The following test shows two problems.

One, the date of the Gregorian cutover changed between JDK 1.4 and later JDKs.

Two, creating a Date from Calendar.getTime(), calling toString on it, then calling getTime on the Date, returns a different time than the Date originally started with.

public class GregoTest {
     public static void main(String args[]) {
         System.out.println("Java:" + System.getProperty("java.version"));
         java.util.TimeZone jdkGMT = java.util.TimeZone.getTimeZone("GMT");
// java.util.TimeZone.setDefault(jdkGMT);
         java.util.Calendar jdkCal = java.util.Calendar.getInstance(jdkGMT);
         jdkCal.clear();
         jdkCal.set(1582, java.util.Calendar.OCTOBER, 15);
         System.out.println("JDK time: " + jdkCal.getTime().getTime() );
         System.out.println("JDK time (str): " + jdkCal.getTime() );
         System.out.println("Day of month: " +
             jdkCal.get(java.util.Calendar.DAY_OF_MONTH));
         Date co = jdkCal.getTime();
         System.out.println("Change over (Oct 15 1582) = " + co + " (" +
             co.getTime() + ")");
         long a = jdkCal.getTime().getTime();
         Date c = jdkCal.getTime();
         c.toString();
         long b = c.getTime();

         if(a!=b) {
             System.out.println("ERR: " + a + " != " + b);
         } else {
             System.out.println(a + " = " + b);
         }
     }
}

###@###.### 2005-05-23 20:39:49 GMT