Most of these problems originate in use of get_unchecked_mut.
When calling ptr::copy_nonoverlapping, using get_unchecked_mut for both
arguments causes the borrow created to make the second pointer to invalid the
first.
The pairs of identical MaybeUninit::slice_as_mut_ptr calls similarly
invalidate each other.
There was also a similar borrow invalidation problem with the use of
slice::get_unchecked_mut to derive the pointer for the CopyOnDrop.
Add `[T]::as_simd(_mut)`
SIMD-style optimizations are the most common use for `[T]::align_to(_mut)`, but that's `unsafe`. So these are *safe* wrappers around it, now that we have the `Simd` type available, to make it easier to use.
```rust
impl [T] {
pub fn as_simd<const LANES: usize>(&self) -> (&[T], &[Simd<T, LANES>], &[T]);
pub fn as_simd_mut<const LANES: usize>(&mut self) -> (&mut [T], &mut [Simd<T, LANES>], &mut [T]);
}
```
They're `cfg`'d out for miri because the `simd` module as a whole is unavailable there.
Make split_inclusive() on an empty slice yield an empty output
`[].split_inclusive()` currently yields a single, empty slice. That's
different from `"".split_inslusive()`, which yields no output at
all. I think that makes the slice version harder to use.
The case where I ran into this bug was when writing code for
generating a diff between two slices of bytes. I wanted to prefix
removed lines with "-" and a added lines with "+". Due to
`split_inclusive()`'s current behavior, that means that my code prints
just a "-" or "+" for empty files. I suspect most existing callers
have similar "bugs" (which would be fixed by this patch).
Closes#89716.
Make certain panicky stdlib functions behave better under panic_immediate_abort
The stdlib has a `panic_immediate_abort` feature that turns panics into immediate aborts, without any formatting/display logic. This feature was [introduced](https://github.com/rust-lang/rust/pull/55011) primarily for codesize-constrained situations.
Unfortunately, this win doesn't quite propagate to `Result::expect()` and `Result::unwrap()`, while the formatting machinery is reduced, `expect()` and `unwrap()` both call `unwrap_failed("msg", &err)` which has a signature of `fn unwrap_failed(msg: &str, error: &dyn fmt::Debug)` and is `#[inline(never)]`. This means that `unwrap_failed` will unconditionally construct a `dyn Debug` trait object even though the object is never used in the function.
Constructing a trait object (even if you never call a method on it!) forces rust to include the vtable and any dependencies. This means that in `panic_immediate_abort` mode, calling expect/unwrap on a Result will pull in a whole bunch of formatting code for the error type even if it's completely unused.
This PR swaps out the function with one that won't require a trait object such that it won't force the inclusion of vtables in the code. It also gates off `#[inline(never)]` in a bunch of other places where allowing the inlining of an abort may be useful (this kind of thing is already done elsewhere in the stdlib).
I don't know how to write a test for this; we don't really seem to have any tests for `panic_immediate_abort` anyway so perhaps it's fine as is.
adjust const_eval_select documentation
"The Rust compiler assumes" indicates that this is language UB, but [I don't think that is a good idea](https://rust-lang.zulipchat.com/#narrow/stream/146212-t-compiler.2Fconst-eval/topic/const_eval_select.20assumptions). This UB would be very hard to test for and looks like a way-too-big footgun. ``@oli-obk`` suggested this is meant to be more like "library UB", so I tried to adjust the docs accordingly.
I also removed all references to "referential transparency". That is a rather vague concept used to mean many different things, and I honestly have no idea what exactly is meant by it in this specific instance. But I assume ``@fee1-dead`` had in their mind a property that all `const fn` code upholds, so by demanding that the runtime code and the const-time code are *observably equivalent*, whatever that property is would also be enforced here.
Cc ``@rust-lang/wg-const-eval``
SIMD-style optimizations are the most common use for `[T]::align_to(_mut)`, but that's `unsafe`. So these are *safe* wrappers around it, now that we have the `Simd` type available, to make it easier to use.
```rust
impl [T] {
pub fn as_simd<const LANES: usize>(&self) -> (&[T], &[Simd<T, LANES>], &[T]);
pub fn as_simd_mut<const LANES: usize>(&mut self) -> (&mut [T], &mut [Simd<T, LANES>], &mut [T]);
}
```
MIRI says `reverse` is UB, so replace it with something LLVM can vectorize
For small types with padding, the current implementation is UB because it does integer operations on uninit values.
