多线程 .NET 队列问题

Multithreaded .NET queue problems(多线程 .NET 队列问题)

问题描述

我的代码中有一个奇怪的错误.这是极其罕见的(可能每隔几周发生一次)，但它就在那里，我不知道为什么.

I'm have a wierd error in my code. It's extremely rare (happens once every few weeks maybe), but it's there and I'm not sure why.

我们有 2 个线程在运行，1 个线程获取网络消息并将它们添加到队列中，如下所示:

We have 2 threads running, 1 thread gets networked messages and adds them to a Queue like this:

DataMessages.Enqueue(new DataMessage(client, msg));

另一个线程从这个队列中取出消息并处理它们，如下所示:

Another thread takes messages off this queue and handles them, like this:

while (NetworkingClient.DataMessages.Count > 0)
{
    DataMessage message = NetworkingClient.DataMessages.Dequeue();

    switch (message.messageType)
    {
       ...
    }
}

但是，我经常在 switch (message.messageType) 行上收到 NullReferenceException，我可以在调试器中看到该消息为空.

However once every so often I get a NullReferenceException on the line switch (message.messageType) and I can see in the debugger that message is null.

不可能将空值放入队列(参见代码的第一部分)，这是唯一使用队列的两件事.

It is not possible that a null value was put onto the queue (see the first bit of code), and these are the only 2 things that use the queue.

队列不是线程安全的吗，是不是我在另一个线程入队的确切时刻出队，这导致了故障?

Is Queue not thread-safe, could it be that I'm dequeuing at the exact moment that the other thread is enqueuing and this causes the glitch?

推荐答案

    while (NetworkingClient.DataMessages.Count > 0)
    {
        // once every two weeks a context switch happens to be here.
        DataMessage message = NetworkingClient.DataMessages.Dequeue();

        switch (message.messageType)
        {
           ...
        }
    }

...当您在该位置获得上下文切换时，第一个表达式的结果(NetworkingClient.DataMessages.Count > 0) 对两个线程都为真，Dequeue() 操作首先获取对象，第二个线程获取对象一个 null(而不是 InvalidOperationException，因为队列的内部状态没有完全更新以抛出正确的异常).

... and when you get that context switch in that location, the result of the first expression (NetworkingClient.DataMessages.Count > 0) is true for both threads, and the one that get's to the Dequeue() operation first get's the object and the second thread get's a null (instead of InvalidOperationException because the Queue's internal state wasn't fully updated to throw the right exception).

现在你有两个选择:

1. 使用 .NET 4.0 ConcurrentQueue

重构你的代码:

让它看起来像这样:

while(true)
{
  DataMessage message = null;

  lock(NetworkingClient.DataMessages.SyncRoot) {
       if(NetworkingClient.DataMessages.Count > 0) {
          message = NetworkingClient.DataMessages.Dequeue();
       } else {
         break;
       }
    }
    // .. rest of your code
}

更新以反映 Heandel 的评论.

updated to reflect Heandel's comment.

这篇关于多线程 .NET 队列问题的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持编程学习网！