Support for type arguments in redirecting factory constuctors

[scheglov]

The kernel I get from front-end has invalid-expression and does not include type arguments.

class A<T> implements B<T> {
  A(int a);
  A.named(double a);
}
class B<U> {
  factory B.one(int b) = A<U>;
  factory B.two(double b) = A<U>.named;
}

library;
import self as self;
import "dart:core" as core;

class A<T extends core::Object> extends core::Object implements self::B<self::A::T> {
  constructor •(core::int a) → void
    : super core::Object::•()
    ;
  constructor named(core::double a) → void
    : super core::Object::•()
    ;
}
class B<U extends core::Object> extends core::Object {
  static field dynamic _redirecting# = <dynamic>[self::B::one, self::B::two];
  static factory one<U extends core::Object>(core::int b) → self::B<self::B::one::U>
    let dynamic #redirecting_factory = self::A::• in invalid-expression;
  static factory two<U extends core::Object>(core::double b) → self::B<self::B::two::U>
    let dynamic #redirecting_factory = self::A::named in invalid-expression;
}

See also #30258

[eernstg]

You may have noted that the grammar in the language specification does not allow for passing type arguments after = in a redirecting factory constructor declaration. Could that in some indirect manner be the source of the problem?

In any case, the specification of checks applied to redirecting constructors have been updated substantially in this CL, which also adds typeParameters? to the syntax.

[scheglov]

Hm... I think the specification allows type arguments in redirection.
See http://www.ecma-international.org/publications/files/ECMA-ST/ECMA-408.pdf

[eernstg]

Good catch, it does indeed allow type arguments already, and the grammar change is reverted now.

[chloestefantsova]

Fixed in ec9c927.