Pass through of target features to llvm-bitcode-linker and handling them
When using the llvm-bitcode-linker (`linker-flavor=llbc`) target-features are not passed through and are not handled by it.
The llvm-bitcode-linker is mainly used as a self contained linker to link llvm bitcode for the nvptx64 target. It uses `llvm-link`, `opt` and `llc` internally. To produce a `.ptx` file of a specific ptx-version it is necessary to pass the version to llc with the `--mattr` option. Without explicitly setting it, the emitted `.ptx`-version is the minimum supported version of the `--target-cpu`.
I would like to be able to explicitly set the ptx version as [some llvm problems only occur in earlier `.ptx`-versions](https://github.com/llvm/llvm-project/issues/112998).
Therefore this pull request adds support for passing target features to llvm-bitcode-linker and handling them.
I was not quite sure if adding these features to `rustc_target/src/target_features.rs` is necessary or not. If so I will gladly add these.
r? ``@kjetilkjeka``
Stabilize `num_midpoint_signed` feature
This PR proposes that we stabilize the signed variants of [`iN::midpoint`](https://github.com/rust-lang/rust/issues/110840#issue-1684506201), the operation is equivalent to doing `(a + b) / 2` in a sufficiently large number.
The stabilized API surface would be:
```rust
/// Calculates the middle point of `self` and `rhs`.
///
/// `midpoint(a, b)` is `(a + b) / 2` as if it were performed in a
/// sufficiently-large signed integer type. This implies that the result is
/// always rounded towards zero and that no overflow will ever occur.
impl i{8,16,32,64,128,size} {
pub const fn midpoint(self, rhs: Self) -> Self;
}
```
T-libs-api previously stabilized the unsigned (and float) variants in #131784, the signed variants were left out because of the rounding that should be used in case of negative midpoint.
This stabilization proposal proposes that we round towards zero because:
- it makes the obvious `(a + b) / 2` in a sufficiently-large number always true
- using another rounding for the positive result would be inconsistent with the unsigned variants
- it makes `midpoint(-a, -b)` == `-midpoint(a, b)` always true
- it is consistent with `midpoint(a as f64, b as f64) as i64`
- it makes it possible to always suggest `midpoint` as a replacement for `(a + b) / 2` expressions *(which we may want to do as a future work given the 21.2k hits on [GitHub Search](https://github.com/search?q=lang%3Arust+%2F%5C%28%5Ba-zA-Z_%5D*+%5C%2B+%5Ba-zA-Z_%5D*%5C%29+%5C%2F+2%2F&type=code&p=1))*
`@scottmcm` mentioned a drawback in https://github.com/rust-lang/rust/pull/132191#issuecomment-2439891200:
> I'm torn, because rounding towards zero makes it "wider" than other values, which `>> 1` avoids -- `(a + b) >> 1` has the nice behaviour that `midpoint(a, b) + 2 == midpoint(a + 2, b + 2)`.
>
> But I guess overall sticking with `(a + b) / 2` makes sense as well, and I do like the negation property 🤷
Which I think is outweigh by the advantages cited above.
Closes#110840
cc `@rust-lang/libs-api`
cc `@scottmcm`
r? `@dtolnay`
Bump sccache in CI to 0.9.1
We haven't updated the used sccache version for years, it has accrued a bunch of fixes and features in the meantime. It now supports the `--show-adv-stats` flag, which gives a more detailed summary of the results of caching. And it can also cache Rust code, which could be useful in the future (https://github.com/rust-lang/rust/pull/136942 - although now there are no large wins).
It also supports caching PGO now, but since the PGO profiles are always different, it won't make any real difference.
https://github.com/rust-lang/rust/pull/133076 previously tried to update the version to 0.3 (CC `@klensy)`
r? `@marcoieni`
Emit dropck normalization errors in borrowck
Borrowck generally assumes that any queries it runs for type checking will succeed, thinking that HIR typeck will have errored first if there was a problem. However as of #98641, dropck isn't run on HIR, so there's no direct guarantee that it doesn't error. While a type being well-formed might be expected to ensure that its fields are well-formed, this is not the case for types containing a type projection:
```rust
pub trait AuthUser {
type Id;
}
pub trait AuthnBackend {
type User: AuthUser;
}
pub struct AuthSession<Backend: AuthnBackend> {
data: Option<<<Backend as AuthnBackend>::User as AuthUser>::Id>,
}
pub trait Authz: Sized {
type AuthnBackend: AuthnBackend<User = Self>;
}
pub fn run_query<User: Authz>(auth: AuthSession<User::AuthnBackend>) {}
// ^ No User: AuthUser bound is required or inferred.
```
While improvements to trait solving might fix this in the future, for now we go for a pragmatic solution of emitting an error from borrowck (by rerunning dropck outside of a query) and making drop elaboration check if an error has been emitted previously before panicking for a failed normalization.
