Possible pretty-printing issue related to #1675

[stew]

When we try to overcome the issue in #1675 with a more explicit type:

ability Bar2 where fun : Nat -> {Bar2}(Nat -> Nat)

this produces:

ability Bar2 where fun : Nat ->{Bar2} Nat -> Nat

but we think this is a pretty-printing issue, because this works:

bar2handler = cases
 { Bar2.fun a -> k } -> handle k (x -> x + a) with bar2handler
 { a } -> a
> handle Bar2.fun 3 4 with bar2handler

(which tells us that the arity of the pattern match is what we wanted, not what it printed)

[anovstrup]

@stew Are those declarations different? I would expect the function type constructor -> to be right-associative, so that those two would be identical. The pattern arity is presumably determined by which argument requires the {Bar2} ability.