Re: [boost] FW: [Boost-users] [threads] Using boost::mutex raises access violation exception

From: Neil Groves (neil_at_[hidden])
Date: 2007-11-24 05:41:23

I have experienced a similar issue with boost::mutex when the mutex was
instantiated from DllMain (in a DLL). The Windows implementation uses a
Critical Section. I have found that under many startup / shutdown cases the
InitializeCriticalSection returns null. I cannot find documentation that
states that this happens. The solution is to create the boost::mutex in
another manner. Hence if you stop the boost::mutex from being a global
variable I think the problem will disappear.

I hope this helps,
Neil Groves

-----Original Message-----
From: boost-bounces_at_[hidden] [mailto:boost-bounces_at_[hidden]]
On Behalf Of v2cechura
Sent: 23 November 2007 14:31
To: boost_at_[hidden]
Subject: [boost] FW: [Boost-users] [threads] Using boost::mutex raises
access violation exception

I posted this to the user's list (no answer till now), but later came to the
conclusion that developer's list might be a better place.
Any help on th issue appreciated.

Vaclav

-----Original Post------

I just recently installed and built boost library (1.34.1) using bcbboost. I
wanted to try the boost::thread class. This simple code gives me "Access
Violation" in Borland C++ Builder 2006 when boost::mutex is
used:

//--------------------------------------------------------------------------
-
#include <iostream>
#include <boost/thread.hpp>
//--------------------------------------------------------------------------
-
boost::xtime xt = {2};

boost::mutex console; // line A

void thr() // int n, boost::mutex & display {
        for (int i = 0; i < 10; i++)
        {
                {
                        boost::mutex::scoped_lock l(console); // line B
                        std::cout << "thr" << 1 << ":" << i << std::endl;
                }
                boost::thread::sleep(xt);
        } // end for
}

int main(int argc, char* argv[])
{
        boost::thread_group tg;
        tg.create_thread(&thr);

        char c;
        std::cin >> c ;
        return 0;
}

If line A and line B are commented-out then the error is gone.

The project is a multithreaded console application, both boost.threads and
rtl are static-linked.

Caller stack is:
:7c812a5b kernel32.RaiseException + 0x52 :004129aa ___raiseDebuggerException
+ 0x1A
:00412a84 ; ___raiseDebuggerException
:7c9037bf ntdll.RtlConvertUlongToLargeInteger + 0x7a :7c90378b
ntdll.RtlConvertUlongToLargeInteger + 0x46 :7c90eafa
ntdll.KiUserExceptionDispatcher + 0xe :0041FA2B
boost::mutex::do_unlock(this=NULL,  =:0012FDE8)
:0041E3A2 boost::detail::thread::lock_ops<boost::mutex>::unlock(m=NULL,
state=:0012FDE8) :0041E12D
boost::condition::do_wait<boost::mutex>(this=:0012FEA4, mutex=NULL)
:0041DE2E
boost::condition::wait<boost::detail::thread::scoped_lock<boost::mutex>
>(this=:0012FEA4, lock=:0012FE5C) :0041D42F
__UNNS__thread_01c827c2c2c9169a::thread_param::wait(this=:0012FE9C)
:0041D201 boost::thread::thread(this=:00000FCC, threadfunc=:00000FC8)
:00000001
:0012fe01 

I put mutex.cpp in my directory and BCB2006 locates the error to be here:

void mutex::do_unlock()
{
    if (m_critical_section) //------------- Access violation HERE 
        release_critical_section(m_mutex);
    else
        release_mutex(m_mutex);
}

The access violation is raised when thread::thread constructor is called.

Looking at the caller stack I noticed that mutex parameter is NULL from some
point and it results in this==NULL for the last call. Note that this mutex
parameter is not the mutex "console" 
object I use in the program. It is a mutex that boost.threads library uses
during thread initialization. On the other hand there is some dependance on
mutex "console" because when I comment it out the error is gone.

I maked console a pointer with appropriate changes to source and created it
using new but it did not help.

Am I missing something?

Any help appreciated. Does anybody have experiences with boost::thread and
mutex under BCB(2006,2007...)?

Vaclav

 

_______________________________________________
Unsubscribe & other changes:
http://lists.boost.org/mailman/listinfo.cgi/boost