c++11 future 与 thread_local 的问题

#pragma once

#include <iostream>
#include <thread>
#include <vector>
#include <memory>
#include <future>

using std::cout;
using std::endl;

class Test_threadLocal
{
    typedef Test_threadLocal self_type;

    class ThreadLocalData
    {
    public:
        ThreadLocalData()
        {
            cout << __FUNCTION__ << " threadId:" << std::this_thread::get_id() << endl;
        }

        ~ThreadLocalData()
        {
            cout << __FUNCTION__ << " threadId:" << std::this_thread::get_id() << endl;
        }
    };

private:
    bool entry()
    {
         thread_local ThreadLocalData data;
         return true;
    }

    void runWithThread()
    {     
        cout << __FUNCTION__ << " threadId:" << std::this_thread::get_id() << " begin ******************************" << endl;
        std::vector<std::shared_ptr<std::thread>> threadGroup;
        const int threadCount = 5;
        for (int i = 0; i < threadCount; ++i)
        {
            std::shared_ptr<std::thread> t(new std::thread(&self_type::entry,this));
            threadGroup.push_back(t);
        }

        for (int i = 0; i < threadCount; ++i)
        {
            threadGroup[i]->join();
        }

        cout << __FUNCTION__ << " threadId:" << std::this_thread::get_id() << " end ******************************" << endl;
        int a = 5;
        ++a;
    }

    void runWithFuture()
    {
        cout << __FUNCTION__ << " threadId:" << std::this_thread::get_id() << " begin ******************************" << endl;
        std::vector< std::future<bool> > futureList;
        const int threadCount = 5;
        for (int i = 0; i < threadCount; ++i)
        {
            futureList.push_back(std::async(std::launch::async, &self_type::entry, this));
        }

        bool bSuccess = true;
        for (auto& f : futureList)
        {
            bool rtn = f.get();
            cout << "rtn:" << rtn << endl;
        }

        cout << __FUNCTION__ << " threadId:" << std::this_thread::get_id() << " end ******************************" << endl;
        int a = 5;
        ++a;
    }

public:
    static void test()
    {
        self_type obj;
        obj.runWithThread();
        obj.runWithFuture();
    }
};

发现： runWithThread执行完毕后，正常调用 ThreadLocalData 析构函数

            runWithFuture执行完毕后，并未调用 ThreadLocalData 析构函数 (这是个问题)