Restrict `#[rustc_box]` to `Box::new` calls
Currently, `#[rustc_box]` can be applied to any call expression with a single argument. This PR only allows it to be applied to calls to `Box::new`
Add support for QNX Neutrino to standard library
This change:
- adds standard library support for QNX Neutrino (7.1).
- upgrades `libc` to version `0.2.139` which supports QNX Neutrino
`@gh-tr`
⚠️ Backtraces on QNX require https://github.com/rust-lang/backtrace-rs/pull/507 which is not yet merged! (But everything else works without these changes) ⚠️
Tested mainly with a x86_64 virtual machine (see qnx-nto.md) and partially with an aarch64 hardware (some tests fail due to constrained resources).
Merge two different equality specialization traits in `core`
Arrays and slices each had their own version of this, without a matching set of `impl`s.
Merge them into one (still-`pub(crate)`) `cmp::BytewiseEq` trait, so we can stop doing all these things twice.
And that means that the `[T]::eq` → `memcmp` specialization picks up a bunch of types where that previously only worked for arrays, so examples like <https://rust.godbolt.org/z/KjsG8MGGT> will use it now instead of emitting loops.
r? the8472
Rollup of 7 pull requests
Successful merges:
- #108143 (rustdoc: search by macro when query ends with `!`)
- #108394 (Make `x doc --open` work on every book)
- #108427 (Recover from for-else and while-else)
- #108462 (Fix `VecDeque::append` capacity overflow for ZSTs)
- #108568 (Make associated_item_def_ids for traits use an unstable option to also return associated types for RPITITs)
- #108604 (Add regression test for #107280)
- #108605 (Add regression test for #105821)
Failed merges:
r? `@ghost`
`@rustbot` modify labels: rollup
Recover from for-else and while-else
This recovers from attempts at writing for-else or while-else loops, which might help users coming from e.g. Python.
```rs
for _ in 0..0 {
// ...
} else {
// ...
}
```
Combined with trying to store it in a let binding, the current diagnostic can be a bit confusing. It mentions let-else and suggests wrapping the loop in parentheses, which the user probably doesn't want. let-else doesn't make sense for `for` and `while` loops, as they are of type `()` (which already is an irrefutable pattern and doesn't need let-else).
<details>
<summary>Current diagnostic</summary>
```rs
error: right curly brace `}` before `else` in a `let...else` statement not allowed
--> src/main.rs:4:5
|
4 | } else {
| ^
|
help: wrap the expression in parentheses
|
2 ~ let _x = (for _ in 0..0 {
3 |
4 ~ }) else {
|
```
</details>
Some questions:
- Can the wording for the error message be improved? Would "for...else loops are not allowed" fit better?
- Should we be more "conservative" in case we want to support this in the future (i.e. say "for...else loops are **currently** not allowed/supported")?
- Is there a better way than storing a `&'static str` for the loop type? It is used for substituting the placeholder in the locale file (since it can emit either `for...else` or `while...else`). Maybe there is an enum I could use that I couldn't find
Make `x doc --open` work on every book
Before this PR, the `--open` flag had to be configured explicitly for every book, and most of them didn't configure it, resulting in the flag silently failing in all but two books.
In this PR, the code to check for the `--open` flag is in the underlying `RustbookSrc` step rather than all the individual steps. This is done by passing the parent step as a field of `RustbookSrc`, so that we can check for the correct step in `maybe_open_in_browser`.
This was part of a larger change that in the end wasn't worth it. Still, I think it could be useful as-is.
Name LLVM anonymous constants by a hash of their contents
This makes the names stable between different versions of a crate unlike the `AllocId` naming, making LLVM IR comparisons with `llvm-diff` more practical.
