The Java VM crashes with segmentation Fault

Name: ddT132432 Date: 09/17/2001


java version "1.3.1"
Java(TM) 2 Runtime Environment, Standard Edition (build 1.3.1-b24)
Java HotSpot(TM) Client VM (build 1.3.1-b24, mixed mode)

This problem still occurs in JDK1.3.1
The VM is causing at random times segmentation faults with our application that
used to run fine on JDK1.1.8-9b on the same machine.
The same application with JDK 1.3.02 was causing the VM to abort with the
following error message:
#
# HotSpot Virtual Machine Error, Unexpected Signal 10
# Please report this error at
# http://java.sun.com/cgi-bin/bugreport.cgi
#
# Error ID: 4F533F534F4C415249530E43505007A6 01
#
# Problematic Thread: prio=5 tid=0x17b400 nid=0x38 runnable
#
Abort (core dumped)

 and we submitted that as a bug to you (actually many people saw that
error in 1.3.02, so that our report was considered a duplicate and we were
advised to upgrade to JDK 1.3.1, where the problem was supposed to disappear).

Description of circumstances that make the bug appear:
Solaris Sun OS 2.6
Entreprise 250 (2x400MHz UltraSparc III - we were not able to reproduce the
problem on Ultra 10 1x450 UltraSparc III)
The application that is causing java to crash is a pure java application:
Part of it was compiled under java 1.3.1 part of it is ibm's xml4j.jar XML
parser that was probably compiled under java 1.1.8.
Our code compiled with java 1.1.8 is not running under java 1.3.x correctly,
throwing an exception about a register containing an invalid value.
Our application is multithreaded (about 60 threads created at startup that
are never supposed to exit).
We have 20 threads accepting connections on the same ServerSocket.
The main java process (the one that crashes) launches 12 child processes with
which it communicates through standard in and standard out (each of the
processes is accessed from any of the threads).
The application is rather big and complex so we cannot send it to you.
We have core files available.
The crash happens rarely very quickly (minutes), most of the times it takes a
couple of days until the crash occurs.
We understand that most likely you cannot find nor fix the bug from this
description but maybe you could take a look at the core files and help us
with information that would allow us to narrow down the scope of the problem.
We have some classes that have only static members.
We use Hotspot, -Xss2m -Xfuture and -Xnoclassgc.

TEST CASE HAS BEEN ATTACHED TO INCIDENT
(Review ID: 130573)
======================================================================