Auto merge of rust-lang#53508 - japaric:maybe-uninit, r=RalfJung

Implement `MaybeUninit`

This PR:

- Adds `MaybeUninit` (see rust-lang#53491) to `{core,std}::mem`.
- Makes `mem::{uninitialized,zeroed}` panic when they are used to instantiate an uninhabited type.
- Does *not* deprecate `mem::{uninitialized,zeroed}` just yet. As per rust-lang#53491 (comment), we should not deprecate them until `MaybeUninit` is stabilized.
- It replaces uses of `mem::{uninitialized,zeroed}` in core and alloc with `MaybeUninit`.

There are still several instances of `mem::{uninitialized,zeroed}` in `std` that *this* PR doesn't address.

r? @RalfJung
cc @eddyb you may want to look at the new panicking logic

src/etc/gdb_rust_pretty_printing.py

@@ -322,8 +322,11 @@ def to_string(self):
     def children(self):
         (length, data_ptr) = \
             rustpp.extract_length_and_ptr_from_std_btreeset(self.__val)
-        val = GdbValue(data_ptr.get_wrapped_value().dereference()).get_child_at_index(3)
-        gdb_ptr = val.get_wrapped_value()
+        leaf_node = GdbValue(data_ptr.get_wrapped_value().dereference())
+        maybe_uninit_keys = leaf_node.get_child_at_index(3)
+        manually_drop_keys = maybe_uninit_keys.get_child_at_index(1)
+        keys = manually_drop_keys.get_child_at_index(0)
+        gdb_ptr = keys.get_wrapped_value()
         for index in xrange(length):
             yield (str(index), gdb_ptr[index])
 
@@ -345,9 +348,14 @@ def to_string(self):
     def children(self):
         (length, data_ptr) = \
             rustpp.extract_length_and_ptr_from_std_btreemap(self.__val)
-        keys = GdbValue(data_ptr.get_wrapped_value().dereference()).get_child_at_index(3)
+        leaf_node = GdbValue(data_ptr.get_wrapped_value().dereference())
+        maybe_uninit_keys = leaf_node.get_child_at_index(3)
+        manually_drop_keys = maybe_uninit_keys.get_child_at_index(1)
+        keys = manually_drop_keys.get_child_at_index(0)
         keys_ptr = keys.get_wrapped_value()
-        vals = GdbValue(data_ptr.get_wrapped_value().dereference()).get_child_at_index(4)
+        maybe_uninit_vals = leaf_node.get_child_at_index(4)
+        manually_drop_vals = maybe_uninit_vals.get_child_at_index(1)
+        vals = manually_drop_vals.get_child_at_index(0)
         vals_ptr = vals.get_wrapped_value()
         for index in xrange(length):
             yield (str(index), keys_ptr[index])

src/liballoc/collections/btree/node.rs

@@ -42,7 +42,7 @@
 //   This implies that even an empty internal node has at least one edge.
 
 use core::marker::PhantomData;
-use core::mem;
+use core::mem::{self, MaybeUninit};
 use core::ptr::{self, Unique, NonNull};
 use core::slice;
 
@@ -73,7 +73,7 @@ struct LeafNode<K, V> {
     /// This node's index into the parent node's `edges` array.
     /// `*node.parent.edges[node.parent_idx]` should be the same thing as `node`.
     /// This is only guaranteed to be initialized when `parent` is nonnull.
-    parent_idx: u16,
+    parent_idx: MaybeUninit<u16>,
 
     /// The number of keys and values this node stores.
     ///
@@ -83,8 +83,8 @@ struct LeafNode<K, V> {
 
     /// The arrays storing the actual data of the node. Only the first `len` elements of each
     /// array are initialized and valid.
-    keys: [K; CAPACITY],
-    vals: [V; CAPACITY],
+    keys: MaybeUninit<[K; CAPACITY]>,
+    vals: MaybeUninit<[V; CAPACITY]>,
 }
 
 impl<K, V> LeafNode<K, V> {
@@ -94,10 +94,10 @@ impl<K, V> LeafNode<K, V> {
         LeafNode {
             // As a general policy, we leave fields uninitialized if they can be, as this should
             // be both slightly faster and easier to track in Valgrind.
-            keys: mem::uninitialized(),
-            vals: mem::uninitialized(),
+            keys: MaybeUninit::uninitialized(),
+            vals: MaybeUninit::uninitialized(),
             parent: ptr::null(),
-            parent_idx: mem::uninitialized(),
+            parent_idx: MaybeUninit::uninitialized(),
             len: 0
         }
     }
@@ -115,10 +115,10 @@ unsafe impl Sync for LeafNode<(), ()> {}
 // ever take a pointer past the first key.
 static EMPTY_ROOT_NODE: LeafNode<(), ()> = LeafNode {
     parent: ptr::null(),
-    parent_idx: 0,
+    parent_idx: MaybeUninit::uninitialized(),
     len: 0,
-    keys: [(); CAPACITY],
-    vals: [(); CAPACITY],
+    keys: MaybeUninit::uninitialized(),
+    vals: MaybeUninit::uninitialized(),
 };
 
