1.4.2_06 Hotspot Considered Harmful

Question

The following 23-line program, containing no library calls and nothing even remotely "tricky", consitently crashes the JVM on every Linux or Windows box we ve tried it on. We tried filing a bug report, but since someone once fixed a problem (5009717) generating the same error code some time back in an early 1.5 beta, this is a "duplicate" (despite the fact that it's on a 1.4 build apparently released after the previous bug was closed). A later attachment to the above bug report suggests that at least one other user has hit this, though.
It is possible to make the problem "disappear" by rearranging the code, but our testing suggests that how long it takes to crop up is a function of Hotspot's optimization heuristics, so it's not easy be sure if you've avoided the problem or just pushed it one iteration past the number of times you've tested some particular path. In other words, you can't trust it.
We're telling our customers to avoid 1.4.2_06 entirely because of this issue. We are big Java advocates in our company and industry and find it most painful and embarassing to have to tell people using our software that the (at this writing) current version of the Java 1.4 environment is
unfit for use and we don't know when (or if) it will ever be fixed.
// run this on 1.4.2_06
public class GoodbyeWorld {
  public static double func1(){
    double r = 0;
    for (int i = 0; i < 1; i++)
      r += func2();
    return r;
  public static double func2(){
    double[] a = {0};
    double r = 0;
    for (int i = 0; i < 1; i++)
      r = a;
return r;
public static void main(String[] argv) {
for (int i = 0; i < 1000; i++)
func1(); // System.out.println("i["+i+"]=" + func1());
$ java -version
java version "1.4.2_06"
Java(TM) 2 Runtime Environment, Standard Edition (build 1.4.2_06-b03)
Java HotSpot(TM) Client VM (build 1.4.2_06-b03, mixed mode)
$ java GoodbyeWorld
# HotSpot Virtual Machine Error, Internal Error
# Please report this error at
# http://java.sun.com/cgi-bin/bugreport.cgi
# Java VM: Java HotSpot(TM) Client VM (1.4.2_06-b03 mixed mode)
# Error ID: 43113F2652414D452D41503F491418160E435050005C
# Problematic Thread: prio=1 tid=0x080a8eb0 nid=0x7954 runnable
Heap at VM Abort:
Heap
def new generation   total 576K, used 156K [0x44770000, 0x44810000, 0x44c50000)
  eden space 512K,  30% used [0x44770000, 0x447971c0, 0x447f0000)
  from space 64K,   0% used [0x447f0000, 0x447f0000, 0x44800000)
  to   space 64K,   0% used [0x44800000, 0x44800000, 0x44810000)
tenured generation   total 1408K, used 0K [0x44c50000, 0x44db0000, 0x48770000)
   the space 1408K,   0% used [0x44c50000, 0x44c50000, 0x44c50200, 0x44db0000)
compacting perm gen  total 4096K, used 975K [0x48770000, 0x48b70000, 0x4c770000)
   the space 4096K,  23% used [0x48770000, 0x48863f80, 0x48864000, 0x48b70000)

Answer 1

Thank you for the test case. I was able to reproduce this on solaris-i586, windows-i586, and on linux-i586. The crash is specific to the -client compiler and the i586 architecture. The expanded text of the Error-ID string is:
    43113F2652414D452D41503F491418160E435050005C
Decodes to:
    c1_FrameMap_i486.cpp, 92The test case works fine using -server (or -Xint) on i586, and it works fine using both -client and -server on SPARC systems.
I opened Bug-ID 6215242 on this issue. Look for is on the Bug-Parade in a day or two.