There is nothing I can find in the specs about how protective attributes change under the constructs known to D and the reference compiler does not follow my assumptions on what is plausible.

For example, if I use more than one protective attribute in a chain of accesses to a member of a class, i find it plausible, that the maximal protection found in the chain will be applied. But this is not true for the reference compiler:

`protected' C2.i is reachable through a `private' alias, whereas `private' C1.i is not reachable although made protected by a `protected' alias.

Therefore neither minimal nor maximal protection found in a chain is used.

A further confusion hits me, when I realize, that one may write to C1.i through C2 and any other class derived from C2 after a maintenance change, because they are in the same module, although one cannot reach C1.i from D1 or any other class derived from C1.

In essence this would mean, that after any maintenance the changed part must be put into a separate module to prevent accesses by mistake or criminal assault.

-manfred


file depth0.d
---------------------
class C1{
  private int i;
  protected alias i j;
}
class C2{
  protected int i;
  private alias i j;
}
---------------------

file depth1.d
---------------------
import depth0;

class D1: public C1{
  void f(){
    j= 1; // error: "i not accessible"
  }
}
class D2: public C2{
  void f(){
    j= 1; // OK
  }
}
--------------------