Closes#103899Closes#135039
r? `@compiler-errors` (feel free to re-assign)
improve cold_path()
#120370 added a new instrinsic `cold_path()` and used it to fix `likely` and `unlikely`
However, in order to limit scope, the information about cold code paths is only used in 2-target switch instructions. This is sufficient for `likely` and `unlikely`, but limits usefulness of `cold_path` for idiomatic rust. For example, code like this:
```
if let Some(x) = y { ... }
```
may generate 3-target switch:
```
switch y.discriminator:
0 => true branch
1 = > false branch
_ => unreachable
```
and therefore marking a branch as cold will have no effect.
This PR improves `cold_path()` to work with arbitrary switch instructions.
Note that for 2-target switches, we can use `llvm.expect`, but for multiple targets we need to manually emit branch weights. I checked Clang and it also emits weights in this situation. The Clang's weight calculation is more complex that this PR, which I believe is mainly because `switch` in `C/C++` can have multiple cases going to the same target.
Move methods from `Map` to `TyCtxt`, part 2.
Continuing the work started in #136466.
Every method gains a `hir_` prefix, though for the ones that already have a `par_` or `try_par_` prefix I added the `hir_` after that.
r? Zalathar
Allow configuring jemalloc per target
In Ferrocene we're trying to switch from `./configure` to a predefined `config.toml` file. One of the limitations of doing that is the `rust.jemalloc` configuration option, which we need to conditionally disable based on the target. This PR adds a `target.$tuple.jemalloc` option to override `rust.jemalloc` to make that possible.
CI: rfl: move job forward to Linux v6.14-rc3
Linux v6.14-rc3 contains commit 6273a058383e ("x86: rust: set rustc-abi=x86-softfloat on rustc>=1.86.0"), which resolves the error from https://github.com/rust-lang/rust/pull/136146.
r? `@lqd` `@Kobzol`
try-job: x86_64-rust-for-linux
`@rustbot` label A-rust-for-linux
`@bors` try
Use `tell` for `<File as Seek>::stream_position`
Some platforms have a more efficient way to get the current offset of the file than by seeking. For example, Wasi has `fd_tell` and SOLID has `SOLID_FS_Ftell`. Implement `<File as Seek>::stream_position()` in terms of those.
I do not use any APIs that were not already used in `std`. Although, the `libc` crate has [`ftell`](https://docs.rs/libc/latest/libc/fn.ftell.html), [`ftello`](https://docs.rs/libc/latest/libc/fn.ftello.html), and [`ftello64`](https://docs.rs/libc/latest/libc/fn.ftello64.html), I do not know platform coverage. It appears that Windows has no `tell`-like API.
I have checked that it builds on each relevant platform.
boostrap: skip no_std targets in Std doc step
This fixes a bug that currently prevents us from adding no_std library targets to rustc in nixpkgs (https://github.com/NixOS/nixpkgs/pull/382166).
When running `./x.py doc`, the `Std` doc step generally fails for no_std targets, logs: https://gist.github.com/niklaskorz/fb83f9503ce19b75e8b1af02cdebd592
Skipping no_std targets in this step will allow using no_std targets such as `bpfel-unknown-none` together with other targets in the same config without blocking the doc generator for them, e.g.
```
./configure --release-channel=stable --tools=rustc,rustdoc,rust-analyzer-proc-macro-srv --build=aarch64-apple-darwin --host=aarch64-apple-darwin --target=aarch64-apple-darwin,bpfel-unknown-none
./x.py doc
```
Logs with this fix applied: https://gist.github.com/niklaskorz/cdd50aaea33ede579f737434286d800b
Pass vendored sources from bootstrap to generate-copyright
In addition to doing the vendoring in bootstrap, this PR also loads the list of manifests to parse from bootstrap (instead of hardcoding a smaller list in generate-copyright). This is best reviewed commit-by-commit.
Fixes https://github.com/rust-lang/rust/issues/136955
Rollup of 7 pull requests
Successful merges:
- #137095 (Replace some u64 hashes with Hash64)
- #137100 (HIR analysis: Remove unnecessary abstraction over list of clauses)
- #137105 (Restrict DerefPure for Cow<T> impl to T = impl Clone, [impl Clone], str.)
- #137120 (Enable `relative-path-include-bytes-132203` rustdoc-ui test on Windows)
- #137125 (Re-add missing empty lines in the releases notes)
- #137145 (use add-core-stubs / minicore for a few more tests)
- #137149 (Remove SSE ABI from i586-pc-windows-msvc)
r? `@ghost`
`@rustbot` modify labels: rollup
I was wrong on #19127, I thought hir-def resolver is enough for them, but it turns out not because of paths like `<Enum>::Variant` and `Type::AssocThatIsEnum::Variant`.
Remove SSE ABI from i586-pc-windows-msvc
As an i586 target, it should not have SSE. This caused the following warning to be emitted:
```
warning: target feature `sse2` must be enabled to ensure that the ABI of the current target can be implemented correctly
|
= note: this was previously accepted by the compiler but is being phased out; it will become a hard error in a future release!
