/// +/// This does so by calling the argument's implementation of [`Drop`][drop]. +/// +/// This effectively does nothing for types which implement `Copy`, e.g. +/// integers. Such values are copied and _then_ moved into the function, so the +/// value persists after this function call. +/// +/// This function is not magic; it is literally defined as +/// +/// ``` +/// pub fn drop(_x: T) { } +/// ``` +/// +/// Because `_x` is moved into the function, it is automatically dropped before +/// the function returns. +/// +/// [drop]: Drop +/// +/// # Examples +/// +/// Basic usage: +/// +/// ``` +/// let v = vec![1, 2, 3]; +/// +/// drop(v); // explicitly drop the vector +/// ``` +/// +/// Since [`RefCell`] enforces the borrow rules at runtime, `drop` can +/// release a [`RefCell`] borrow: +/// +/// ``` +/// use std::cell::RefCell; +/// +/// let x = RefCell::new(1); +/// +/// let mut mutable_borrow = x.borrow_mut(); +/// *mutable_borrow = 1; +/// +/// drop(mutable_borrow); // relinquish the mutable borrow on this slot +/// +/// let borrow = x.borrow(); +/// println!("{}", *borrow); +/// ``` +/// +/// Integers and other types implementing [`Copy`] are unaffected by `drop`. +/// +/// ``` +/// #[derive(Copy, Clone)] +/// struct Foo(u8); +/// +/// let x = 1; +/// let y = Foo(2); +/// drop(x); // a copy of `x` is moved and dropped +/// drop(y); // a copy of `y` is moved and dropped +/// +/// println!("x: {}, y: {}", x, y.0); // still available +/// ``` +/// +/// [`RefCell`]: crate::cell::RefCell +#[inline] +#[stable(feature = "rust1", since = "1.0.0")] +#[cfg_attr(not(test), rustc_diagnostic_item = "mem_drop")] +pub fn drop(_x: T) {} + +/// Bitwise-copies a value. +/// +/// This function is not magic; it is literally defined as +/// ``` +/// pub fn copy(x: &T) -> T { *x } +/// ``` +/// +/// It is useful when you want to pass a function pointer to a combinator, rather than defining a new closure. +/// +/// Example: +/// ``` +/// #![feature(mem_copy_fn)] +/// use core::mem::copy; +/// let result_from_ffi_function: Result<(), &i32> = Err(&1); +/// let result_copied: Result<(), i32> = result_from_ffi_function.map_err(copy); +/// ``` +#[inline] +#[unstable(feature = "mem_copy_fn", issue = "98262")] +pub fn copy(x: &T) -> T { + *x +} + +/// Interprets `src` as having type `&U`, and then reads `src` without moving +/// the contained value. +/// +/// This function will unsafely assume the pointer `src` is valid for [`size_of::`][size_of] +/// bytes by transmuting `&T` to `&U` and then reading the `&U` (except that this is done in a way +/// that is correct even when `&U` has stricter alignment requirements than `&T`). It will also +/// unsafely create a copy of the contained value instead of moving out of `src`. +/// +/// It is not a compile-time error if `T` and `U` have different sizes, but it +/// is highly encouraged to only invoke this function where `T` and `U` have the +/// same size. This function triggers [undefined behavior][ub] if `U` is larger than +/// `T`. +/// +/// [ub]: ../../reference/behavior-considered-undefined.html +/// +/// # Examples +/// +/// ``` +/// use std::mem; +/// +/// #[repr(packed)] +/// struct Foo { +/// bar: u8, +/// } +/// +/// let foo_array = [10u8]; +/// +/// unsafe { +/// // Copy the data from 'foo_array' and treat it as a 'Foo' +/// let mut foo_struct: Foo = mem::transmute_copy(&foo_array); +/// assert_eq!