rustfmt thinks it should be able to modify string values

[matthiaskrgr]

fn main() {
    println!(
        "Hello, world!
a    
    a
"
    );
}

note how line 2 inside the format string is actually "a ".
apparently rustfmt things it should be able to change this which is scary because it could potentially alter program behaviour

cargo fmt -- --error-on-unformatted --unstable-features

error[internal]: left behind trailing whitespace
 --> /tmp/frustfmt/src/main.rs:4:4:2
  |
4 | a
  |  ^^^^
  |
  = note: set `error_on_unformatted = false` to suppress the warning against comments or string literals

[ytmimi]

@matthiaskrgr thanks for the report.

From what I can tell rustfmt doesn't modify the string lit. Running rustfmt with --check on your input snippet doesn't produce a diff:

rustfmt --config=error_on_unformatted=false --check

rustfmt is simply warning the user that there's trailing whitespace in the input that it was unable to remove, but as far as I know it will never try to modify the string lit.

[simonwuelker]

for completeness, the same issue also arises inside macro invocations , without the need to pass --error-on-unformatted :

fn main() {
    a!(
        aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()   
    ) 
}

(Not sure why the identifier has to be this long)

$ rustfmt src/lib.rs
error[internal]: left behind trailing whitespace
 --> <redacted>/src/lib.rs:3:3:109
  |
3 |         aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()   
  |                                                                                                             ^^^
  |

warning: rustfmt has failed to format. See previous 1 errors.

[ytmimi]

@Wuelle I want to point out that what you're experiencing is unrelated to the current issue. In your case there is trailing whitespace in the input file that rustfmt is unable to remove because of the long identifier name. Bumping the max_width to some sufficiently high value should alleviate your issue.

[JakkuSakura]

I think I encountered the same issue combined with #5711

error[internal]: left behind trailing whitespace
   |
11 |                                            &strategy.strategy_id as &(dyn ToSql + Sync), 
   |                                              ^
   |

#[cfg(test)]
mod tests {
    #[tokio::test]
    async fn test_foo() -> Result<()> {
        db.query("
                                                INSERT INTO tbl.bar 
                                                (fkey_strategy_id, pkey_id, blockchain, token_name, token_address, quantity, updated_at, created_at) 
                                                VALUES 
                                                ($1, $2, $3, $4, $5, $6, $7, $8);
                                                ", &[
                                                &strategy.strategy_id as &(dyn ToSql + Sync), 
                                ]).await?;

        Ok(())
    }
}

example2.zip

[ytmimi]

@qiujiangkun this is not the same issue. The trailing whitespace that rustfmt is warning you about isn't contained in the query stinrg, it's after the of &strategy.strategy_id as &(dyn ToSql + Sync),.

The following line is way to long to fit into the default width of 100, and therefore rustfmt is unable to format the chain db.query(...).await.

(fkey_strategy_id, pkey_id, blockchain, token_name, token_address, quantity, updated_at, created_at)

rustfmt doesn't modify whitespace within string literals so you could fix this problem by dedenting the SQL query or by setting the max_width to some sufficiently large value. In testing 200 worked.

[JakkuSakura]

I don't expect internal errors while formatting a valid program.
In my case, it refused to format the rest of the code.

I can accept behavior like this:
give a friendly non-internal warning and proceed to format the rest of the code