Terse diagnostic "unexpected end of macro invocation" when misusing `assert_eq!`

[jruderman]

Given the following code (playground):

fn main() {
    assert_eq!(1 + 2 == 3);
}

Current output: terse

error: unexpected end of macro invocation
 --> src/main.rs:2:26
  |
2 |     assert_eq!(1 + 2 == 3);
  |                          ^ missing tokens in macro arguments

Better output: show macro correspondence

note:          This expression was treated as the $left argument to assert_eq!
  |
2 |     assert_eq!(1 + 2 == 3);
  |                ^^^^^^^^^^

note:          Following $left, assert_eq! expects ","

(See also #97108, which also asks for more verbose "missing tokens in macro arguments" errors.)

Ideal output: fix my mistake

help: To assert equality, separate the two expressions with a comma:
  |
2 |     assert_eq!(1 + 2, 3);
  |                     ^

(See also #57507, which asks for specific diagnostics when making a similar mistake with the assert! macro.)

[lyming2007]

@rustbot claim

[lyming2007]

I'm afraid we can't suggest some thing like:

help: To assert equality, separate the two expressions with a comma:
  |
2 |     assert_eq!(1 + 2, 3);
  |                     ^

Because 1 + 2 == 3 is a legal expression for RUST so we don't know if the programmer really wanted to write code like:
assert_eq!(1+2 == 3, true); or assert_eq!(1+2 , 3)

[jruderman]

I think it would be okay to suggest the more likely intended code.

Otoh, Clippy has disabled lints suggesting to replace assert with assert_eq in anticipation of Rust changes. Perhaps it would be better to hold off for now on suggestions that commit to preferring a particular assert syntax.

[jruderman]

#103439 will improve diagnostics for macro use by showing what token or ast node was expected :)

[Noratrieb]

With that PR, the new error is

error: unexpected end of macro invocation
 --> aaa.rs:2:26
  |
2 |     assert_eq!(1 + 2 == 3);
  |                          ^ missing tokens in macro arguments
  |
note: while trying to match `,`

error: aborting due to previous error

which is better but still not great