(foo_struct.bar, 10); +/// +/// // Modify the copied data +/// foo_struct.bar = 20; +/// assert_eq!(foo_struct.bar, 20); +/// } +/// +/// // The contents of 'foo_array' should not have changed +/// assert_eq!(foo_array, [10]); +/// ``` +#[inline] +#[must_use] +#[stable(feature = "rust1", since = "1.0.0")] +#[rustc_const_unstable(feature = "const_transmute_copy", issue = "83165")] +pub const unsafe fn transmute_copy(src: &T) -> U { + assert!(size_of::() >= size_of::(), "cannot transmute_copy if U is larger than T"); + + // If U has a higher alignment requirement, src might not be suitably aligned. + if align_of::() > align_of::() { + // SAFETY: `src` is a reference which is guaranteed to be valid for reads. + // The caller must guarantee that the actual transmutation is safe. + unsafe { ptr::read_unaligned(src as *const T as *const U) } + } else { + // SAFETY: `src` is a reference which is guaranteed to be valid for reads. + // We just checked that `src as *const U` was properly aligned. + // The caller must guarantee that the actual transmutation is safe. + unsafe { ptr::read(src as *const T as *const U) } + } +} + +/// Opaque type representing the discriminant of an enum. +/// +/// See the [`discriminant`] function in this module for more information. +#[stable(feature = "discriminant_value", since = "1.21.0")] +pub struct Discriminant(::Discriminant); + +// N.B. These trait implementations cannot be derived because we don't want any bounds on T. + +#[stable(feature = "discriminant_value", since = "1.21.0")] +impl Copy for Discriminant {} + +#[stable(feature = "discriminant_value", since = "1.21.0")] +impl clone::Clone for Discriminant { + fn clone(&self) -> Self { + *self + } +} + +#[stable(feature = "discriminant_value", since = "1.21.0")] +impl cmp::PartialEq for Discriminant { + fn eq(&self, rhs: &Self) -> bool { + self.0 == rhs.0 + } +} + +#[stable(feature = "discriminant_value", since = "1.21.0")] +impl cmp::Eq for Discriminant {} + +#[stable(feature = "discriminant_value", since = "1.21.0")] +impl hash::Hash for Discriminant { + fn hash(&self, state: &mut H) { + self.0.hash(state); + } +} + +#[stable(feature = "discriminant_value", since = "1.21.0")] +impl fmt::Debug for Discriminant { + fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result { + fmt.debug_tuple("Discriminant").field(&self.0).finish() + } +} + +/// Returns a value uniquely identifying the enum variant in `v`. +/// +/// If `T` is not an enum, calling this function will not result in undefined behavior, but the +/// return value is unspecified. +/// +/// # Stability +/// +/// The discriminant of an enum variant may change if the enum definition changes. A discriminant +/// of some variant will not change between compilations with the same compiler. +/// +/// # Examples +/// +/// This can be used to compare enums that carry data, while disregarding +/// the actual data: +/// +/// ``` +/// use std::mem; +/// +/// enum Foo { A(&'static str), B(i32), C(i32) } +/// +/// assert_eq!(mem::discriminant(&Foo::A("bar")), mem::discriminant(&Foo::A("baz"))); +/// assert_eq!(mem::discriminant(&Foo::B(1)), mem::discriminant(&Foo::B(2))); +/// assert_ne!