Remove allocations from case-insensitive comparison to keywords
Follows up on work in 99d02fb40f, expanding the alloc-free comparisons to more cases of case-insensitive keyword matching.
r? ghost for perf
Rollup of 7 pull requests
Successful merges:
- #135348 (rustdoc-json: Include items in stripped modules in `Crate::paths`.)
- #135365 (Update the explanation for why we use box_new in vec!)
- #135383 (De-abstract tagged ptr and make it covariant)
- #135401 (Remove some empty expected files to fix blessing)
- #135406 (Update unstable lint docs to include required feature attributes)
- #135407 (Deny various clippy lints)
- #135411 (run_make_support: add `#![warn(unreachable_pub)]`)
r? `@ghost`
`@rustbot` modify labels: rollup
Update unstable lint docs to include required feature attributes
closes#135298
## Summary
This PR updates the documentation examples for the following unstable lints to ensure they include the necessary feature attributes for proper usage:
- fuzzy_provenance_casts
- lossy_provenance_casts
- unqualified_local_imports
- test_unstable_lint
## Changes Made:
- Added the appropriate #![feature(...)] attributes to the example code for each lint.
- Updated the examples to produce correct and meaningful warnings, ensuring they align with current lint behavior.
Reference:
- Used the `must_not_suspend` lint documentation as a template for these updates.
Remove some empty expected files to fix blessing
https://github.com/rust-lang/rust/pull/134808 made --bless remove empty
expected files. Remove some empty files that were causing noise in
unrelated `--bless` invocations.
De-abstract tagged ptr and make it covariant
In #135272 I needed to use a tagged ptr in `hir::TyKind` in order to not regress hir type sizes. Unfortunately the existing `CopyTaggedPtr` abstraction is insufficient as it makes the `'hir` lifetime invariant.
I spent some time trying to keep existing functionality while making it covariant but in the end I realised that actually we dont use *any* of this code *anywhere* in rustc, so I've just removed everything and replaced it with a much less general abstraction that is suitable for what I need in #135272.
Idk if anyone has a preference for just keeping all the abstractions here in case anyone needs them in the future 🤷♀️
Update the explanation for why we use box_new in vec!
The perf run in this PR demonstrates that there is no longer a dramatic change in compile time with the intrinsic `box_new` vs calling `Box::new`, but I've locally confirmed that there is still a dramatic change in stack use.
rustdoc-json: Include items in stripped modules in `Crate::paths`.
Closes#135309
When we're running rustdoc-json, we should err on the side of adding more items to `Cache::paths`, as that directly becomes `Crate::paths` in the output.
r? ``@GuillaumeGomez.`` Best reviewed commit-by-commit.
Add #[inline] to copy_from_slice
I'm doing cooked things to CGU partitioning for compiler-builtins (https://github.com/rust-lang/rust/pull/135395) and this was the lone symbol in my compiler-builtins rlib that wasn't an intrinsic. Adding `#[inline]` makes it go away.
Perf report indicates a marginal but chaotic effect on compile time, marginal improvement in codegen. As expected.
centralize build stamp logic
This PR brings all the stamp file handling into one place inside `build_stamp` module, which takes care of everything related to build stamps. By doing this, we cut down on duplicated code and types and keep the codebase easier to maintain and more consistent.
Main goals are:
- Make stamp handling stricter so we don't have to pass `Path`s around and manually `join` on arbitrary directories
- Keep all stamp-related logic in one place
- Make it easier to test and debug
- Avoid duplication
- Keep things simple and well-documented
Resolves#134962
Rollup of 6 pull requests
Successful merges:
- #129259 (Add inherent versions of MaybeUninit methods for slices)
- #135374 (Suggest typo fix when trait path expression is typo'ed)
- #135377 (Make MIR cleanup for functions with impossible predicates into a real MIR pass)
- #135378 (Remove a bunch of diagnostic stashing that doesn't do anything)
- #135397 (compiletest: add erroneous variant to `string_enum`s conversions error)
- #135398 (add more crash tests)
r? `@ghost`
`@rustbot` modify labels: rollup
ci: Move dist-aarch64-linux to an aarch64 runner
Move the dist-aarch64-linux CI job to an aarch64 runner instead of cross-compiling it from an x86 one. This will make it possible to perform optimisations such as LTO, PGO and BOLT later on.
r? `@Kobzol`
Reland of #133809 now that the higher page sizes are fixed.
try-job: dist-aarch64-linux
try-job: dist-x86_64-linux
try-job: dist-i686-linux
compiletest: add erroneous variant to `string_enum`s conversions error
As requested in #135392, this adds which variant caused the string conversion failure.
r? jieyouxu
fixes#135392
Remove a bunch of diagnostic stashing that doesn't do anything
#121669 removed a bunch of conditional diagnostic stashing/canceling, but left around the `steal` calls which just emitted the error eagerly instead of canceling the diagnostic. I think that these no-op `steal` calls don't do much and are confusing to encounter, so let's remove them.
