incorrect inference of `NamedTuple{(), <:Any}`

[JeffBezanson]

See https://github.com/JuliaLang/julia/pull/47059/files/b8a6b104b224da3efc9a85c975feeffc86681bd9#r1006058629

julia> (()->NamedTuple{(), <:Any})()
NamedTuple{(), Tuple{}}

[aviatesk]

I may be making a serious misunderstanding. My understanding is that NamedTuple{(), <:Any} can only represent NamedTuple{(), Tuple{}} (since NamedTuple{(), T} raises an error when T isn't Tuple{} making it valid for inference to assume NamedTuple{(), T} to be NamedTuple{(), Tuple{}} always.

[martinholters]

julia> NamedTuple{(), <:Any}
NamedTuple{(), <:Any}

julia> (() -> NamedTuple{(), <:Any})()
NamedTuple{(), Tuple{}}

We cannot create instances of NamedTuple{(), <:Any}, but we can still create the type.

[KristofferC]

I've reverted #47059 on 1.9 so moving the milestone.

[vtjnash]

It should probably be reverted on master too then, since the commit itself is faulty (closing this issue)