(mem::discriminant(&Foo::B(3)), mem::discriminant(&Foo::C(3))); +/// ``` +#[stable(feature = "discriminant_value", since = "1.21.0")] +#[rustc_const_unstable(feature = "const_discriminant", issue = "69821")] +#[cfg_attr(not(test), rustc_diagnostic_item = "mem_discriminant")] +#[cfg_attr(miri, track_caller)] // even without panics, this helps for Miri backtraces +pub const fn discriminant(v: &T) -> Discriminant { + Discriminant(intrinsics::discriminant_value(v)) +} + +/// Returns the number of variants in the enum type `T`. +/// +/// If `T` is not an enum, calling this function will not result in undefined behavior, but the +/// return value is unspecified. Equally, if `T` is an enum with more variants than `usize::MAX` +/// the return value is unspecified. Uninhabited variants will be counted. +/// +/// Note that an enum may be expanded with additional variants in the future +/// as a non-breaking change, for example if it is marked `#[non_exhaustive]`, +/// which will change the result of this function. +/// +/// # Examples +/// +/// ``` +/// # #![feature(never_type)] +/// # #![feature(variant_count)] +/// +/// use std::mem; +/// +/// enum Void {} +/// enum Foo { A(&'static str), B(i32), C(i32) } +/// +/// assert_eq!(mem::variant_count::(), 0); +/// assert_eq!(mem::variant_count::(), 3); +/// +/// assert_eq!(mem::variant_count::>(), 2); +/// assert_eq!(mem::variant_count::>(), 2); +/// ``` +#[inline(always)] +#[must_use] +#[unstable(feature = "variant_count", issue = "73662")] +#[rustc_const_unstable(feature = "variant_count", issue = "73662")] +#[rustc_diagnostic_item = "mem_variant_count"] +pub const fn variant_count() -> usize { + intrinsics::variant_count::() +} diff --git a/library/core/src/mem/transmutability.rs b/library/core/src/mem/transmutability.rs new file mode 100644 index 000000000..b59a5b89d --- /dev/null +++ b/library/core/src/mem/transmutability.rs @@ -0,0 +1,43 @@ +/// Are values of a type transmutable into values of another type? +/// +/// This trait is implemented on-the-fly by the compiler for types `Src` and `Self` when the bits of +/// any value of type `Self` are safely transmutable into a value of type `Dst`, in a given `Context`, +/// notwithstanding whatever safety checks you have asked the compiler to [`Assume`] are satisfied. +#[unstable(feature = "transmutability", issue = "99571")] +#[cfg_attr(not(bootstrap), lang = "transmute_trait")] +#[rustc_on_unimplemented( + message = "`{Src}` cannot be safely transmuted into `{Self}` in the defining scope of `{Context}`.", + label = "`{Src}` cannot be safely transmuted into `{Self}` in the defining scope of `{Context}`." +)] +pub unsafe trait BikeshedIntrinsicFrom< + Src, + Context, + const ASSUME_ALIGNMENT: bool, + const ASSUME_LIFETIMES: bool, + const ASSUME_VALIDITY: bool, + const ASSUME_VISIBILITY: bool, +> where + Src: ?Sized, +{ +} + +/// What transmutation safety conditions shall the compiler assume that *you* are checking? +#[unstable(feature = "transmutability", issue = "99571")] +#[derive(PartialEq, Eq, Clone, Copy, Debug)] +pub struct Assume { + /// When `true`, the compiler assumes that *you* are ensuring (either dynamically or statically) that + /// destination referents do not have stricter alignment requirements than source referents. + pub alignment: bool, + + /// When `true`, the compiler assume that *you* are ensuring that lifetimes are not extended in a manner + /// that violates Rust's memory model. + pub lifetimes: bool, + + /// When `true`, the compiler assumes that *you* are ensuring that the source type is actually a valid + /// instance of the destination type. + pub validity: bool, + + /// When `true`, the compiler assumes that *you* have ensured that it is safe for you to violate the + /// type and field privacy of the destination type (and sometimes of the source type, too). + pub visibility: bool, +} diff --git a/library/core/src/mem/valid_align.rs b/library/core/src/mem/valid_align.rs new file mode 100644 index 000000000..fcfa95120 --- /dev/null +++ b/library/core/src/mem/valid_align.rs @@ -0,0 +1,247 @@ +use crate::convert::TryFrom; +use crate::num::NonZeroUsize; +use crate::{cmp, fmt, hash, mem, num}; + +/// A type storing a `usize` which is a power of two, and thus +/// represents a possible alignment in the rust abstract machine. +/// +/// Note that particularly large alignments, while representable in this type, +/// are likely not to be supported by actual allocators and linkers. +#[derive(Copy, Clone)] +#[repr(transparent)] +pub(crate) struct ValidAlign(ValidAlignEnum); + +// ValidAlign is `repr(usize)`, but via extra steps. +const _: () = assert!(mem::size_of::() == mem::size_of::()); +const _: () = assert!(mem::align_of::() == mem::align_of::()); + +impl ValidAlign { + /// Creates a `ValidAlign` from a power-of-two `usize`. + /// + /// # Safety + /// + /// `align` must be a power of two. + /// + /// Equivalently, it must be `1 << exp` for some `exp` in `0..usize::BITS`. + /// It must *not* be zero. + #[inline] + pub(crate) const unsafe fn new_unchecked(align: usize) -> Self { + debug_assert!(align.is_power_of_two()); + + // SAFETY: By precondition, this must be a power of two, and + // our variants encompass all possible powers of two. + unsafe { mem::transmute::(align) } + } + + #[inline] + pub(crate) const fn as_nonzero(self) -> NonZeroUsize { + // SAFETY: All the discriminants are non-zero. + unsafe { NonZeroUsize::new_unchecked(self.0 as usize) } + } + + /// Returns the base 2 logarithm of the alignment. + /// + /// This is always exact, as `self` represents a power of two. + #[inline] + pub(crate) fn log2(self) -> u32 { + self.as_nonzero().trailing_zeros() + } +} + +impl fmt::Debug for ValidAlign { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + write!(f, "{:?} (1 << {:?})", self.as_nonzero(), self.log2()) + } +} + +impl TryFrom for ValidAlign { + type Error = num::TryFromIntError; + + #[inline] + fn try_from(align: NonZeroUsize) -> Result { + if align.is_power_of_two() { + // SAFETY: Just checked for power-of-two + unsafe { Ok(ValidAlign::new_unchecked(align.get())) } + } else { + Err(num::TryFromIntError(())) + } + } +} + +impl TryFrom for ValidAlign { + type Error = num::TryFromIntError; + + #[inline] + fn try_from(align: usize) -> Result { + if align.is_power_of_two() { + // SAFETY: Just checked for power-of-two + unsafe { Ok(ValidAlign::new_unchecked(align)) } + } else { + Err(num::TryFromIntError(())) + } + } +} + +impl cmp::Eq for ValidAlign {} + +impl cmp::PartialEq for ValidAlign { + #[inline] + fn eq(&self, other: &Self) -> bool { + self.as_nonzero() == other.as_nonzero() + } +} + +impl cmp::Ord for ValidAlign { + #[inline] + fn cmp(&self, other: &Self) -> cmp::Ordering { + self.as_nonzero().cmp(&other.as_nonzero()) + } +} + +impl cmp::PartialOrd for ValidAlign { + #[inline] + fn partial_cmp(&self, other: &Self) -> Option { + Some(self.cmp(other)) + } +} + +impl hash::Hash for ValidAlign { + #[inline] + fn