= note: for more information, see issue #116344 <https://github.com/rust-lang/rust/issues/116344>
warning: 1 warning emitted
```
see #116344.
r? RalfJung
use add-core-stubs / minicore for a few more tests
See https://github.com/rust-lang/rust/issues/131485 for context. These are some tests I worked on in the past so I figured I'd see if `minicore` works for them. :)
Enable `relative-path-include-bytes-132203` rustdoc-ui test on Windows
The problem with the error message on Windows is:
- The path separators are different
- The OS error message string is different
Normalizing those two things makes the test pass on Windows.
Restrict DerefPure for Cow<T> impl to T = impl Clone, [impl Clone], str.
Fixes#136046
`feature(deref_patterns)` tracking issue: https://github.com/rust-lang/rust/issues/87121
`Cow<'_, T>` should only implement `DerefPure` if its `Deref` impl is pure, which requires `<T::Owned as Borrow<T>>::borrow` to be pure. This PR restricts `impl DerefPure for Cow<'_, T>` to `T: Sized + Clone`, `T = [U: Clone]`, and `T = str` (for all of whom `<T::Owned as Borrow<T>>::borrow` is implemented in the stdlib and is pure).
cc ``@Nadrieril``
------
An alternate approach would be to introduce a new `unsafe trait BorrowPure<T>` analogous to `DerefPure` that could be implemented for `T: Sized`, `&T`, `&mut T`, `String`, `Vec`, `Box`, `PathBuf`, `OsString`, etc. https://github.com/rust-lang/rust/compare/master...zachs18:borrow-pure-trait
HIR analysis: Remove unnecessary abstraction over list of clauses
`rustc_hir_analysis::bounds::Bounds` with its methods is nowadays a paper-thin wrapper around `Vec<(Clause, Span)>`s and `Vec::push` essentially.
Its existence slightly annoyed me (and I keep opening its corresp. file instead of the identically named `bounds.rs` in `hir_ty_lowering/` that I actually want most of the time :P).
Opening to check if you agree with inlining it.
r? compiler-errors or reassign
Replace some u64 hashes with Hash64
I introduced the Hash64 and Hash128 types in https://github.com/rust-lang/rust/pull/110083, essentially as a mechanism to prevent hashes from landing in our leb128 encoding paths. If you just have a u64 or u128 field in a struct then derive Encodable/Decodable, that number gets leb128 encoding. So if you need to store a hash or some other value which behaves very close to a hash, don't store it as a u64.
This reverts part of https://github.com/rust-lang/rust/pull/117603, which turned an encoded Hash64 into a u64.
Based on https://github.com/rust-lang/rust/pull/110083, I don't expect this to be perf-sensitive on its own, though I expect that it may help stabilize some of the small rmeta size fluctuations we currently see in perf reports.
Fix const items not being allowed to be called `r#move` or `r#static`
Because of an ambiguity with const closures, the parser needs to ensure that for a const item, the `const` keyword isn't followed by a `move` or `static` keyword, as that would indicate a const closure:
```rust
fn main() {
const move // ...
}
```
This check did not take raw identifiers into account, therefore being unable to distinguish between `const move` and `const r#move`. The latter is obviously not a const closure, so it should be allowed as a const item.
This fixes the check in the parser to only treat `const ...` as a const closure if it's followed by the *proper keyword*, and not a raw identifier.
Additionally, this adds a large test that tests for all raw identifiers in all kinds of positions, including `const`, to prevent issues like this one from occurring again.
fixes#137128
Use `const_error!` when possible
Replace usages of `io::Error::new(io::ErrorKind::Variant, "constant string")` with `io::const_error!(io::ErrorKind::Variant, "constant string")` to avoid allocations when possible. Additionally, fix `&&str` error messages in SGX and missing/misplaced trailing commas in `const_error!`.
Pattern Migration 2024: clean up and comment
This follows up on #136577 by moving the pattern migration logic to its own module, removing a bit of unnecessary complexity, and adding comments. Since there's quite a bit of pattern migration logic now (and potentially more in #136496), I think it makes sense to keep it separate from THIR construction, at least as much as is convenient.
r? ``@Nadrieril``
Overhaul `rustc_middle::limits`
In particular, to make `pattern_complexity` work more like other limits, which then enables some other simplifications.
r? ``@Nadrieril``
There were two mistakes: first, tests were sorted before test modules, and second, we re-sorted based on the name only, which cancelled the sort based on the kind.
It's similar to the other limits, e.g. obtained via `get_limit`. So it
makes sense to handle it consistently with the other limits. We now use
`Limit`/`usize` in most places instead of `Option<usize>`, so we use
`Limit::new(usize::MAX)`/`usize::MAX` to emulate how `None` used to work.
The commit also adds `Limit::unlimited`.
And add a new diagnostic for non-`Fn` parenthesized generic args.
Path lowering started to look like a mess, with each function carrying additional parameters for the diagnostic callback (since paths can occur both in type and in expression/pattern position, and their diagnostic handling is different) and the segment index, for the diagnostics report. So I refactored it from stateless functions on `TyLoweringContext` into stateful struct, `PathLoweringContext`, that tracks the process of lowering a path from resolution til assoc types selection.
