Inlining/Extension/Inheritance issue

[pbroadbery]

.. this comes from investingating issues with Complex Integer.

Start with a domain like:

    Complex(T: ..): with {
      *: (T, %) -> %
      *: (I, %) -> % 
    }

Now inherit these definitions...

    extend Complex(T: ...) with {
       f: (T, %) -> %
    } == add {
      f(t: T, x: %): % == t * x;
    }

All good so far.

   f$Complex I

Now call f. If the compiler has chosen to inline "_" in the definition of f, it hopes to get the first definition of _. Unfortunately, both definitions of "*" have the same hash code in the case of Complex I, and if we get the second, mayhem will result.

For a real example, see the initial version of reciprocal in sal_complex.as.

[hemmecke]

Why is this a particular problem with inlining? If the semantics of *: (T, %) -> % and *: (I, %) -> % is different for T=I, then that is problematic anyway. I'm not aware of the fact that in Aldor one can have two different actions (I, %) -> % that are denoted by the same identifier *.