Static extensions vs. constrained monomorphs

[Simn]

using Main;

class Main {
	static public function main() {
		var obj = merge({foo: 12}, {bar: "bar"});
		obj.staticExtension();
	}

	static function merge<A:{}, B:{}, C:A & B>(a:A, b:B):C {
		return null;
	}

	static function staticExtension(a:{foo:Int, bar:String}) {}
}

source/Main.hx:6: characters 7-22 : Unknown<0> : { foo : Int, bar : String } has no field staticExtension

This is probably just a missing using_field in the TMono code path of type_field.

[Simn]

I thought I found a regression with code like this:

using Main;

class Main {
	static public function main() {
		infer({foo: 12});
	}

	static function infer(a) {
		a.foo = 12;
		a.staticExtension();
	}

	static function staticExtension(a:{foo:Int}) {}
}

source/Main.hx:5: characters 9-18 : Object requires field staticExtension
source/Main.hx:5: characters 9-18 : For function argument 'a'

But that was failing before as well. I guess it makes sense that it does.

However, completion shows staticExtension as a field on a, so there's some discrepancy here.

[frabbit]

It would be great to have this working:

static function merge<A:{}, B:{}>(a:A, b:B):A & B {
		return null;
	}

It's different to the constraint C Example, because C opens the type on the call side which means 

var x: { foo: Int, bar: Int, ohno: Int } = merge({ foo: 1}, { bar: 2});

compiles just fine.