 /// The underlying representation of internal nodes. As with `LeafNode`s, these should be hidden
@@ -430,7 +430,7 @@ impl<BorrowType, K, V, Type> NodeRef<BorrowType, K, V, Type> {
                     root: self.root,
                     _marker: PhantomData
                 },
-                idx: self.as_leaf().parent_idx as usize,
+                idx: unsafe { usize::from(*self.as_leaf().parent_idx.get_ref()) },
                 _marker: PhantomData
             })
         } else {
@@ -567,7 +567,7 @@ impl<'a, K: 'a, V: 'a, Type> NodeRef<marker::Immut<'a>, K, V, Type> {
             // the node, which is allowed by LLVM.
             unsafe {
                 slice::from_raw_parts(
-                    self.as_leaf().keys.as_ptr(),
+                    self.as_leaf().keys.as_ptr() as *const K,
                     self.len()
                 )
             }
@@ -578,7 +578,7 @@ impl<'a, K: 'a, V: 'a, Type> NodeRef<marker::Immut<'a>, K, V, Type> {
         debug_assert!(!self.is_shared_root());
         unsafe {
             slice::from_raw_parts(
-                self.as_leaf().vals.as_ptr(),
+                self.as_leaf().vals.as_ptr() as *const V,
                 self.len()
             )
         }
@@ -605,7 +605,7 @@ impl<'a, K: 'a, V: 'a, Type> NodeRef<marker::Mut<'a>, K, V, Type> {
         } else {
             unsafe {
                 slice::from_raw_parts_mut(
-                    &mut self.as_leaf_mut().keys as *mut [K] as *mut K,
+                    self.as_leaf_mut().keys.get_mut() as *mut [K] as *mut K,
                     self.len()
                 )
             }
@@ -616,7 +616,7 @@ impl<'a, K: 'a, V: 'a, Type> NodeRef<marker::Mut<'a>, K, V, Type> {
         debug_assert!(!self.is_shared_root());
         unsafe {
             slice::from_raw_parts_mut(
-                &mut self.as_leaf_mut().vals as *mut [V] as *mut V,
+                self.as_leaf_mut().vals.get_mut() as *mut [V] as *mut V,
                 self.len()
             )
         }
@@ -1013,7 +1013,7 @@ impl<'a, K, V> Handle<NodeRef<marker::Mut<'a>, K, V, marker::Internal>, marker::
         let ptr = self.node.as_internal_mut() as *mut _;
         let mut child = self.descend();
         child.as_leaf_mut().parent = ptr;
-        child.as_leaf_mut().parent_idx = idx;
+        child.as_leaf_mut().parent_idx.set(idx);
     }
 
     /// Unsafely asserts to the compiler some static information about whether the underlying
@@ -1152,12 +1152,12 @@ impl<'a, K, V> Handle<NodeRef<marker::Mut<'a>, K, V, marker::Leaf>, marker::KV>
 
             ptr::copy_nonoverlapping(
                 self.node.keys().as_ptr().add(self.idx + 1),
-                new_node.keys.as_mut_ptr(),
+                new_node.keys.as_mut_ptr() as *mut K,
                 new_len
             );
             ptr::copy_nonoverlapping(
                 self.node.vals().as_ptr().add(self.idx + 1),
-                new_node.vals.as_mut_ptr(),
+                new_node.vals.as_mut_ptr() as *mut V,
                 new_len
             );
 
@@ -1210,12 +1210,12 @@ impl<'a, K, V> Handle<NodeRef<marker::Mut<'a>, K, V, marker::Internal>, marker::
 
             ptr::copy_nonoverlapping(
                 self.node.keys().as_ptr().add(self.idx + 1),
-                new_node.data.keys.as_mut_ptr(),
+                new_node.data.keys.as_mut_ptr() as *mut K,
                 new_len
             );
             ptr::copy_nonoverlapping(
                 self.node.vals().as_ptr().add(self.idx + 1),
-                new_node.data.vals.as_mut_ptr(),
+                new_node.data.vals.as_mut_ptr() as *mut V,
                 new_len
             );
             ptr::copy_nonoverlapping(

src/liballoc/lib.rs

@@ -119,6 +119,7 @@
 #![feature(exact_chunks)]
 #![feature(rustc_const_unstable)]
 #![feature(const_vec_new)]
+#![feature(maybe_uninit)]
 
