The following comment has been added to this issue:
Author: Michael Good
Created: Mon, 18 Oct 2004 3:18 PM
Body:
I took another look at this with Java 5.0 on Windows XP SP2. We added an
-XX:+ShowMessageBoxOnError
flag to our JNI invocation. This causes the JVM to crash with the message:
EXCEPTION_FLT_STACK_CHECK (0xc0000092) at pc=0x04d30758, pid=2268, tid=3760
where the pc, pid, and tid change each invocation. Since this is not a null pointer
exception, it doesn't seem to be a duplicate of 27454. This is again a crash during
NodeList.getLength() after doing a node.getElementsByTagName() call.
If you need more data from the crash, please let me know what to do,
especially if the Java 5.0 serviceability improvements can help.
---------------------------------------------------------------------
View this comment:
http://issues.apache.org/jira/browse/XERCESJ-912?page=comments#action_54317
---------------------------------------------------------------------
View the issue:
http://issues.apache.org/jira/browse/XERCESJ-912
Here is an overview of the issue:
---------------------------------------------------------------------
Key: XERCESJ-912
Summary: DOM call via JNI Invocation can crash calling program
Type: Bug
Status: Resolved
Resolution: DUPLICATE
Project: Xerces2-J
Components:
DOM
Versions:
2.6.2
Assignee: Xerces-J Developers Mailing List
Reporter: Michael Good
Created: Wed, 10 Mar 2004 3:07 AM
Updated: Mon, 18 Oct 2004 3:18 PM
Environment: Operating System: Windows XP
Platform: PC
Description:
We have an XML parsing program that is a plug-in for another program. Our main
plug-in is written in C++, calling Java via the JNI Invocation interface. We
then make Xerces-J calls from the Java portion of our plug-in. The Java
portion then calls back into the C++ portion via JNI.
There are some places where Xerces DOM calls are surrounding the container
program. That is, we do not get an exception or a throwable in our Java try-
catch block. Instead, the exception is sometimes picked up in the C++ try-
catch block, and sometimes crashes the whole application without any exception
handler anywhere.
The two calls where we have seen this are in NodeList.getLength() - possibly
related to bug 27454? - and DomImplementation.createDocumentType. When we
upgraded from 2.6.1 to 2.6.2, the bug in our test case moved from
createDocumentType to getLength.
It is currently 100% reproducible here with a test file and our specific
source / machine environment. But it does seem timing-dependent. If we pop up
a log of debug dialogs, the bug goes away. It also goes away if we run the C++
portion from a debugger. If we run another part of our plug-in first before
this part, it also goes away. This getLength() call is used all the time - in
this case, following a node.getElementsByTagName() call - and usually works
fine.
I haven't tested on other Windows OSes, but we don't see this behavior on Mac
OS X.
Thanks for any assistance you can offer. With this type of bug it's usually
something in our JNI that causes the problem, but we have narrowed it down to
the Xerces nodeList.getLength() call in this situation.
Michael Good
Recordare LLC
---------------------------------------------------------------------
JIRA INFORMATION:
This message is automatically generated by JIRA.
If you think it was sent incorrectly contact one of the administrators:
http://issues.apache.org/jira/secure/Administrators.jspa
If you want more information on JIRA, or have a bug to report see:
http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
