About overriding and overloading - Advanced Topic

Question

Please look at this thread
http://forum.java.sun.com/thread.jsp?forum=31&thread=159240
Someone posted a code snippet that I can't fully explain, perhaps some Java Guru can answer!

Answer 1

This is the program:
The question is why the compiler error in the indicated line but not when upcating b to Homer.
<code>
class Homer {
     float doh(float f) {
          System.out.println("doh(float)");
          return 1.0f;
     char doh(char c) {
          System.out.println("doh(string)");
          return 'd';
class Bart extends Homer {
     float doh(float f) {
          System.out.println("doh(float)");
     return 1.0f;
class Hide {
     public static void main(String[] args) {
          Bart b = new Bart();
          b.doh('x');//compiler error in this line
          b.doh(1);
          b.doh(1.0f);
</code>
I looked for an answer in the Java Language Specification and I found the section 15.12.2.2
I will try to sum up in the following:
Steps performed by the compiler to spawn the symbolic reference for a given method invocation:
1) It finds out which class or interface looked for the method declaration.
In the example b.doh('x'); is the type given of b, that is Bart.
2) Looks for the methods that are both applicable and accesible in the methods declared and inherited in that type.
A method declaration is applicable to a method invocation if they have the same number of parameters, and, the types of the arguments in the declaration are the same or can be converted via a widenning convertion to those of the declaration.
Whether a method declaration is accessible to a expression of method invocation depends on the modifiers specified in the declaration and where the invocation occurs.
If there are several methods declaration both applicable and accessible the one more specific is chosen. If no more specific method exist, the compiler generates an error.
This how to find out which method is more specific:
Let T be the class where a method m is declared, and let U be another class where a method called as well m is declared. Both have the same number of parameters because they are applicable, but they have different types in the argument list. If it happens that T is a subclass of (or the same as) U, and, all the parameters types in the m declared in T can be converted via widenning convertion to those in the m declared in U (or are the same); the m in T is more specific than the m in U.
3) The checks described in the section 15.2.3 in the JLS are performed. They do not apply in our example.
Back to the example, the compiler finds two applicable and accessible methods in Bart:
float doh(float f)      declared in Bart
char doh(char c)     declared in Homer     
The last one inhertited from Homer.
So which is the more specific? let�s see first if doh declared in Bart is more specific than doh declared in Homer: Bart is a subclass of Homer, but float is not of the same type nor widenning convertible to char . Now, if doh in Homer is more specific than doh in Bart: well, Homer is not the same nor widenning convertible to bart. So there is no a more specific method and the compiler complains.
If you apply these rules to b upcasted to Homer doh(char) is more specific. The same happens if doh(char) is added to Bart.