split MaybeUninit into several features, expand docs a bit
This splits the `maybe_uninit` feature gate into several:
* `maybe_uninit` for what we will hopefully stabilize soon-ish.
* `maybe_uninit_ref` for creating references into `MaybeUninit`, for which the rules are not yet clear.
* `maybe_uninit_slice` for handling slices of `MaybeUninit`, which needs more API design work.
* `maybe_uninit_array` for creating arrays of `MaybeUninit` using a macro (because we don't have https://github.com/rust-lang/rust/issues/49147 yet).
Is that an okay thing to do? The goal is to help people avoid APIs we do not want to stabilize yet. I used this to make sure rustc itself does not use `get_ref` and `get_mut`.
I also extended the docs to advise against uninitialized integers -- again this is something for which the rules are still being discussed.
rustdoc: mask `compiler_builtins` docs
Fixes#46783.
I wasn't able to fully confirm the underlying cause, but my theory is that functions in `compiler_builtins` were overwriting functions with the same names in libcore in the search index. Since the functions in `compiler_builtins` didn't have docs, that's why they weren't appearing in the results.
Masking the `compiler_builtins` crate fixes the search results. It appears that this crate was accidentally unmasked in #49503.
Use posix_spawn_file_actions_addchdir_np when possible
This is a non-POSIX extension implemented in Solaris and in glibc 2.29.
With this we can still use `posix_spawn()` when `Command::current_dir()`
has been set, otherwise we fallback to `fork(); chdir(); exec()`.
fix Box::into_unique effecitvely transmuting to a raw ptr
Miri/Stacked Borrows treat `Box` specially: they assert that it is unique, and tag it appropriately. However, currently, `Box::into_inner` is not aware of that and returns a raw pointer (wrapped in a `Unique`) that carries the same tag as the box, meaning it carries a `Uniq` tag. This leads to all sorts of problems when people use the raw pointer they get out of the `Unique` type.
In the future, it'd be interesting to make `Unique` also carry some kind of uniqueness. In that case, something like this would instead be needed whenever a raw pointer is extracted from a `Unique`. However, that is out-of-scope for the current version of Stacked Borrows. So until then, this changes `into_unique` to perform a proper reference-to-raw-ptr-cast, which clears the tag.
Add specific feature gate error for const-unstable features
Before:
```
error: `impl Trait` in const fn is unstable
--> src/lib.rs:7:19
|
7 | const fn foo() -> impl T {
| ^^^^^^
error: aborting due to previous error
```
After:
```
error[E0723]: `impl Trait` in const fn is unstable (see issue #57563)
--> src/lib.rs:7:19
|
7 | const fn foo() -> impl T {
| ^^^^^^
= help: add #![feature(const_fn)] to the crate attributes to enable
error: aborting due to previous error
```
This improves the situation with https://github.com/rust-lang/rust/issues/57563. Fixes https://github.com/rust-lang/rust/issues/57544. Fixes https://github.com/rust-lang/rust/issues/54469.
r? @oli-obk
Stabilize slice_sort_by_cached_key
I was going to ask on the tracking issue (https://github.com/rust-lang/rust/issues/34447), but decided to just send this and hope for an FCP here. The method was added last March by https://github.com/rust-lang/rust/pull/48639.
Signature: https://doc.rust-lang.org/std/primitive.slice.html#method.sort_by_cached_key
```rust
impl [T] {
pub fn sort_by_cached_key<K, F>(&mut self, f: F)
where F: FnMut(&T) -> K, K: Ord;
}
```
That's an identical signature to the existing `sort_by_key`, so I think the questions are just naming, implementation, and the usual "do we want this?".
The implementation seems to have proven its use in rustc at least, which many uses: https://github.com/rust-lang/rust/search?l=Rust&q=sort_by_cached_key
(I'm asking because it's exactly what I just needed the other day:
```rust
all_positions.sort_by_cached_key(|&n|
data::CITIES.iter()
.map(|x| *metric_closure.get_edge(n, x.pos).unwrap())
.sum::<usize>()
);
```
since caching that key is a pretty obviously good idea.)
Closes#34447
With `NllLivenessMap` and `LiveVar` removed, the `IdentityMap` (remaining
structure implementing the `LiveVariableMap` trait) loses its meaning.
Specialize the `LiveVarSet` to a `BitSet<Local>` removing the `V` and related
parameters. The `LiveVarSet<V>` was only being used as `LiveVarSet<Local>` so
this commit doesn't bring any change to the logic, it just removes an unused
parameter (that without `LiveVar` now, it couldn't have been specialized to
anything but `Local`).
Extend `LocalUseMap`'s `IndexVec`s that track def/use/drop data to store the
original `Local` indexes and not the compacted `LiveVar` ones (favoring speed
and code simplicity over space). Remove the `NllLivenessMap` embedded inside it
since it's no longer needed to perform the `LiveVar`/`Local` conversion.
It was used in `compute_for_all_locals` to iterate only the `Local`s that need
liveness analysis (filtered through `compute`). Instead, explicitly extract that
reduced set (as `live_locals`) in `trace` and pass it to
`compute_for_all_locals`.
Change the variable type used in `compute_for_all_locals` from `LiveVar` to
`Local` and do the same for its helper functions (and the functions in
`LocalUseMap` they rely on):
* `add_defs_for` -> `LocalUseMap::defs`
* `compute_use_live_points_for` -> `LocalUseMap::uses`
* `compute_drop_live_points_for` -> `LocalUseMap::drops`
Push back the use of `LiveVar` to the `LocalUseMap` (where the other
`NllLivenessMap` remains embedded) functions which internally do the
`from_local` conversion.
In our type inference system, when we "generalize" a type T to become
a suitable value for a type variable V, we sometimes wind up creating
new inference variables. So, for example, if we are making V be some
subtype of `&'X u32`, then we might instantiate V with `&'Y u32`.
This generalized type is then related `&'Y u32 <: &'X u32`, resulting
in a region constriant `'Y: 'X`. Previously, however, we were making
these fresh variables like `'Y` in the "current universe", but they
should be created in the universe of V. Moreover, we sometimes cheat
in an invariant context and avoid creating fresh variables if we know
the result must be equal -- we can only do that when the universes
work out.
rustc_mir: split qualify_consts' "value qualification" bitflags into separate computations.
Prerequisite for computing those bits through a dataflow algorithm ~~(which I might do in this PR later)~~.
This PR should not change behavior overall, other than treating `simd_shuffle*` identically to `#[rustc_args_required_const]` (maybe we should just have `#[rustc_args_required_const]` on the intrinsic imports of `simd_shuffle*`? cc @gnzlbg)
cc @oli-obk @alexreg
