`cargo fix` doesn't apply suggestions from mismatched_lifetime_syntaxes

[kornelski]

Rust is about to start warning about this by default, which can trigger a lot of warnings, but unfortunately the warnings can't be fixed automatically:

#![warn(mismatched_lifetime_syntaxes)]

fn lifetimes(x: &str) -> std::borrow::Cow<str> {
    x.into()
}

Running these commands has no effect:

cargo fix
cargo fix --broken-code
cargo clippy --fix
cargo clippy --fix --broken-code

The code remains unfixed, and the warning keeps triggering.

---

cargo 1.90.0-nightly (9b296973b 2025-07-25)
rustc 1.90.0-nightly (f8e355c 2025-07-27)

[Noratrieb]

@shepmaster would you be interested in implementing a suggestion?

[shepmaster]

This is a deliberate decision; cargo fix is not yet capable of applying the suggestions that the lint is currently outputting. See #142998 and #142980 and ultimately #53934. Rust-Analyzer does support applying the fixes. I think there's work ongoing for an interactive cargo fix that may also theoretically support them.

As such, I don't think there's anything to be done for this issue and I'd vote to close it.

Running these commands has no effect

Note that cargo does not print the run `cargo fix ...` to apply 1 suggestion text, so that's expected.

[kornelski]

I'm surprised that it doesn't do that even with --broken-code, or that there isn't any other opt-in to apply the "maybe incorrect" suggestions.

I'm absolutely fine with invalid suggestions and broken syntax being applied. I know the suggestions are buggy, e.g. they suggest #'_ [attr] for macro-generated code, and I don't mind that at all. I still test and review these changes.

But I really really want to avoid the laborious process of finding and updating every lifetime by hand. Rust-analyzer doesn't solve it, because that is still a manual process that requires opening every file and clicking around.

I have hundreds of crates to update.

[shepmaster]

I'm absolutely fine with invalid suggestions and broken syntax being applied

avoid the laborious process of finding and updating every lifetime by hand

I'm pretty sure these would be in conflict with each other — by breaking the syntax on every warning site, you'd then have to visit each and fix it manually.

---

I have hundreds of crates to update.

A practical solution is to allow the lint — one change per crate. If the absence of the lint hasn't been causing you any problems to date across hundreds of crates, then you may not be the target audience for the lint.

On the flip side, I've been missing the core aspect of this new lint so much that I've habitually been denying elided_lifetimes_in_paths for years, even though that's overly broad. I could choose to reduce the pedanticness of my linting by switching to mismatched_lifetime_syntaxes.

---

that there isn't any other opt-in

As I understand the problem, we produce data of this shape

{
  suggestions: [
    [[place1, change1], [place2, change2]],
    [[place3, change3], [place4, change4]],
 ]
}

But when it gets to cargo fix, it's closer to

{
  suggestions: [
    [place1, change1], 
    [place2, change2],
    [place3, change3],
    [place4, change4],
  ]
}

This means if you try to apply the suggestions, they would interfere with each other. In this lint's case, the suggestions literally overlap at the exact same bytes, so cargo fix abandons the change.

[kornelski]

Thanks for the explanation.

Simply silencing the lint is a temporary solution, but I generally agree with these suggestions, so I'd prefer to fix the code instead.

I wasn't aware of the overlapping alternative suggestions, because they were also hidden from the user. If the problem is caused by overlapping alternatives, these alternatives can also be hidden from cargo fix in the same way — #144601