Name: tb29552 Date: 09/11/98


/* The problem is that an expression of the form
 *
 * ClassName.staticFieldName.staticMethodName()
 *
 * is (sometimes?) not executed correctly.
 *
 *The problem can be reproduced with the following code:
 */
public class StaticBug
{
    public
        StaticBug()
        {
            System.err.println("Constructing an instance of class StaticBug");
        }
    
    public static void
        main(String[] args)
        {
            SomeClass.someStaticField.someStaticMethod();
        
            // Everything goes allright when the following two
            // lines are used instead:
            //
            //StaticBug b = SomeClass.someStaticField;
            //b.someStaticMethod();
        }
    
    public static void
        someStaticMethod()
        {
            System.err.println("In StaticBug.someStaticMethod()");
        }
}

class SomeClass
{
    static
        {
            System.err.println("Initializing class SomeClass");
        }
        
    public static StaticBug someStaticField = new StaticBug();
}
/*
 *
 * The output should be:
 *
 * Initializing class SomeClass
 * Constructing an instance of class StaticBug
 * In StaticBug.someStaticMethod()
 *
 * but it is:
 *
 * In StaticBug.someStaticMethod()
 *
 * So class SomeClass is not initialized.
 *
 * According to the JLS (First printing, August 1996)
 * section 15.11.4.1, the expression
 * 'SomeClass.someStaticField' should be "evaluated, but
 * the result is then discarded." To me, this implies
 * that SomeClass is loaded and initialized, and that an
 * instance of StaticBug is created.
 *
 * As shown above, if the single statement is split
 * into two parts, everything goes as specified.
 *
 * I marked this bug as a compiler bug, but I'm not
 * sure. Since all Java compiler/VM combinations I've
 * tried behave the same, it may be a VM or VM spec bug.
 */

(Review ID: 37734)
======================================================================