Suggest adding semicolon in user code rather than macro impl details

[Urgau]

This PR tries to find the right span (by peeling expansion) so that the suggestion for adding a semicolon is suggested in user code rather than in the expanded code (in the example a macro impl).

Fixes #139049
r? @fmease

[rustbot]

fmease is not on the review rotation at the moment.
They may take a while to respond.

[fmease]

Thanks! r=me with two nits addressed in one way or another

[Urgau]

@bors r=fmease

[bors]

📌 Commit 6ff3713 has been approved by fmease

It is now in the queue for this repository.

[m-ou-se]

@Urgau #140748 makes this just work, so this new test blocks me from merging that PR. Should I remove it? Is there a different way to test what you want to test?

No wait, the error still appears. Only the suggestion doesn't appear. Investigating..

[m-ou-se]

I just noticed that this test passes in Rust 2024, due to the new rules for temporaries in tail expressions.

[fmease]

I just noticed that this test passes in Rust 2024

Sure, but we still want to provide non-butchered / good diagnostics in older editions, they have "long-term support".

[fmease]

We could replace that test with the following one:

// Make sure the generated suggestion suggest editing the user code instead of
// the macro implementation (which might come from an external crate).
// issue: <https://github.com/rust-lang/rust/issues/139049>

// You could assume that this comes from an extern crate (it doesn't
// because an aux crate would be overkill for this test).
macro_rules! perform { ($e:expr) => { D(&$e).end() } }

fn main() {
    { let l = (); perform!(l) }; //~ ERROR does not live long enough
    //~^ SUGGESTION ;
    let _x = { let l = (); perform!(l) }; //~ ERROR does not live long enough
    //~| SUGGESTION let x
}

struct D<T>(T);
impl<T> Drop for D<T> { fn drop(&mut self) {} }
impl<T> D<T> { fn end(&self) -> String { String::new()} }

[fmease]

Or modify tests/ui/nll/issue-54556-used-vs-unused-tails.rs and add it there

[m-ou-se]

@fmease That test doesn't work. Then it just suggests to add the ; inside the macro definition.