As reported on:

http://stackoverflow.com/questions/10620680/why-volatile-in-java-5-doesnt-synchronize-cached-copies-of-variables-with-main

and discussed on:

http://cs.oswego.edu/pipermail/concurrency-interest/2012-May/009440.html

and

http://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/2012-May/007704.html

The following test program reports "Error" where under the Java Memory Model rules it never should:

public class VolatileBug {
    volatile static private int a;
    static private int b;

    public static void main(String [] args) throws Exception {
        for (int i = 0; i < 100; i++) {
            new Thread() {

                @Override
                public void run() {
                    int tt = b; // makes the jvm cache the value of b

                    while (a==0) {

                    }

                    if (b == 0) {
                        System.out.println("error");
                    }
                }

            }.start();
        }

        b = 1;
        a = 1;
    }
}

Investigation shows the problem is in the GlobalValueNumbering code where the load of the volatile field doesn't kill the prevous load of a non-volatile field, resulting in the old value incorrectly being re-used after the volatile field is read.