Add `Option::as_`(`mut_`)`slice`
This adds the following functions:
* `Option<T>::as_slice(&self) -> &[T]`
* `Option<T>::as_mut_slice(&mut self) -> &[T]`
The `as_slice` and `as_mut_slice_mut` functions benefit from an optimization that makes them completely branch-free. ~~Unfortunately, this optimization is not available on by-value Options, therefore the `into_slice` implementations use the plain `match` + `slice::from_ref` approach.~~
Note that the optimization's soundness hinges on the fact that either the niche optimization makes the offset of the `Some(_)` contents zero or the mempory layout of `Option<T>` is equal to that of `Option<MaybeUninit<T>>`.
The idea has been discussed on [Zulip](https://rust-lang.zulipchat.com/#narrow/stream/219381-t-libs/topic/Option.3A.3Aas_slice). Notably the idea for the `as_slice_mut` and `into_slice´ methods came from `@cuviper` and `@Sp00ph` hardened the optimization against niche-optimized Options.
The [rust playground](https://play.rust-lang.org/?version=nightly&mode=release&edition=2021&gist=74f8e4239a19f454c183aaf7b4a969e0) shows that the generated assembly of the optimized method is basically only a copy while the naive method generates code containing a `test dx, dx` on x86_64.
---
EDIT from reviewer: ACP is https://github.com/rust-lang/libs-team/issues/150
Remove or document uses of #[rustc_box] in library
r? `@thomcc`
Only one of these uses is tested for in the rustc-perf benchmark suite. The impact there on compile time is somewhat dramatic, but I am inclined to make this change as a simplification to the library and wait for people to complain if it explodes their compilation time. I think in the absence of data or reports from users about what code paths really matter, if we are optimizing for compilation time, it's hard to argue against using `#[rustc_box]` everywhere we currently call `Box::new`.
Rollup of 8 pull requests
Successful merges:
- #108297 (Exit when there are unmatched delims to avoid noisy diagnostics)
- #108531 (rustdoc: Show that repeated expression arrays can be made with constant values)
- #108536 (Update books)
- #108550 (Remove the `capture_disjoint_fields` feature)
- #108551 (Descriptive error when users try to combine RPITIT/AFIT with specialization)
- #108554 (Only look for param in item's generics if it actually comes from generics)
- #108555 (Fix a race in the query system)
- #108558 (add missing feature in core/tests)
Failed merges:
r? `@ghost`
`@rustbot` modify labels: rollup
compiler/rustc_session: fix sysroot detection logic
This pull request fixes the sysroot detection logic on systems where `/usr/lib` contains a multi-arch structure (e.g. installs `rustc_driver` into `/usr/lib/x86_64-linux-gnu` folder).
This fixes a regression for various Linux systems introduced in #103660. On Debian and Ubuntu systems, the logic in the pull request, as mentioned earlier, will resolve the sysroot to `/usr/lib`, making `rustc --print target-libdir` to return `/usr/lib/lib/rustlib/x86_64-unknown-linux-gnu/lib` (notice the extra `lib` at the beginning).
The fix is not very "clean" according to the standard. If you have any suggestions on improving the logic, you are more than welcome to comment below!
add missing feature in core/tests
https://github.com/rust-lang/rust/pull/104265 introduced the `ip_in_core` feature. For some reason core tests seem to still build without that feature -- no idea how that is possible. Might be related to https://github.com/rust-lang/rust/issues/15702? I was under the impression that `pub use` with different stability doesn't actually work. That's why `intrinsics::transmute` is stable, for example.
Either way, core tests fail to build in miri-test-libstd, and adding the feature fixes that.
r? ```@thomcc```
Fix a race in the query system
This fixes an issue where in between the `job` removal and `complete` call the query neither has a job nor a result, allowing another thread to start executing it again.
r? ``@cjgillot``
Only look for param in item's generics if it actually comes from generics
Record whether a `hir::GenericParam` comes from an item's generics, or from a `for<...>` binder. Then, only look for the param in `object_lifetime_default` if it actually comes from the item's generics.
Fixes#108177
