[ https://issues.apache.org/activemq/browse/AMQCPP-128?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_39435
]
Albert Strasheim commented on AMQCPP-128:
-----------------------------------------
I look at a bit at the "Expression: list erase iterator outside range" crash.
In the case I looked at, one thread was receiving from a socket. Another thread was waiting
on a mutex and a third thread was doing notifyAll on that same mutex.
The thread waiting on the mutex was doing so from ActiveMQSessionExecutor::run(). Stack trace:
> vs2005-activemq-example.exe!std::list<void *,std::allocator<void *> >::_Insert(std::list<void
*,std::allocator<void *> >::_Iterator<1> _Where=..., void * const & _Val=0x00f2fe30)
Line 706 + 0xf bytes C++
kernel32.dll!7c802532()
vs2005-activemq-example.exe!activemq::concurrent::Mutex::wait(unsigned long millisecs=4294967295)
Line 199 C++
vs2005-activemq-example.exe!activemq::concurrent::Mutex::wait() Line 106 C++
vs2005-activemq-example.exe!activemq::core::ActiveMQSessionExecutor::run() Line 231 C++
vs2005-activemq-example.exe!activemq::concurrent::Thread::runCallback(void * param=0x00d31d10)
Line 152 + 0x13 bytes C++
msvcr80d.dll!_callthreadstartex() Line 348 + 0xf bytes C
msvcr80d.dll!_threadstartex(void * ptd=0x00d31a90) Line 331 C
The thread doing notifyAll was doing so from ActiveMQSessionExecutor::stop(), called as part
of a rollback. Stack trace (the _Debug_message call at the top of the stack is the MSVC checked
iterators kicking in to report the error):
> msvcp80d.dll!std::_Debug_message(const wchar_t * message=0x006a0fb0, const wchar_t *
file=0x006a0d88, unsigned int line=776) Line 24 C++
vs2005-activemq-example.exe!std::list<void *,std::allocator<void *> >::erase(std::list<void
*,std::allocator<void *> >::_Iterator<1> _Where=0xcdcdcdcd) Line 776 + 0x15
bytes C++
vs2005-activemq-example.exe!std::list<void *,std::allocator<void *> >::pop_front()
Line 649 + 0x20 bytes C++
vs2005-activemq-example.exe!activemq::concurrent::Mutex::notifyAll() Line 258 C++
vs2005-activemq-example.exe!activemq::core::ActiveMQSessionExecutor::stop() Line 154 C++
vs2005-activemq-example.exe!activemq::core::ActiveMQSession::stop() Line 849 C++
vs2005-activemq-example.exe!activemq::core::ActiveMQTransaction::rollback() Line 239 C++
vs2005-activemq-example.exe!activemq::core::ActiveMQSession::rollback() Line 169 + 0x19
bytes C++
vs2005-activemq-example.exe!test_transaction(const std::basic_string<char,std::char_traits<char>,std::allocator<char>
> & url="tcp://localhost:61613?wireFormat=stomp") Line 56 + 0x15 bytes C++
vs2005-activemq-example.exe!main(int argc=1, char * * argv=0x00355898) Line 85 + 0x30 bytes
C++
vs2005-activemq-example.exe!__tmainCRTStartup() Line 597 + 0x19 bytes C
vs2005-activemq-example.exe!mainCRTStartup() Line 414 C
The strange part: it looks like both threads synchronize on this mutex before they call wait
or notifyAll, so why are they both inside the synchronized block at the same time?
> Transacted session crashes or loses messages
> --------------------------------------------
>
> Key: AMQCPP-128
> URL: https://issues.apache.org/activemq/browse/AMQCPP-128
> Project: ActiveMQ C++ Client
> Issue Type: Bug
> Affects Versions: 2.1
> Reporter: Albert Strasheim
> Assignee: Timothy Bish
> Fix For: 2.1
>
> Attachments: main.cpp
>
>
> Attached is some code that does a few things with transacted sessions, in a loop. This
sometimes reproduces the crash I mentioned in one of my latest comments in AMQCPP-122.
> I'm testing on Windows XP SP2 with Visual Studio 8 on Core 2 Duo machine.
> I've also seen all of the following happen in subsequent runs of the code:
> - Where I expect to receive a message, nothing turns up (even with a long timeout period).
Cleanup goes fine.
> - Where I expect to receive a message, nothing turns up. Cleanup hangs.
> - Sometimes it prints:
> Mutex::unlock - Failed, not Lock Owner!
> FILE: ..\src\main\activemq\concurrent\Mutex.cpp, LINE: 83
> FILE: activemq-cpp\src\main\activemq/concurrent/Lock.h, LINE: 110
> FILE: ..\src\main\activemq\core\ActiveMQConsumer.cpp, LINE: 443
> FILE: ..\src\main\activemq\core\ActiveMQSessionExecutor.cpp, LINE: 192
> and keeps on going
> - Sometimes it crashes with the following when running a Debug build on Windows:
> ---------------------------
> Microsoft Visual C++ Debug Library
> ---------------------------
> Debug Assertion Failed!
> Program: ...
> File: C:\Program Files\Microsoft Visual Studio 8\VC\include\list
> Line: 776
> Expression: list erase iterator outside range
> For information on how your program can cause an assertion
> failure, see the Visual C++ documentation on asserts.
> (Press Retry to debug the application)
> ---------------------------
> Abort Retry Ignore
> ---------------------------
> - Sometimes it crashes with something along these lines:
> ---------------------------
> vs2005-activemq-example.exe - Application Error
> ---------------------------
> The instruction at "0x0053dd80" referenced memory at "0xdddddddd". The memory could not
be "read".
> Click on OK to terminate the program
> Click on CANCEL to debug the program
> ---------------------------
> OK Cancel
> ---------------------------
> The relevant thread's stack trace looks like this:
> > vs2005-activemq-example.exe!activemq::core::ActiveMQSession::close() Line 110 +
0x8 bytes C++
> vs2005-activemq-example.exe!activemq::core::ActiveMQSession::~ActiveMQSession() Line
74 + 0x8 bytes C++
> vs2005-activemq-example.exe!activemq::core::ActiveMQSession::`scalar deleting destructor'()
+ 0xf bytes C++
> vs2005-activemq-example.exe!test_transaction(const std::basic_string<char,std::char_traits<char>,std::allocator<char>
> & url="tcp://localhost:61613?wireFormat=stomp") Line 78 + 0x37 bytes C++
> vs2005-activemq-example.exe!main(int argc=1, char * * argv=0x003558c8) Line 85 + 0x30
bytes C++
> vs2005-activemq-example.exe!__tmainCRTStartup() Line 597 + 0x19 bytes C
> vs2005-activemq-example.exe!mainCRTStartup() Line 414 C
> It crashes where it's doing resource->close() because resource points to garbage.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
|