Reimplement and complete spanless_eq for the AST

[Jarcho]

This replaces the many exported functions that used to exist with a single spanless_eq function. The macros implementing this would be derives, but that's not possible so nearly copying the struct/enum definitions is required.

The new implementation has full exhaustiveness checking on both fields and variants. This way any changes made to the AST in rustc will require changes here.

changelog: none

[rustbot]

r? @Alexendoo

rustbot has assigned @Alexendoo.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[Alexendoo]

Could we use ${concat} here to allow a similar syntax to the struct definitions?

[Jarcho]

I ran into hygiene issues when I tried it the first time, but it might be a thing I can work around.

[Jarcho]

Tried some other options. The feature is just not implemented enough currently to work.

[Jarcho]

concat doesn't work in nested repetitions and pushing the concat into an inner macro means the generated identifier can only the inner macro can access the name due to hygiene rules. Basically neither of these work:

// with repetitions: ice
$(
  (
    $var { $($name: ${concat(l, $name)}),*  },
    $var { $($name: ${concat(r, $name)}),*  },
  ) => $(eq!(${concat(l, $name)} ${concat(r, $name)}))&&*,
)*

// with inner: binding names can't be used outside the `name` macro
$(
  (
    $var { $($name: name!(l, $name)),*  },
    $var { $($name: name!(r, $name)),*  },
  ) => $(eq!($name))&&*,
)*

[Alexendoo]

We could accept it as is then, but I'm worried the syntax will be confusing for rustc contributions

A temporary paste dependency could also work, it's a pretty small dependency at least

[rustbot]

☔ The latest upstream changes (possibly #15773) made this pull request unmergeable. Please resolve the merge conflicts.