Rewriting `r#try!` at root of chain generates bad code

[matthewhughes934]

Summary

Just recording something I ran into during my (ongoing) dive into chain formatting

I tried to format this code:

fn main() {
    let _ = r#try!(root)/* this is a comment*/.bar;
}

Expected behavior

It would be formatted to something like

fn main() {
    let _ = root? /* this is a comment*/
        .bar;
}

Actual behavior

It was formatted to invalid code:

fn main() {
    let _ = root? )/* this is a comment*/
        .bar;
}

Configuration

Only needs the use_try_shorthand option

Reproduction Steps

rustfmt --config use_try_shorthand=true ./file.rs

Meta

Also reproduced on 1141ee711fc87bf51ff8934c09a9d608b34e50bf (a recent HEAD of master)

$ rustfmt --version
rustfmt 1.10.0-nightly (bff8e12ff5 2026-08-26)

[jieyouxu]

Huh, I wonder if this is related to handling of raw ident spans 🤔

[ytmimi]

Could be! I think we'd need to investigate further to know for sure, but I wouldn't be surprised if this was a span issue.

[matthewhughes934]

I think it's because when parsing the chain items we call Chain::convert_try, that function calls convert_try_mac and that function just copies the original macro span. When formatting the chain we try to recover the comments from the item's span, and things get confused.

I discovered this during my (continued) poking around of the chain code, I'll keep looking to get a better idea.