Get rid of `mir::Const::from_ty_const`
This function is strange, because it turns valtrees into `mir::Const::Value`, but the rest of the const variants stay as type system consts.
All of the callsites except for one in `instsimplify` (array length simplification of `ptr_metadata` call) just go through the valtree arm of the function, so it's easier to just create a `mir::Const` directly for those.
For the instsimplify case, if we have a type system const we should *keep* having a type system const, rather than turning it into a `mir::Const::Value`; it doesn't really matter in practice, though, bc `usize` has no padding, but it feels more principled.
This lint detects and removes the unnecessary semicolon after a `match`
or `if` statement returning `()`. It seems to be quite a common
"mistake", given the number of hits (88) we had in the Clippy sources
themselves.
The lint doesn't bother about loops, as `rustfmt` already removes the
extra semicolon. It doesn't handle blocks either, as an extra block
level, followed or not by a semicolon, is likely intentional.
I propose to put the lint in `pedantic`, as putting it in `style` seems
quite hazardous given the number of hits.
Note: there exists a `redundant-semicolon` lint in the compiler, but it
is an early lint and cannot check that the expression evaluates to `()`,
so it ignores the cases we're handling here.
----
changelog: [`unnecessary_semicolon`]: new lint
changelog: [`useless-nonzero-new_unchecked`]: new lint
Close#13991
### What it does
Checks for `NonZero*::new_unchecked(<literal>)` being used in a `const`
context.
### Why is this bad?
Using `NonZero*::new_unchecked()` is an `unsafe` function and requires
an `unsafe` context. When used with an
integer literal in a `const` context, `NonZero*::new().unwrap()` will
provide the same result with identical
runtime performances while not requiring `unsafe`.
### Example
```no_run
const PLAYERS: NonZeroUsize = unsafe { NonZeroUsize::new_unchecked(3) };
```
Use instead:
```no_run
const PLAYERS: NonZeroUsize = NonZeroUsize::new(3).unwrap();
```
Commit 9ef6e2199c introduced a check to
ensure that Clippy doesn't consider a lifetime present in an explicit
self type as being the default for an elided output lifetime. For
example, elision did not work in the case like:
```rust
fn func(self: &Rc<Self>, &str) -> &str { … }
```
Since Rust 1.81.0, the lifetime in the self type is now considered
the default for elision. Elision should then be suggested when
appropriate.
changelog: [`manual_ok_err`]: new lint
Detect manual implementations of `.ok()` or `.err()`, as in
```rust
let a = match func() {
Ok(v) => Some(v),
Err(_) => None,
};
let b = if let Err(v) = func() {
Some(v)
} else {
None
};
```
which can be replaced by
```rust
let a = func().ok();
let b = func().err();
```
This pattern was detected in the wild in the Rust reimplementation of
coreutils:
https://github.com/uutils/coreutils/pull/6886#pullrequestreview-2465160137
Fixes#11846.
This PR has three commits:
- The first commit adds an `initializer-suggestions` configuration to
control suggestion applicability when initializers are present. The
following are the options:
- "none": do not suggest
- "maybe-incorrect": suggest, but do not apply suggestions with `--fix`
- "machine-applicable": suggest and apply suggestions with `--fix`
- The second commit fixes suggestions to handle field attributes
(problem [noticed by
@samueltardieu](https://github.com/rust-lang/rust-clippy/pull/13737#discussion_r1859261645)).
- The third commit adds `initializer-suggestions = "machine-applicable"`
to Clippy's `clippy.toml` and applies the suggestions. (Nothing seems to
break.)
---
changelog: make `inconsistent_struct_constructor` "all fields are
shorthand" requirement configurable
Begin to implement type system layer of unsafe binders
Mostly TODOs, but there's a lot of match arms that are basically just noops so I wanted to split these out before I put up the MIR lowering/projection part of this logic.
r? oli-obk
Tracking:
- https://github.com/rust-lang/rust/issues/130516
Fixes#13375
I've added the lint next to the other attribute-related ones. Not sure
if this is the correct place, since while we are looking after the
`packed`-attribute (there is nothing we can do about types defined
elsewhere), we are more concerned about the type's representation set by
the attribute (instead of "duplicate attributes" and such).
The lint simply looks at the attributes themselves without concern for
the item-kind, since items where `repr` is not allowed end up in a
compile-error anyway.
I'm somewhat concerned about the level of noise this lint would cause
if/when it goes into stable, although it does _not_ come up in
`lintcheck`.
```
changelog: [`repr_packed_without_abi`]: Initial implementation
```
