Name: iaR10016 Date: 05/04/2000



JDK version:
java version "1.3.0beta"
Java(TM) 2 Runtime Environment, Standard Edition (build 1.3.0beta-b03)
Java HotSpot(TM) Client VM (build 1.3.0beta-b01, mixed mode)

The following test example shows the incorrect floating-point
calculations in JDK 1.3.

Correct result for (double)(1.1107651257113993E-16+1.) is
1.0000000000000002.

If the test runs with -Xcomp option, the result of calculations is
correct on
single-processor system and incorrect on multy-processor system.

In Linux JDK 1.2.2 and in Solaris JDK 1.2.2, 1.3 the result is correct.

------------------ test.java ----------------------------------

public class test {
    public static void main (String args[]) {
        double x = 1.1107651257113993E-16;
        double y = 1.;
        System.out.println(x+y);
    }
}

--------------- Output (Solaris JDK 1.3) ----------------------

Correct result:

$ java test
1.0000000000000002
$

---- Output (Linux JDK 1.3, multy-processor system) -----------

Incorrect result:

$ java test
1.0
$ java -Xint test
1.0
#
# HotSpot Virtual Machine Error, Unexpected Signal 11
# Please report this error at
# http://java.sun.com/cgi-bin/bugreport.cgi
#
# Error ID: 4F533F4C494E55580E43505005D8
#
# Problematic Thread: prio=1 tid=0x8097388 nid=0x400 runnable
#
$ java -Xcomp test
1.0
$

------ Output (Linux JDK 1.3, single-processor system) --------

Incorrect result:

$ java test
1.0
$ java -Xint test
1.0


Correct result:

$ java -Xcomp test
1.0000000000000002
$

JCK 1.3 compiler tests /lang/FP/lang/FP/fpl035/fpl03502m2/fpl03502m2
and lang/FP/fpl035/fpl03502m4/fpl03502m4 fail if Linux JDK 1.3 because
of this bug.

======================================================================