Mailing-List: contact dev-help@activemq.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@activemq.apache.org
Message-ID: <990831277.1211210169130.JavaMail.jira@brutus>
Date: Mon, 19 May 2008 08:16:09 -0700 (PDT)
From: "Timothy Bish (JIRA)" <jira@apache.org>
To: dev@activemq.apache.org
Subject: [jira] Resolved: (AMQCPP-174) Deadlock on transacted session
In-Reply-To: <71375912.1208223583457.JavaMail.jira@brutus>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit


     [ https://issues.apache.org/activemq/browse/AMQCPP-174?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Timothy Bish resolved AMQCPP-174.
---------------------------------

      Assignee: Timothy Bish  (was: Nathan Mittler)
    Resolution: Fixed

Made some changes to the session executer that seems to have resolved the dedlock issue.  Changes are in SVN rev: 657835

> Deadlock on transacted session
> ------------------------------
>
>                 Key: AMQCPP-174
>                 URL: https://issues.apache.org/activemq/browse/AMQCPP-174
>             Project: ActiveMQ C++ Client
>          Issue Type: Bug
>          Components: CMS Impl
>    Affects Versions: 2.1.3
>         Environment: Windows XP
> Visual Studio 2003 & 2005
> activemq-cpp 2.1.3
> apache-activemq 5.0.0
>            Reporter: Nick Cooper
>            Assignee: Timothy Bish
>             Fix For: 2.2
>
>         Attachments: request_response.zip
>
>
> I am implementing a request/response client and require the ability to do transactions. I initially created one connection with one session with a producer and a consumer. This worked fine when creating the session with Session::AUTO_ACKNOWLEDGE. When I changed this to use Session::SESSION_TRANSACTED
> with session->commit() the application stopped working. To help work out the issue I created separate connections and sessions for the producer and consumer which did not help.
> Looking at the application in the debugger it looks like when session->commit() is called, a dead lock occurs.
> In ActiveMQSessionExecutor::stop()  the following line is executed:
>         // Wakeup the thread so that it can acknowledge the stop request.
>         mutex.notifyAll();
> but eventQ.size() is 0, that is no thread is notified. The next line is then executed:
>         // Wait for the thread to notify us that it has acknowledged
>         // the stop request.
>         mutex.wait();
> but there is no other thread to ever come and wake it up, thus it hangs and the session is never committed.
> Am I using the API incorrectly or is this a real issue?
> Note this looks very similar to issue 155 https://issues.apache.org/activemq/browse/AMQCPP-155
> Thread Summary
> --------------
>  	2000	main	             tool::DefaultCounter::`scalar deleting destructor'	Normal	0
>  	1896	_threadstartex	activemq::concurrent::Mutex::wait	                       Normal	0
>  	3984	_threadstartex	activemq::network::SocketInputStream::read	           Normal	0
> >	1544	_threadstartex	activemq::concurrent::Mutex::notifyAll	                       Normal	0
> Thread 1544 Stacktrace....
> >	ppl_worker_d.exe!activemq::concurrent::Mutex::notifyAll()  Line 233	C++
>  	ppl_worker_d.exe!activemq::core::ActiveMQSessionExecutor::stop()  Line 150 + 0x11	C++
>  	ppl_worker_d.exe!activemq::core::ActiveMQSession::stop()  Line 847 + 0x1b	C++
>  	ppl_worker_d.exe!activemq::core::ActiveMQTransaction::commit()  Line 205	C++
>  	ppl_worker_d.exe!activemq::core::ActiveMQSession::commit()  Line 150 + 0x13	C++
>  	ppl_worker_d.exe!dist::TransactedConnector::onMessage(const cms::Message * message=0x014df8c0)  Line 176 + 0x30	C++
>  	ppl_worker_d.exe!activemq::core::ActiveMQConsumer::dispatch(activemq::core::DispatchData & data={...})  Line 429 + 0x3a	C++
>  	ppl_worker_d.exe!activemq::core::ActiveMQSessionExecutor::dispatch(activemq::core::DispatchData & data={...})  Line 188 + 0x15	C++
>  	ppl_worker_d.exe!activemq::core::ActiveMQSessionExecutor::dispatchAll()  Line 273 + 0x11	C++
>  	ppl_worker_d.exe!activemq::core::ActiveMQSessionExecutor::run()  Line 213	C++
>  	ppl_worker_d.exe!activemq::concurrent::Thread::runCallback(void * param=0x014dc3e8)  Line 152 + 0x13	C++
>  	msvcr71d.dll!_threadstartex(void * ptd=0x014df798)  Line 241 + 0xd	C
>  	kernel32.dll!7c80b683() 	
> Thread 3984 Stacktrace....
> >	ppl_worker_d.exe!activemq::network::SocketInputStream::read(unsigned char * buffer=0x014d8f70, unsigned int bufferSize=1)  Line 152 + 0x19	C++
>  	ppl_worker_d.exe!activemq::io::BufferedInputStream::bufferData()  Line 190 + 0x2c	C++
>  	ppl_worker_d.exe!activemq::io::BufferedInputStream::read(unsigned char * targetBuffer=0x017ff6f4, unsigned int targetBufferSize=4)  Line 125	C++
>  	ppl_worker_d.exe!activemq::io::DataInputStream::readAllData(unsigned char * buffer=0x017ff6f4, unsigned int length=4)  Line 369 + 0x21	C++
>  	ppl_worker_d.exe!activemq::io::DataInputStream::readInt()  Line 168	C++
>  	ppl_worker_d.exe!activemq::connector::openwire::OpenWireFormat::unmarshal(activemq::io::DataInputStream * dis=0x014dcbe8)  Line 218 + 0xd	C++
>  	ppl_worker_d.exe!activemq::connector::openwire::OpenWireCommandReader::readCommand()  Line 71 + 0x1a	C++
>  	ppl_worker_d.exe!activemq::transport::IOTransport::run()  Line 166 + 0x13	C++
>  	ppl_worker_d.exe!activemq::concurrent::Thread::runCallback(void * param=0x014dce40)  Line 152 + 0x13	C++
>  	msvcr71d.dll!_threadstartex(void * ptd=0x014ddd10)  Line 241 + 0xd	C
>  	kernel32.dll!7c80b683() 	
> Thread 1896 Stacktrace....
> >	ppl_worker_d.exe!activemq::concurrent::Mutex::wait(unsigned long millisecs=4294967295)  Line 178 + 0x10	C++
>  	ppl_worker_d.exe!activemq::concurrent::Mutex::wait()  Line 102 + 0xf	C++
>  	ppl_worker_d.exe!activemq::util::Queue<activemq::transport::Command *>::wait()  Line 206 + 0x11	C++
>  	ppl_worker_d.exe!activemq::transport::filters::AsyncSendTransport::run()  Line 118 + 0x11	C++
>  	ppl_worker_d.exe!activemq::concurrent::Thread::runCallback(void * param=0x014dcad0)  Line 152 + 0x13	C++
>  	msvcr71d.dll!_threadstartex(void * ptd=0x014dcd78)  Line 241 + 0xd	C
>  	kernel32.dll!7c80b683() 

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.