rewrittenUninitializedType(uninitializedThis, Environment,
                           MethodClass, MethodClass) :-
    MethodClass = class(MethodClassName, CurrentLoader),
    thisClass(Environment, class(thisClassName, thisLoader)),
    superclassChain(thisClassName, thisLoader, [MethodClass | Rest]).

This says that, for an invokespecial of a superclass <init> method, the type 'unintializedThis' should be replaced with 'MethodClass', the type of the *superclass*. But that's incorrect: the new type should be the type of the *current class*.

This bug appears to have been introduced by JDK-7160765, which recognized the need to constrain the invocation to the current class or the superclass. But it inadvertently changed the result type at the same time.

Ever since JDK-8138821, which spun off the 'uninitializedThis' case separately, there hasn't really been a need for a 'rewrittenUninitializedType' predicate that works on all uninitialized types. It may make sense to strip this check down to a simpler 'validInvokespecialTarget' or something like that, and leave it to the 'instructionIsTypeSafe' rule to always map 'uninitializedThis' to 'CurrentClass'.