```
error: Undefined Behavior: using uninitialized data, but this operation requires initialized memory
--> /playground/.rustup/toolchains/nightly-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/core/src/num/mod.rs:836:5
|
836 | / uint_impl! { u32, u32, i32, 32, 4294967295, 8, "0x10000b3", "0xb301", "0x12345678",
837 | | "0x78563412", "0x1e6a2c48", "[0x78, 0x56, 0x34, 0x12]", "[0x12, 0x34, 0x56, 0x78]", "", "" }
| |________________________________________________________________________________________________^ using uninitialized data, but this operation requires initialized memory
|
= help: this indicates a bug in the program: it performed an invalid operation, and caused Undefined Behavior
= help: see https://doc.rust-lang.org/nightly/reference/behavior-considered-undefined.html for further information
= note: inside `core::num::<impl u32>::rotate_left` at /playground/.rustup/toolchains/nightly-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/core/src/num/uint_macros.rs:211:13
= note: inside `core::slice::<impl [Foo]>::reverse` at /playground/.rustup/toolchains/nightly-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/core/src/slice/mod.rs:701:58
```
<https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=340739f22ca5b457e1da6f361768edc6>
But LLVM has gotten smarter since I wrote the previous implementation in 2017, so this PR removes all the manual magic and just writes it in such a way that LLVM will vectorize. This code is much simpler and has very little `unsafe`, and is actually faster to boot!
If you're curious to see the codegen: <https://rust.godbolt.org/z/Pcn13Y9E3>
Before:
```
running 7 tests
test slice::reverse_simd_f64x4 ... bench: 17,940 ns/iter (+/- 481) = 58448 MB/s
test slice::reverse_u128 ... bench: 17,758 ns/iter (+/- 205) = 59048 MB/s
test slice::reverse_u16 ... bench: 158,234 ns/iter (+/- 6,876) = 6626 MB/s
test slice::reverse_u32 ... bench: 62,047 ns/iter (+/- 1,117) = 16899 MB/s
test slice::reverse_u64 ... bench: 31,582 ns/iter (+/- 552) = 33201 MB/s
test slice::reverse_u8 ... bench: 81,253 ns/iter (+/- 1,510) = 12905 MB/s
test slice::reverse_u8x3 ... bench: 270,615 ns/iter (+/- 11,463) = 3874 MB/s
```
After:
```
running 7 tests
test slice::reverse_simd_f64x4 ... bench: 17,731 ns/iter (+/- 306) = 59137 MB/s
test slice::reverse_u128 ... bench: 17,919 ns/iter (+/- 239) = 58517 MB/s
test slice::reverse_u16 ... bench: 43,160 ns/iter (+/- 607) = 24295 MB/s
test slice::reverse_u32 ... bench: 21,065 ns/iter (+/- 371) = 49778 MB/s
test slice::reverse_u64 ... bench: 21,118 ns/iter (+/- 482) = 49653 MB/s
test slice::reverse_u8 ... bench: 76,878 ns/iter (+/- 1,688) = 13639 MB/s
test slice::reverse_u8x3 ... bench: 264,723 ns/iter (+/- 5,544) = 3961 MB/s
```
Those are the existing benches, <14a2fd640e/library/alloc/benches/slice.rs (L322-L346)>
For small types with padding, the current implementation is UB because it does integer operations on uninit values. But LLVM has gotten smarter since I wrote the previous implementation in 2017, so remove all the manual magic and just write it in such a way that LLVM will vectorize. This code is much simpler (albeit nuanced) and has very little `unsafe`, and is actually faster to boot!
Add #[must_use] to remaining core functions
I've run out of compelling reasons to group functions together across crates so I'm just going to go module-by-module. This is everything remaining from the `core` crate.
Ignored by clippy for reasons unknown:
```rust
core::alloc::Layout unsafe fn for_value_raw<T: ?Sized>(t: *const T) -> Self;
core::any const fn type_name_of_val<T: ?Sized>(_val: &T) -> &'static str;
```
Ignored by clippy because of `mut`:
```rust
str fn split_at_mut(&mut self, mid: usize) -> (&mut str, &mut str);
```
<del>
Ignored by clippy presumably because a caller might want `f` called for side effects. That seems like a bad usage of `map` to me.