 // Allow testing this library
 

src/libcore/fmt/float.rs

@@ -9,7 +9,7 @@
 // except according to those terms.
 
 use fmt::{Formatter, Result, LowerExp, UpperExp, Display, Debug};
-use mem;
+use mem::MaybeUninit;
 use num::flt2dec;
 
 // Don't inline this so callers don't use the stack space this function
@@ -20,11 +20,11 @@ fn float_to_decimal_common_exact<T>(fmt: &mut Formatter, num: &T,
     where T: flt2dec::DecodableFloat
 {
     unsafe {
-        let mut buf: [u8; 1024] = mem::uninitialized(); // enough for f32 and f64
-        let mut parts: [flt2dec::Part; 4] = mem::uninitialized();
+        let mut buf = MaybeUninit::<[u8; 1024]>::uninitialized(); // enough for f32 and f64
+        let mut parts = MaybeUninit::<[flt2dec::Part; 4]>::uninitialized();
         let formatted = flt2dec::to_exact_fixed_str(flt2dec::strategy::grisu::format_exact,
                                                     *num, sign, precision,
-                                                    false, &mut buf, &mut parts);
+                                                    false, buf.get_mut(), parts.get_mut());
         fmt.pad_formatted_parts(&formatted)
     }
 }
@@ -38,10 +38,11 @@ fn float_to_decimal_common_shortest<T>(fmt: &mut Formatter, num: &T,
 {
     unsafe {
         // enough for f32 and f64
-        let mut buf: [u8; flt2dec::MAX_SIG_DIGITS] = mem::uninitialized();
-        let mut parts: [flt2dec::Part; 4] = mem::uninitialized();
+        let mut buf = MaybeUninit::<[u8; flt2dec::MAX_SIG_DIGITS]>::uninitialized();
+        let mut parts = MaybeUninit::<[flt2dec::Part; 4]>::uninitialized();
         let formatted = flt2dec::to_shortest_str(flt2dec::strategy::grisu::format_shortest, *num,
-                                                 sign, precision, false, &mut buf, &mut parts);
+                                                 sign, precision, false, buf.get_mut(),
+                                                 parts.get_mut());
         fmt.pad_formatted_parts(&formatted)
     }
 }
@@ -75,11 +76,11 @@ fn float_to_exponential_common_exact<T>(fmt: &mut Formatter, num: &T,
     where T: flt2dec::DecodableFloat
 {
     unsafe {
-        let mut buf: [u8; 1024] = mem::uninitialized(); // enough for f32 and f64
-        let mut parts: [flt2dec::Part; 6] = mem::uninitialized();
+        let mut buf = MaybeUninit::<[u8; 1024]>::uninitialized(); // enough for f32 and f64
+        let mut parts = MaybeUninit::<[flt2dec::Part; 6]>::uninitialized();
         let formatted = flt2dec::to_exact_exp_str(flt2dec::strategy::grisu::format_exact,
                                                   *num, sign, precision,
-                                                  upper, &mut buf, &mut parts);
+                                                  upper, buf.get_mut(), parts.get_mut());
         fmt.pad_formatted_parts(&formatted)
     }
 }
@@ -94,11 +95,11 @@ fn float_to_exponential_common_shortest<T>(fmt: &mut Formatter,
 {
     unsafe {
         // enough for f32 and f64
-        let mut buf: [u8; flt2dec::MAX_SIG_DIGITS] = mem::uninitialized();
-        let mut parts: [flt2dec::Part; 6] = mem::uninitialized();
+        let mut buf = MaybeUninit::<[u8; flt2dec::MAX_SIG_DIGITS]>::uninitialized();
+        let mut parts = MaybeUninit::<[flt2dec::Part; 6]>::uninitialized();
         let formatted = flt2dec::to_shortest_exp_str(flt2dec::strategy::grisu::format_shortest,
                                                      *num, sign, (0, 0), upper,
-                                                     &mut buf, &mut parts);
+                                                     buf.get_mut(), parts.get_mut());
         fmt.pad_formatted_parts(&formatted)
     }
 }

src/libcore/lib.rs

@@ -246,6 +246,8 @@ macro_rules! test_v512 { ($item:item) => {}; }
 #[allow(unused_macros)]
 macro_rules! vector_impl { ($([$f:ident, $($args:tt)*]),*) => { $($f!($($args)*);)* } }
 #[path = "../stdsimd/coresimd/mod.rs"]
+// replacing uses of mem::{uninitialized,zeroed} with MaybeUninit needs to be in the stdsimd repo
+#[allow(deprecated)]
 #[allow(missing_docs, missing_debug_implementations, dead_code, unused_imports)]
 #[unstable(feature = "stdsimd", issue = "48556")]
 #[cfg(not(stage0))] // allow changes to how stdsimd works in stage0