hash(&self, state: &mut H) { + self.as_nonzero().hash(state) + } +} + +#[cfg(target_pointer_width = "16")] +type ValidAlignEnum = ValidAlignEnum16; +#[cfg(target_pointer_width = "32")] +type ValidAlignEnum = ValidAlignEnum32; +#[cfg(target_pointer_width = "64")] +type ValidAlignEnum = ValidAlignEnum64; + +#[derive(Copy, Clone)] +#[repr(u16)] +enum ValidAlignEnum16 { + _Align1Shl0 = 1 << 0, + _Align1Shl1 = 1 << 1, + _Align1Shl2 = 1 << 2, + _Align1Shl3 = 1 << 3, + _Align1Shl4 = 1 << 4, + _Align1Shl5 = 1 << 5, + _Align1Shl6 = 1 << 6, + _Align1Shl7 = 1 << 7, + _Align1Shl8 = 1 << 8, + _Align1Shl9 = 1 << 9, + _Align1Shl10 = 1 << 10, + _Align1Shl11 = 1 << 11, + _Align1Shl12 = 1 << 12, + _Align1Shl13 = 1 << 13, + _Align1Shl14 = 1 << 14, + _Align1Shl15 = 1 << 15, +} + +#[derive(Copy, Clone)] +#[repr(u32)] +enum ValidAlignEnum32 { + _Align1Shl0 = 1 << 0, + _Align1Shl1 = 1 << 1, + _Align1Shl2 = 1 << 2, + _Align1Shl3 = 1 << 3, + _Align1Shl4 = 1 << 4, + _Align1Shl5 = 1 << 5, + _Align1Shl6 = 1 << 6, + _Align1Shl7 = 1 << 7, + _Align1Shl8 = 1 << 8, + _Align1Shl9 = 1 << 9, + _Align1Shl10 = 1 << 10, + _Align1Shl11 = 1 << 11, + _Align1Shl12 = 1 << 12, + _Align1Shl13 = 1 << 13, + _Align1Shl14 = 1 << 14, + _Align1Shl15 = 1 << 15, + _Align1Shl16 = 1 << 16, + _Align1Shl17 = 1 << 17, + _Align1Shl18 = 1 << 18, + _Align1Shl19 = 1 << 19, + _Align1Shl20 = 1 << 20, + _Align1Shl21 = 1 << 21, + _Align1Shl22 = 1 << 22, + _Align1Shl23 = 1 << 23, + _Align1Shl24 = 1 << 24, + _Align1Shl25 = 1 << 25, + _Align1Shl26 = 1 << 26, + _Align1Shl27 = 1 << 27, + _Align1Shl28 = 1 << 28, + _Align1Shl29 = 1 << 29, + _Align1Shl30 = 1 << 30, + _Align1Shl31 = 1 << 31, +} + +#[derive(Copy, Clone)] +#[repr(u64)] +enum ValidAlignEnum64 { + _Align1Shl0 = 1 << 0, + _Align1Shl1 = 1 << 1, + _Align1Shl2 = 1 << 2, + _Align1Shl3 = 1 << 3, + _Align1Shl4 = 1 << 4, + _Align1Shl5 = 1 << 5, + _Align1Shl6 = 1 << 6, + _Align1Shl7 = 1 << 7, + _Align1Shl8 = 1 << 8, + _Align1Shl9 = 1 << 9, + _Align1Shl10 = 1 << 10, + _Align1Shl11 = 1 << 11, + _Align1Shl12 = 1 << 12, + _Align1Shl13 = 1 << 13, + _Align1Shl14 = 1 << 14, + _Align1Shl15 = 1 << 15, + _Align1Shl16 = 1 << 16, + _Align1Shl17 = 1 << 17, + _Align1Shl18 = 1 << 18, + _Align1Shl19 = 1 << 19, + _Align1Shl20 = 1 << 20, + _Align1Shl21 = 1 << 21, + _Align1Shl22 = 1 << 22, + _Align1Shl23 = 1 << 23, + _Align1Shl24 = 1 << 24, + _Align1Shl25 = 1 << 25, + _Align1Shl26 = 1 << 26, + _Align1Shl27 = 1 << 27, + _Align1Shl28 = 1 << 28, + _Align1Shl29 = 1 << 29, + _Align1Shl30 = 1 << 30, + _Align1Shl31 = 1 << 31, + _Align1Shl32 = 1 << 32, + _Align1Shl33 = 1 << 33, + _Align1Shl34 = 1 << 34, + _Align1Shl35 = 1 << 35, + _Align1Shl36 = 1 << 36, + _Align1Shl37 = 1 << 37, + _Align1Shl38 = 1 << 38, + _Align1Shl39 = 1 << 39, + _Align1Shl40 = 1 << 40, + _Align1Shl41 = 1 << 41, + _Align1Shl42 = 1 << 42, + _Align1Shl43 = 1 << 43, + _Align1Shl44 = 1 << 44, + _Align1Shl45 = 1 << 45, + _Align1Shl46 = 1 << 46, + _Align1Shl47 = 1 << 47, + _Align1Shl48 = 1 << 48, + _Align1Shl49 = 1 << 49, + _Align1Shl50 = 1 << 50, + _Align1Shl51 = 1 << 51, + _Align1Shl52 = 1 << 52, + _Align1Shl53 = 1 << 53, + _Align1Shl54 = 1 << 54, + _Align1Shl55 = 1 << 55, + _Align1Shl56 = 1 << 56, + _Align1Shl57 = 1 << 57, + _Align1Shl58 = 1 << 58, + _Align1Shl59 = 1 << 59, + _Align1Shl60 = 1 << 60, + _Align1Shl61 = 1 << 61, + _Align1Shl62 = 1 << 62, + _Align1Shl63 = 1 << 63, +} -- cgit v1.2.3