test case is attached
found in 1.4.1 and 1.3.1
Run this on a regular 1.4.1 or 1.3.1_03 and one of three things can happen:
1. It runs (about 25% of the time).
2. It crashes with a seggie on a lduh (the observed strcmp problem - about
25% of the time, usually when the count is at 49152)
3. It crashes with a seggie in a lw (a new and interesting problem - about
50% of the time, the count could be anything)

Sometimes it just works... c'est la vie. Sometimes it crashes with the lduh.
This doesn't happen in the patched build. Sometimes it crashes it what looks
like preamble code to the strcmp stuff. This looks like an optimizer problem
and happens on both releases, with monotonous regularity.

Run as "java -server -Xmx256M StringTest 2".


You will need fix for bug 4701709 from esc. 538410 prior to running in order to see this bug.


###@###.### 2002-06-27:

Note: When the customer says above that the problem does not occur in
      the "patched build" He is referring to a 1.3.1_hotspot_server_VM
      that I shipped
him with bug #4701709 fixed in... See Comments section.


###@###.### 2002-07-09:

Smaller test case. fails every time:

// $JAVA_HOME/bin/java_g -server -Xbatch -Xcomp -XX:+PrintCompilation -XX:CompileOnly=.test StringTest2
// Correctly running program will exit silently.
// Incorrect program SEGV's and gives "implicit exception happened..." message.

import java.util.*;

public class StringTest2 {
  static final int LEN = 100;
  static final int LAST = LEN-1;

  static void test(String[] myArray) {
    int i;
    for (i=0; i<LAST; i++) {
      if (myArray[i].compareTo(myArray[LAST]) != 0) {
System.out.println("D'oh!");
      }
    }
  }
  
  public static void main(String[] args) {
    String[] myArray = new String[LEN];
    String s = "foo";

    for (int j=0; j<LAST; j++) {
      if (j < LEN / 2) {
myArray[j] = s;
      } else {
myArray[j] = null;
      }
    }
    myArray[LAST] = s;
    
    try {
      test(myArray);
    } catch (NullPointerException e) {
    }
  }
}