Postfix function type annotations don't make it past type inference

[hmontero1205]

In the following example:

foo (w: Int) (z: Int) -> Int =
  w + z

I had to annotate the function args "inline". The following equivalent:

foo w z: Int -> Int -> Int =
  w + z

Results in this error being thrown: TypeError "Cannot change empty Ann type to Classes type".

[hmontero1205]

@XijiaoLi , just a thought:

Should we have an explicit pass that attempts to unify the function args' type signatures with the postfix type signature (if one was provided)?

[XijiaoLi]

Sure, that sounds good to me.

[j-hui]

This looks like a bug with the lowering phase. I'll take a look tomorrow.

Should we have an explicit pass that attempts to unify the function args' type signatures with the postfix type signature (if one was provided)?

This shouldn't be necessary (as its own thing). The lowering phase should essentially distribute all the argument type annotations into a stack of type annotations. We just need to handle the general case of unifying multiple type annotations (which Xijiao and I already spoke about on Tuesday).

[j-hui]

This is being handled in #103