```rust
core::cell::Ref<'b, T> fn map<U: ?Sized, F>(orig: Ref<'b, T>, f: F) -> Ref<'b, T>;
core::cell::Ref<'b, T> fn map_split<U: ?Sized, V: ?Sized, F>(orig: Ref<'b, T>, f: F) -> (Ref<'b, U>, Ref<'b, V>);
```
</del>
Parent issue: #89692
r? ```@joshtriplett```
Add #[must_use] to expensive computations
The unifying theme for this commit is weak, admittedly. I put together a list of "expensive" functions when I originally proposed this whole effort, but nobody's cared about that criterion. Still, it's a decent way to bite off a not-too-big chunk of work.
Given the grab bag nature of this commit, the messages I used vary quite a bit. I'm open to wording changes.
For some reason clippy flagged four `BTreeSet` methods but didn't say boo about equivalent ones on `HashSet`. I stared at them for a while but I can't figure out the difference so I added the `HashSet` ones in.
```rust
// Flagged by clippy.
alloc::collections::btree_set::BTreeSet<T> fn difference<'a>(&'a self, other: &'a BTreeSet<T>) -> Difference<'a, T>;
alloc::collections::btree_set::BTreeSet<T> fn symmetric_difference<'a>(&'a self, other: &'a BTreeSet<T>) -> SymmetricDifference<'a, T>
alloc::collections::btree_set::BTreeSet<T> fn intersection<'a>(&'a self, other: &'a BTreeSet<T>) -> Intersection<'a, T>;
alloc::collections::btree_set::BTreeSet<T> fn union<'a>(&'a self, other: &'a BTreeSet<T>) -> Union<'a, T>;
// Ignored by clippy, but not by me.
std::collections::HashSet<T, S> fn difference<'a>(&'a self, other: &'a HashSet<T, S>) -> Difference<'a, T, S>;
std::collections::HashSet<T, S> fn symmetric_difference<'a>(&'a self, other: &'a HashSet<T, S>) -> SymmetricDifference<'a, T, S>
std::collections::HashSet<T, S> fn intersection<'a>(&'a self, other: &'a HashSet<T, S>) -> Intersection<'a, T, S>;
std::collections::HashSet<T, S> fn union<'a>(&'a self, other: &'a HashSet<T, S>) -> Union<'a, T, S>;
```
Parent issue: #89692
r? ```@joshtriplett```
track_caller for slice length assertions
`clone_from_slice` was missing `#[track_caller]`, and its assert did not report a useful location.
These are small generic methods, so hopefully track_caller gets inlined into nothingness, but it may be worth running a benchmark on this.
Make RSplit<T, P>: Clone not require T: Clone
This addresses a TODO comment. The behavior of `#[derive(Clone)]` *does* result in a `T: Clone` requirement. Playground example:
https://play.rust-lang.org/?version=stable&mode=debug&edition=2018&gist=a8b1a9581ff8893baf401d624a53d35b
Add a manual `Clone` implementation, mirroring `Split` and `SplitInclusive`.
`(R)?SplitN(Mut)?` don't have any `Clone` implementations, but I'll leave that for its own pull request.
This addresses a TODO comment. The behavior of #[derive(Clone)]
*does* result in a T: Clone requirement.
Add a manual Clone implementation, matching Split and SplitInclusive.
The unifying theme for this commit is weak, admittedly. I put together a
list of "expensive" functions when I originally proposed this whole
effort, but nobody's cared about that criterion. Still, it's a decent
way to bite off a not-too-big chunk of work.
Given the grab bag nature of this commit, the messages I used vary quite
a bit.
`[].split_inclusive()` currently yields a single, empty slice. That's
different from `"".split_inslusive()`, which yields no output at
all. I think that makes the slice version harder to use.
The case where I ran into this bug was when writing code for
generating a diff between two slices of bytes. I wanted to prefix
removed lines with "-" and a added lines with "+". Due to
`split_inclusive()`'s current behavior, that means that my code prints
just a "-" or "+" for empty files. I suspect most existing callers
have similar "bugs" (which would be fixed by this patch).
Closes#89716.
