activemq-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From huntc <hu...@internode.on.net>
Subject Access Violation - activemq::io::SocketInputStream::read - The connection is broken
Date Fri, 29 Jun 2007 23:22:36 GMT

Hi there,

I'm getting some sort of access violation/segmentation fault if I have a
connection open and I shut the broker down. I think that the abend is the
result of some problem when the session is closed. I'm using v.2.0.1.

A note on convention for the following: I'm using boost's shared_ptr class
to manage my heap allocations.

On with the story:

An exception is raised and my handler captures it thus:

void JMSSubscribeIFeatureSourceImp::onException(const CMSException&
inException)
{
    cerr << BUNDLE_IDENTIFIER " - " << inException.getMessage() << endl;

    {
        mJMSConnectedMutex.Lock();
        MutexUnlockSentry theInitedUnlocker(mJMSConnectedMutex);

        mJMSConnected = false;  // Better had try and re-establish the
connection
    }
}

When I next try and send a message, if the connection is signaled as closed
then I attempt to clean up and re-open the connection. Thus:

void JMSSubscribeIFeatureSourceImp::MaintainJMSConnection() throw()
{
    {
        mJMSConnectedMutex.Lock();
        auto_ptr<MutexUnlockSentry> theUnlockerP(new
MutexUnlockSentry(mJMSConnectedMutex));
        
        if (!mJMSConnected)
        {
            try
            {
                // Ensure any connection that was previously established is
now gone
				
				CloseJMSConnection();
                
                {
                    // Create a ConnectionFactory
                    auto_ptr<ActiveMQConnectionFactory>
theConnectionFactoryP(new ActiveMQConnectionFactory());
                    
                    // Create a Connection
                    mConnectionP =
shared_ptr<Connection>(theConnectionFactoryP->createConnection());
                }
                

                // Indicate that we're now connected
                mJMSConnected = true;
                theUnlockerP.reset();   // Free the lock in case anything
subsequently causes a problem
                                        // and our exception listener is
called.
                
                // Start up the connection and listen in on problems
                mConnectionP->start();
                mConnectionP->setExceptionListener(this);

                // Create a Session
                mSessionP =
shared_ptr<Session>(mConnectionP->createSession(Session::AUTO_ACKNOWLEDGE));

                // Create the destination (Topic or Queue)
                mDestinationP =
shared_ptr<Destination>(mSessionP->createTopic(GetTopicName()));

                // Create a MessageConsumer from the Session to the Topic
                mConsumerP =
shared_ptr<MessageConsumer>(mSessionP->createConsumer(mDestinationP.get()));

                mConsumerP->setMessageListener(this);
                
            }
            catch (const CMSException& inException)
            {
                cerr << BUNDLE_IDENTIFIER " - During connection - " <<
inException.getMessage() << endl;
            }
            catch (const bad_alloc&)
            {
                cerr << BUNDLE_IDENTIFIER " - bad alloc when attempting to
get factory" << endl;
            }
        }
    }
}

My close connection routine looks like this (I've been explicit closing
things in the reverse order to which they're opened - I should just be able
to reset the connection shared ptr causing a dealloc and that's all that is
required. However by stepping through the below I find that my problem
occurs during the mSessionP->close()):

void JMSSubscribeIFeatureSourceImp::CloseJMSConnection() throw()
{

	if (mConsumerP.get() != 0)
	{
		try
		{
			mConsumerP->setMessageListener(0);
		}
		catch (const CMSException&)
		{}
		
	}

	if (mSessionP.get() != 0)
	{
		try
		{
			mSessionP->close();
		}
		catch (const CMSException&)
		{}
		
	}
	
	if (mConnectionP.get() != 0)
	{
		try
		{
			mConnectionP->setExceptionListener(0);
		}
		catch (const CMSException&)
		{}

		try
		{
			mConnectionP->stop();
		}
		catch (const CMSException&)
		{}
		
		try
		{
			mConnectionP->close();
		}
		catch (const CMSException&)
		{}
		
	}

    mConsumerP.reset();
    mDestinationP.reset();
	mSessionP.reset();
	mConnectionP.reset();
}

The most interesting point to note is that the problem does not occur on my
Intel Quad Xeon Mac OS X machine; only on my slow old PowerBook G4. I'm thus
suspecting a race condition.

Any thoughts before I investigate this further?

Kind regards,
Christopher
-- 
View this message in context: http://www.nabble.com/Access-Violation---activemq%3A%3Aio%3A%3ASocketInputStream%3A%3Aread---The-connection-is-broken-tf4002670s2354.html#a11368797
Sent from the ActiveMQ - User mailing list archive at Nabble.com.


Mime
View raw message