src/libcore/mem.rs

@@ -514,6 +514,7 @@ pub fn needs_drop<T>() -> bool {
 /// assert_eq!(0, x);
 /// ```
 #[inline]
+#[rustc_deprecated(since = "2.0.0", reason = "use `mem::MaybeUninit::zeroed` instead")]
 #[stable(feature = "rust1", since = "1.0.0")]
 pub unsafe fn zeroed<T>() -> T {
     intrinsics::init()
@@ -608,6 +609,7 @@ pub unsafe fn zeroed<T>() -> T {
 /// [copy_no]: ../intrinsics/fn.copy_nonoverlapping.html
 /// [`Drop`]: ../ops/trait.Drop.html
 #[inline]
+#[rustc_deprecated(since = "2.0.0", reason = "use `mem::MaybeUninit::uninitialized` instead")]
 #[stable(feature = "rust1", since = "1.0.0")]
 pub unsafe fn uninitialized<T>() -> T {
     intrinsics::uninit()
@@ -1024,3 +1026,97 @@ impl<T: ?Sized> DerefMut for ManuallyDrop<T> {
         &mut self.value
     }
 }
+
+/// A newtype to construct uninitialized instances of `T`
+#[allow(missing_debug_implementations)]
+#[unstable(feature = "maybe_uninit", issue = "53491")]
+// NOTE after stabilizing `MaybeUninit` proceed to deprecate `mem::{uninitialized,zeroed}`
+pub union MaybeUninit<T> {
+    uninit: (),
+    value: ManuallyDrop<T>,
+}
+
+impl<T> MaybeUninit<T> {
+    /// Create 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.
+    #[unstable(feature = "maybe_uninit", issue = "53491")]
+    pub const fn uninitialized() -> MaybeUninit<T> {
+        MaybeUninit { uninit: () }
+    }
+
+    /// Create 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<usize>::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.
+    #[unstable(feature = "maybe_uninit", issue = "53491")]
+    pub fn zeroed() -> MaybeUninit<T> {
+        let mut u = MaybeUninit::<T>::uninitialized();
+        unsafe {
+            u.as_mut_ptr().write_bytes(0u8, 1);
+        }
+        u
+    }
+
+    /// Set the value of the `MaybeUninit`. This overwrites any previous value without dropping it.
+    #[unstable(feature = "maybe_uninit", issue = "53491")]
+    pub fn set(&mut self, val: T) {
+        unsafe {
+            self.value = ManuallyDrop::new(val);
+        }
+    }
+
+    /// Extract 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.
+    ///
+    /// # Unsafety
+    ///
+    /// It is up to the caller to guarantee that the the `MaybeUninit` really is in an initialized
+    /// state, otherwise this will immediately cause undefined behavior.
+    #[unstable(feature = "maybe_uninit", issue = "53491")]
+    pub unsafe fn into_inner(self) -> T {
+        ManuallyDrop::into_inner(self.value)
+    }
+
+    /// Get a reference to the contained value.
+    ///
+    /// # Unsafety
+    ///
+    /// It is up to the caller to guarantee that the the `MaybeUninit` really is in an initialized
+    /// state, otherwise this will immediately cause undefined behavior.
+    #[unstable(feature = "maybe_uninit", issue = "53491")]
+    pub unsafe fn get_ref(&self) -> &T {
+        &*self.value
+    }
+
+    /// Get a mutable reference to the contained value.
+    ///
+    /// # Unsafety
+    ///
+    /// It is up to the caller to guarantee that the the `MaybeUninit` really is in an initialized
+    /// state, otherwise this will immediately cause undefined behavior.
+    #[unstable(feature = "maybe_uninit", issue = "53491")]
+    pub unsafe fn get_mut(&mut self) -> &mut T {
+        &mut *self.value
+    }
+
+    /// Get a pointer to the contained value. Reading from this pointer will be undefined
+    /// behavior unless the `MaybeUninit` is initialized.
+    #[unstable(feature = "maybe_uninit", issue = "53491")]
+    pub fn as_ptr(&self) -> *const T {
+        unsafe { &*self.value as *const T }
+    }
+
+    /// Get a mutable pointer to the contained value. Reading from this pointer will be undefined
+    /// behavior unless the `MaybeUninit` is initialized.
+    #[unstable(feature = "maybe_uninit", issue = "53491")]
+    pub fn as_mut_ptr(&mut self) -> *mut T {
+        unsafe { &mut *self.value as *mut T }
+    }
+}