Overriding functions with Dynamic param

[kevinresol]

class Main extends Test
{
    static function main(){}
    override function test(_, _){}
}

class Test
{
    public function new(){}
    public function test(a:Int, b:Dynamic){}
}

Fails with: Field test overloads parent class with different or incomplete type (Haxe: b0b16f2)

Compiles fine when changing type of b to non-Dynamic, or specify (_, b:Dynamic) in the override

[Simn]

That happens because monomorphs are not bound to Dynamic. Not worth fixing in my opinion, it's good practice to type-hint your function arguments and not doing so can lead to other inference issues as well.

[Gama11]

Would it help if there were special semantics for _? If the programmer says "I don't care about the argument", that includes "I don't care about the argument type", so it seems annoying that you'd have to specify it.

[Simn]

Do you mean automatically inferring _ to be Dynamic?

[kevinresol]

btw this is a breaking change because it worked in 3.2.1

[Gama11]

Perhaps, yeah? Although that seems potentially dangerous if the compiler doesn't also make sure _ really is unused.

[kevinresol]

Oh man... I thought _ has a special semantic meaning from the beginning and that explains why I can put two parameters of the same name. And I just found that I have been WRONG. I can actually write a function like function test(a:Int, a:Dynamic){} ....

[Gama11]

Yeah, that seems like something that can introduce subtle bugs. Maybe it's a good idea to only allow that for _ and if _ is unused?

[Simn]

@kevinresol: Yes, that comes from fixing #4378 which intentionally allowed Dynamic variance. I'll have to check the regression though.

[Simn]

@Gama11: See #4381.

[kevinresol]

I am not sure what I expect here, given that _ actually doesn't have special meaning...
Please close this at will.

[Simn]

I still don't think it's worth fixing this, so I'll go ahead and close it.