Announcing Rust 1.89.0

Aug. 7, 2025 · The Rust Release Team

The Rust team is happy to announce a new version of Rust, 1.89.0. Rust is a programming language empowering everyone to build reliable and efficient software.

If you have a previous version of Rust installed via rustup, you can get 1.89.0 with:

$ rustup update stable

If you don't have it already, you can get rustup from the appropriate page on our website, and check out the detailed release notes for 1.89.0.

If you'd like to help us out by testing future releases, you might consider updating locally to use the beta channel (rustup default beta) or the nightly channel (rustup default nightly). Please report any bugs you might come across!

What's in 1.89.0 stable

Explicitly inferred arguments to const generics

Rust now supports _ as an argument to const generic parameters, inferring the value from surrounding context:

pub fn all_false<const LEN: usize>() -> [bool; LEN] {
  [false; _]
}

Similar to the rules for when _ is permitted as a type, _ is not permitted as an argument to const generics when in a signature:

// This is not allowed
pub const fn all_false<const LEN: usize>() -> [bool; _] {
  [false; LEN]
}

// Neither is this
pub const ALL_FALSE: [bool; _] = all_false::<10>();

Mismatched lifetime syntaxes lint

Lifetime elision in function signatures is an ergonomic aspect of the Rust language, but it can also be a stumbling point for newcomers and experts alike. This is especially true when lifetimes are inferred in types where it isn't syntactically obvious that a lifetime is even present:

// The returned type `std::slice::Iter` has a lifetime, 
// but there's no visual indication of that.
//
// Lifetime elision infers the lifetime of the return 
// type to be the same as that of `scores`.
fn items(scores: &[u8]) -> std::slice::Iter<u8> {
   scores.iter()
}

Code like this will now produce a warning by default:

warning: hiding a lifetime that's elided elsewhere is confusing
 --> src/lib.rs:1:18
  |
1 | fn items(scores: &[u8]) -> std::slice::Iter<u8> {
  |                  ^^^^^     -------------------- the same lifetime is hidden here
  |                  |
  |                  the lifetime is elided here
  |
  = help: the same lifetime is referred to in inconsistent ways, making the signature confusing
  = note: `#[warn(mismatched_lifetime_syntaxes)]` on by default
help: use `'_` for type paths
  |
1 | fn items(scores: &[u8]) -> std::slice::Iter<'_, u8> {
  |                                             +++

We first attempted to improve this situation back in 2018 as part of the rust_2018_idioms lint group, but strong feedback about the elided_lifetimes_in_paths lint showed that it was too blunt of a hammer as it warns about lifetimes which don't matter to understand the function:

use std::fmt;

struct Greeting;

impl fmt::Display for Greeting {
    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
        //                -----^^^^^^^^^ expected lifetime parameter
        // Knowing that `Formatter` has a lifetime does not help the programmer
        "howdy".fmt(f)
    }
}

We then realized that the confusion we want to eliminate occurs when both

  1. lifetime elision inference rules connect an input lifetime to an output lifetime
  2. it's not syntactically obvious that a lifetime exists

There are two pieces of Rust syntax that indicate that a lifetime exists: & and ', with ' being subdivided into the inferred lifetime '_ and named lifetimes 'a. When a type uses a named lifetime, lifetime elision will not infer a lifetime for that type. Using these criteria, we can construct three groups:

Self-evident it has a lifetimeAllow lifetime elision to infer a lifetimeExamples
NoYesContainsLifetime
YesYes&T, &'_ T, ContainsLifetime<'_>
YesNo&'a T, ContainsLifetime<'a>

The mismatched_lifetime_syntaxes lint checks that the inputs and outputs of a function belong to the same group. For the initial motivating example above, &[u8] falls into the second group while std::slice::Iter<u8> falls into the first group. We say that the lifetimes in the first group are hidden.

Because the input and output lifetimes belong to different groups, the lint will warn about this function, reducing confusion about when a value has a meaningful lifetime that isn't visually obvious.

The mismatched_lifetime_syntaxes lint supersedes the elided_named_lifetimes lint, which did something similar for named lifetimes specifically.

Future work on the elided_lifetimes_in_paths lint intends to split it into more focused sub-lints with an eye to warning about a subset of them eventually.

More x86 target features

The target_feature attribute now supports the sha512, sm3, sm4, kl and widekl target features on x86. Additionally a number of avx512 intrinsics and target features are also supported on x86:

#[target_feature(enable = "avx512bw")]
pub fn cool_simd_code(/* .. */) -> /* ... */ {
    /* ... */
}

Cross-compiled doctests

Doctests will now be tested when running cargo test --doc --target other_target, this may result in some amount of breakage due to would-be-failing doctests now being tested.

Failing tests can be disabled by annotating the doctest with ignore-<target> (docs):

/// ```ignore-x86_64
/// panic!("something")
/// ```
pub fn my_function() { }

i128 and u128 in extern "C" functions

i128 and u128 no longer trigger the improper_ctypes_definitions lint, meaning these types may be used in extern "C" functions without warning. This comes with some caveats:

  • The Rust types are ABI- and layout-compatible with (unsigned) __int128 in C when the type is available.
  • On platforms where __int128 is not available, i128 and u128 do not necessarily align with any C type.
  • i128 is not necessarily compatible with _BitInt(128) on any platform, because _BitInt(128) and __int128 may not have the same ABI (as is the case on x86-64).

This is the last bit of follow up to the layout changes from last year: https://blog.rust-lang.org/2024/03/30/i128-layout-update.

Demoting x86_64-apple-darwin to Tier 2 with host tools

GitHub will soon discontinue providing free macOS x86_64 runners for public repositories. Apple has also announced their plans for discontinuing support for the x86_64 architecture.

In accordance with these changes, the Rust project is in the process of demoting the x86_64-apple-darwin target from Tier 1 with host tools to Tier 2 with host tools. This means that the target, including tools like rustc and cargo, will be guaranteed to build but is not guaranteed to pass our automated test suite.

We expect that the RFC for the demotion to Tier 2 with host tools will be accepted between the releases of Rust 1.89 and 1.90, which means that Rust 1.89 will be the last release of Rust where x86_64-apple-darwin is a Tier 1 target.

For users, this change will not immediately cause impact. Builds of both the standard library and the compiler will still be distributed by the Rust Project for use via rustup or alternative installation methods while the target remains at Tier 2. Over time, it's likely that reduced test coverage for this target will cause things to break or fall out of compatibility with no further announcements.

Standards Compliant C ABI on the wasm32-unknown-unknown target

extern "C" functions on the wasm32-unknown-unknown target now have a standards compliant ABI. See this blog post for more information: https://blog.rust-lang.org/2025/04/04/c-abi-changes-for-wasm32-unknown-unknown.

Platform Support

Refer to Rust’s platform support page for more information on Rust’s tiered platform support.

Stabilized APIs

These previously stable APIs are now stable in const contexts:

Other changes

Check out everything that changed in Rust, Cargo, and Clippy.

Contributors to 1.89.0

Many people came together to create Rust 1.89.0. We couldn't have done it without all of you. Thanks!