C++ 条件等待停止执行

Question

我有以下类方法:

myclass::concurFnc(bool changingVar )
{
    int i;
    mIdMutex.lock();
    i = mId++;
    mIdMutex.unlock;

    std::cout << "Try Waiting: " << i << std::endl;
    std::mutex waitValidMutex;
    std::unique_lock<std::mutex> l(waitValidMutex);
    mNoOtherThreadDiffCompareVal.wait(l, [this, &changingVar]()
    {
        mSemMutex.lock();
        bool result = false;
        if(myclass::staticVar == changingVar)
        {
            std::cout << "Same var : " << changingVar << i << std::endl;
            result = true;
            --mSem;
        } else if(mSem == 1)
        {
            std::cout << "Only one in crit section: " << i << std::endl;
            --mSem;
            myclass::staticVar = changingVar;
            result = true;
        } else
        {
            std::cout << "wait: " << i << std::endl;
            result = false;
        }
        std::cout << "Sem is now " << mSem << std::endl;
        mSemMutex.unlock();
        return result;
    });

    //DO STUFF

    mSemMutex.lock();
    ++mSem;
    std::cout << "In the end sem is: " << mSem << std::endl;
    mSemMutex.unlock();
    mNoOtherThreadDiffCompareVal.notify_all();
}

//class Member Variables
std::mutex mSemMutex;
std::mutex mIdMutex;
std::condition_variable mNoOtherThreadDiffCompareVal;
int32_t mSem;
int32_t mId;

myclass::myclass() : mSem(1), mId(0)
{
}

void myclass::startThreads()
{
    const int AMOUNT_OF_PROCESSES = 3
    std::vector<std::shared_ptr<boost::thread> > processes;
    for (size_t t_id = 0; t_id < AMOUNT_OF_PROCESSES; ++t_id) 
    {
        bool changingVar = getVarSomewhere();
        std::shared_ptr<boost::thread> thread_ptr(new boost::thread(&myclass::concurFnc, this, changingVar));
        processes.push_back(thread_ptr);
    }

    for (size_t t_id = 0; t_id < AMOUNT_OF_PROCESSES; ++t_id) {
        processes[t_id]->join();
    }
}

如果 changingVar 与当前在//DO STUFF 中的所有其他调用者相同(或者如果它是唯一的)，它应该允许进入//DO STUFF 部分该函数是从 boost::threads 中的函数并发调用的。

有时会发生执行停止的情况，因为它一直在等待，尽管 mSem 为 1 并且本应调用 notify_all()

执行停止前的输出是:

wait: 3248
Sem is now 0
Try Waiting: 3249
wait: 3249
Sem is now 0
In the end sem is: 1
Only one in crit section: 3249
Sem is now 0
Try Waiting: 3250
wait: 3250
Sem is now 0
In the end sem is: 1
Only one in crit section: 3250
Sem is now 0
In the end sem is: 1
Try Waiting: 3251
Same comp val : 0 of: 3251
Sem is now 0
.In the end sem is: 1

到目前为止我还不能真正重现它。

程序又开始工作了。也许是因为我将 gdb 附加到它，或者不知何故有一个巨大的延迟。

Answer 1

您对条件变量的使用模式很奇怪。

传递给条件变量的wait函数的互斥量应该总是保护你正在等待的条件的互斥量，这里好像是mSemMutex .

这是有道理的，因为为了检查等待的条件是否满足，您无论如何都需要锁定这个互斥体。考虑到这一点，您可能可以完全摆脱 waitValidMutex 并用 mSemMutex 替换它的所有用途。

另请注意，目前，由于 waitValidMutex 是本地互斥锁，您实际上是 risk undefined behavior when calling wait :

Calling this function if lock.mutex() is not the same mutex as the one used by all other threads that are currently waiting on the same condition variable is undefined behavior.