Adapt to the state when FunctionTypeAliasElement does not implement FunctionTypedElement.

[scheglov]

No description provided.

[jcollins-g]

I think ModelElement.modelType needs a small additional adjustment along the same lines.

I am pretty certain ModelFunctionTypedef's construction isn't going to work right when the eventual change lands, and am surprised that the tests pass. We explicitly call ModelFunctionTypedef's constructor with a FunctionTypeAliasElement in the giant ModelElement.from dispatching constructor, and its constructor expects a FunctionTypedElement.

[scheglov]

1. It works in ModelElement.modelType because FunctionTypeAliasElement still implements TypeDefiningElement. It'd like it to stop doing this, but this is a bigger task than I'd like to do now (specifically there are many references in DDC, and I hope that DDC will die off sooner than I have to change it :-)). I still might do some tweaks there though, because I'd like to remove TypeDefiningElement.type (yes, type defining, but without a type :-)).

2. Places where ModelFunctionTypedef is created, are inside if (e is GenericFunctionTypeElement), which just cannot be true both, at least now. It might have been the case in the past, but now FunctionTypeAliasElement does not implement GenericFunctionTypeElement. So, we hit the next block if (e is FunctionTypeAliasElement) and create Typedef instead, which sounds like a good name :-)

[jcollins-g]

1. It works in ModelElement.modelType because FunctionTypeAliasElement still implements TypeDefiningElement. It'd like it to stop doing this, but this is a bigger task than I'd like to do now (specifically there are many references in DDC, and I hope that DDC will die off sooner than I have to change it :-)). I still might do some tweaks there though, because I'd like to remove TypeDefiningElement.type (yes, type defining, but without a type :-)).

OK.

2. Places where ModelFunctionTypedef is created, are inside if (e is GenericFunctionTypeElement), which just cannot be true both, at least now. It might have been the case in the past, but now FunctionTypeAliasElement does not implement GenericFunctionTypeElement. So, we hit the next block if (e is FunctionTypeAliasElement) and create Typedef instead, which sounds like a good name :-)

If you can drop a todo for me to clean this up I'd appreciate it. Otherwise, sounds good.