The net effect is:
1. We emit more duplicated errors, since stashing has the side effect of duplicating diagnostics. This is not a big deal, since outside of `-Zdeduplicate-diagnostics=no`, the errors are already being deduplicated by the compiler.
2. It changes the order of diagnostics, since we're no longer stashing and then later stealing the errors. I don't think this matters much for the changes that the UI test suite manifests, and it makes these errors less order dependent.
Make MIR cleanup for functions with impossible predicates into a real MIR pass
It's a bit jarring to see the body of a function with an impossible-to-satisfy where clause suddenly go to a single `unreachable` terminator when looking at the MIR dump output in order, and I discovered it's because we manually replace the body outside of a MIR pass.
Let's make it into a fully flegded MIR pass so it's more clear what it's doing and when it's being applied.
Suggest typo fix when trait path expression is typo'ed
When users write something like `Default::defualt()` (notice the typo), failure to resolve the erroneous `defualt` item will cause resolution + lowering to interpret this as a type-dependent path whose self type is `Default` which is a trait object without `dyn`, rather than a trait function like `<_ as Default>::default()`.
Try to provide a bit of guidance in this situation when we can detect the typo.
Fixes https://github.com/rust-lang/rust/issues/135349
Add inherent versions of MaybeUninit methods for slices
This is my attempt to un-stall #63569 and #79995, by creating methods that mirror the existing `MaybeUninit` API:
```rust
impl<T> MaybeUninit<T> {
pub fn write(&mut self, value: T) -> &mut T;
pub fn as_bytes(&self) -> &[MaybeUninit<u8>];
pub fn as_bytes_mut(&mut self) -> &mut [MaybeUninit<u8>];
pub unsafe fn assume_init_drop(&mut self);
pub unsafe fn assume_init_ref(&self) -> &T;
pub unsafe fn assume_init_mut(&mut self) -> &mut T;
}
```
Adding these APIs:
```rust
impl<T> [MaybeUninit<T>] {
// replacing copy_from_slice; renamed to avoid conflict
pub fn write_copy_of_slice(&mut self, value: &[T]) -> &mut [T] where T: Copy;
// replacing clone_from_slice; renamed to avoid conflict
pub fn write_clone_of_slice(&mut self, value: &[T]) -> &mut [T] where T: Clone;
// identical to non-slice versions; no conflict
pub fn as_bytes(&self) -> &[MaybeUninit<u8>];
pub fn as_bytes_mut(&mut self) -> &mut [MaybeUninit<u8>];
pub unsafe fn assume_init_drop(&mut self);
pub unsafe fn assume_init_ref(&self) -> &[T];
pub unsafe fn assume_init_mut(&mut self) -> &mut [T];
}
```
Since the `assume_init` methods are identical to those on non-slices, they feel pretty natural. The main issue with the write methods is naming, as discussed in #79995 among other places. My rationale:
* The term "write" should be in them somewhere, to mirror the other API, and this pretty much automatically makes them not collide with any other inherent slice methods.
* I chose `write_clone_of_slice` and `write_copy_of_slice` since `clone` and `copy` are being used as objects here, whereas they're being used as actions in `clone_from_slice` and `copy_from_slice`.
The final "weird" thing I've done in this PR is remove a link to `Vec<T>` from `assume_init_drop` (both copies, since they're effectively copied docs), since there's no good way to link to `Vec` for something that can occur both on the page for `std/primitive.slice.html` and `std/vec/struct.Vec.html`, since the code here lives in libcore and can't use intra-doc-linking to mention `Vec`. (see: #121436)
The reason why this method shows up both on `Vec<T>` and `[T]` is because the `[T]` docs are automatically inlined on `Vec<T>`'s page, since it implements `Deref`. It's unfortunate that rustdoc doesn't have a way of dealing with this at the moment, but it is what it is, and it's a reasonable compromise for now.
compiletest: include stage0-sysroot libstd dylib in recipe dylib search path
To fix some of the failures in `COMPILETEST_FORCE_STAGE0=1 ./x test run-make --stage 0`. Specifically,
```
COMPILETEST_FORCE_STAGE0=1 ./x test tests/run-make/rustdoc-default-output/ --stage 0
```
should now pass.
Fixes#135373. (As in, make *some* of the `run-make` tests pass, other `run-make` tests fail for various reasons against stage0, and generally `run-make` tests are not guaranteed to pass at stage 0.)
cc `@lolbinarycat`
r? bootstrap
Cleanup `suggest_binding_for_closure_capture_self` diag in borrowck
Mostly grammar fix/improvement, but also a small cleanup to use iterators instead of for loops for collecting into a vector.
Remove emsdk version update from 1.84.0 relnotes
See [this comment](https://github.com/rust-lang/rust/issues/131467#issuecomment-2529314603). The reproducer in that comment does indeed show that rustup's `rust-std` component is still compiled with the old emscripten ABI because libc's config flag `emscripten_new_stat_abi` is not set.
#131533 presumably had no effect because the wrong CI file was modified. So nothing has changed since 1.83.0. The PR author (workingjubilee) is currently on vacation.
Also the issue #131467 should be reopened.
