All fixed-size array impls show up as `<unevaluated[]>`

[eira-fransham]

After #46882, when a trait bound fails and the trait was implemented by fixed-size arrays, the compiler will display an error like:

   = help: the following implementations were found:
             <[T; <unevaluated[]>] as Trait>
             <[T; <unevaluated[]>] as Trait>
             <[T; <unevaluated[]>] as Trait>
             <[T; <unevaluated[]>] as Trait>
           and 31 others

Which is pretty unhelpful. A solution would be to evaluate these and/or to show evaluated bounds preferably.

[oli-obk]

I can't reproduce triggering an error with these messages, do you have an example? I need it for a regression test, the fix was indeed as simple

to evaluate these and/or to show evaluated bounds preferably.

[eira-fransham]

It only appears if you explicitly disambiguate the trait, like <[T; N] as Foo>::foo(). Playground link here.

[estebank]

@oli-obk are you displaying these as <[(); _] as Foo>? I feel that'd be a reasonable output for this.

[oli-obk]

@estebank that would be a better fallback than <unevaluated[]>, but I'm currently trying to get at the actual value (though that triggers cycle errors atm when building libcore's array impls ^^)

[eira-fransham]

@estebank That's better in some ways, but kinda implies that for<const T> [(); T]: Foo