As everyone knows, private fields are not inherited between classes. What intrigues me, is how it works for inner static classes. Consider the following code:
Inner classes were a late entry to Java. When they were added they were added as a compiler extension only, no change to the JVM.
The language spec stated that an inner class was allowed to access the private members of the class that it was declared within; that includes other inner classes.
To make it work, the compiler generates bridge methods. Using javap with Main$A from the example above looks like this:
Notice the addition of access$200 and access$300. They provide back door access to the private method and field respectively.
class Main$A {
Main$A(Main$1);
Code:
0: aload_0
1: invokespecial #3 // Method "":()V
4: return
static void access$200(Main$A);
Code:
0: aload_0
1: invokespecial #2 // Method foo:()V
4: return
static int access$300(Main$A);
Code:
0: aload_0
1: getfield #1 // Field a:I
4: ireturn
}
For completeness, here is the generated code for Main$B. Notice the calls to access$200 and 300, they appear where super.a and super.foo() appeared in the Java code.
class Main$B extends Main$A {
public Main$B();
Code:
0: aload_0
1: aconst_null
2: invokespecial #1 // Method Main$A."":(LMain$1;)V
5: aload_0
6: invokestatic #2 // Method Main$A.access$100:(LMain$A;)V
9: getstatic #3 // Field java/lang/System.out:Ljava/io/PrintStream;
12: aload_0
13: invokestatic #4 // Method Main$A.access$200:(LMain$A;)I
16: invokevirtual #5 // Method java/io/PrintStream.println:(I)V
19: return
}
And if it is legal, why it is possible only via "super.XXX" construct?
private fields are not usually part of the compilers resolution path for fields, by forcing developers to specify super the compiler is making certain that private access is what was meant and not a mistake.