C ++ Map迭代和堆栈损坏

时间:2012-05-27 02:03:45

标签: c++ visual-c++ memory map corruption

我正在尝试使用地图系统来存储和更新聊天服务器的数据。应用程序是mutlithreaded并使用锁定系统来防止多个线程访问数据。

问题在于:当从地图中单独删除客户端时,就可以了。但是,当我尝试调用多个关闭时,它会在内存中留下一些。如果我在地图上调用:: clear(),它会导致调试断言错误“Iterator not compatible”或类似错误。该代码将首次运行(使用80多个控制台作为测试连接进行测试),但由于它留下了大块,因此无法再次运行。我尝试过研究方法,并且已经编写了系统来停止代码执行,直到每个进程完成。到目前为止,我感谢任何帮助,并附上了相关的代码片段。

//portion of server code that handles shutting down
DWORD WINAPI runserver(void *params) {  
    runserverPARAMS *p = (runserverPARAMS*)params;  
    /*Server stuff*/                            

    serverquit = 0; 
    //client based cleanup
    vector<int> tokill;
    map<int,int>::iterator it = clientsockets.begin();

    while(it != clientsockets.end()) {      
        tokill.push_back(it->first);
        ++it;
    }
    for(;;) {
        for each (int x in tokill) {
            clientquit[x] = 1;
            while(clientoffline[x] != 1) {
                //haulting execution until thread has terminated
            }
            destoryclient(x);
        }
    }
    //client thread based cleanup complete.
    return 0;
}


//clientioprelim
DWORD WINAPI clientioprelim(void* params) {
    CLIENTthreadparams *inparams = (CLIENTthreadparams *)params;
    /*Socket stuff*/
    for(;;) {       
        /**/
        }
        else {
            if(clientquit[inparams->clientid] == 1)
                break;
        }
    }
    clientoffline[inparams->clientid] = 1;
    return 0;
}

int LOCKED; //exported as extern via libraries.h so it's visible to other source files

void destoryclient(int clientid) {
    for(;;) {
        if(LOCKED == 0) {
            LOCKED = 1;         
            shutdown(clientsockets[clientid], 2);
            closesocket(clientsockets[clientid]);
            if((clientsockets.count(clientid) != 0) && (clientsockets.find(clientid) != clientsockets.end()))
                clientsockets.erase(clientsockets.find(clientid));                  
            if((clientname.count(clientid) != 0) && (clientname.find(clientid) != clientname.end()))
                clientname.erase(clientname.find(clientid));
            if((clientusername.count(clientid) != 0) && (clientusername.find(clientid) != clientusername.end()))
                clientusername.erase(clientusername.find(clientid));
            if((clientaddr.count(clientid) != 0) && (clientaddr.find(clientid) != clientaddr.end()))
                clientaddr.erase(clientusername.find(clientid));
            if((clientcontacts.count(clientid) != 0) && (clientcontacts.find(clientid) != clientcontacts.end())) 
                clientcontacts.erase(clientcontacts.find(clientid));
            if((clientquit.count(clientid) != 0) && (clientquit.find(clientid) != clientquit.end()))
                clientquit.erase(clientquit.find(clientid));    
            if((clientthreads.count(clientid) != 0) && (clientthreads.find(clientid) != clientthreads.end())) 
                clientthreads.erase(clientthreads.find(clientid));
            LOCKED = 0;
            break;          
        }
    }
    return;
}

1 个答案:

答案 0 :(得分:0)

您是否真的使用int进行锁定,还是仅仅是对代码的简化?如果你真的使用int:如果两个线程在分配变量之前检查变量(简化),那么这将无法同时输入两次(或更多)临界区。请参阅mutexes in Wikipedia以供参考。您可以使用由Windows提供的某种互斥锁,也可以使用boost thread而不是int