pthread z klasy

86

Powiedzmy, że mam taką klasę jak

class c { 
    // ...
    void *print(void *){ cout << "Hello"; }
}

A potem mam wektor c

vector<c> classes; pthread_t t1;
classes.push_back(c());
classes.push_back(c());

Teraz chcę utworzyć wątek na c.print();

A oto poniższy problem: pthread_create(&t1, NULL, &c[0].print, NULL);

Błąd Ouput: nie można przekonwertować 'void * (tree_item ::) (void )' na 'void * ( ) (void )' dla argumentu '3' na 'int pthread_create (pthread_t *, const pthread_attr_t *, void * ( ) (void ), nieważne *) '

Anioł Król 47
źródło

Odpowiedzi:

147

Nie możesz tego zrobić tak, jak to napisałeś, ponieważ funkcje składowe klasy C ++ mają thisprzekazany ukryty parametr. Nie pthread_create()ma pojęcia, jakiej wartości thisużyć, więc jeśli spróbujesz obejść kompilator, rzutując metodę na funkcję wskaźnika odpowiedniego typu, pojawi się błąd segmetnacji. Musisz użyć statycznej metody klasy (która nie ma thisparametrów) lub zwykłej zwykłej funkcji, aby załadować klasę:

class C
{
public:
    void *hello(void)
    {
        std::cout << "Hello, world!" << std::endl;
        return 0;
    }

    static void *hello_helper(void *context)
    {
        return ((C *)context)->hello();
    }
};
...
C c;
pthread_t t;
pthread_create(&t, NULL, &C::hello_helper, &c);
Adam Rosenfield
źródło
czy powyższe działałoby z wektorami w następujący sposób: pthread_create (& t, NULL, & C :: hello_helper, & vector_c [0]); ?
Angel.King. 47
Wszystkie powyższe komentarze są przydatne. Użyłem kombinacji wszystkich, aby rozwiązać problem .. Nadal nie działa tak prosto, jak próbowałem to zrobić ... Ale niestety mogę oznaczyć tylko jeden jako poprawny, W przeciwnym razie każdy otrzyma kredyt ... Dzięki
Angel.King. 47
Chciałem głosować za tą odpowiedzią, ale używa ona rzutów w stylu C, które muszą zostać zakończone ze skrajnym uprzedzeniem. Poza tym ta odpowiedź jest poprawna.
Chris Jester-Young
@Chris: Nie chcę wdawać się w świętą wojnę o style obsady, ale jest całkowicie poprawne semantycznie, aby użyć w tym przypadku obsady w stylu C.
Adam Rosenfield
2
@AdamRosenfield jest również doskonale semantycznie poprawne, aby połączyć przysłówki w łańcuchy, ale to nie sprawia, że ​​jest to dobry styl! xD
ACK_stoverflow
82

Moim ulubionym sposobem obsługi wątku jest hermetyzowanie go wewnątrz obiektu C ++. Oto przykład:

class MyThreadClass
{
public:
   MyThreadClass() {/* empty */}
   virtual ~MyThreadClass() {/* empty */}

   /** Returns true if the thread was successfully started, false if there was an error starting the thread */
   bool StartInternalThread()
   {
      return (pthread_create(&_thread, NULL, InternalThreadEntryFunc, this) == 0);
   }

   /** Will not return until the internal thread has exited. */
   void WaitForInternalThreadToExit()
   {
      (void) pthread_join(_thread, NULL);
   }

protected:
   /** Implement this method in your subclass with the code you want your thread to run. */
   virtual void InternalThreadEntry() = 0;

private:
   static void * InternalThreadEntryFunc(void * This) {((MyThreadClass *)This)->InternalThreadEntry(); return NULL;}

   pthread_t _thread;
};

To use it, you would just create a subclass of MyThreadClass with the InternalThreadEntry() method implemented to contain your thread's event loop. You'd need to call WaitForInternalThreadToExit() on the thread object before deleting the thread object, of course (and have some mechanism to make sure the thread actually exits, otherwise WaitForInternalThreadToExit() would never return)

Jeremy Friesner
źródło
1
That is a great way i can understand the use of the above Virtual Class, But i have much deaper problems..I have threads that spawn off other threads that need to be all put in a vector. And then a recursive loop to go and join all the threads. Im sure i could implement the above to do that as well by calling the wait in the proper place, But il try it to see where i get to
Angel.King.47
4
This solution is so very elegant. I will use it from now on. Thank you Jeremy Friesner. +1
Armada
hello Jeremy Friesner, how to pass a reference to InternalThreadEntry(aclass_ref& refobj) ? what changes I should make?
sree
@sree Add the reference (or a pointer) to the MyThreadClass as a member variable; then InternalThreadEntry() can access it directly, without having to worry about passing it via the (void *) argument.
Jeremy Friesner
10

You'll have to give pthread_create a function that matches the signature it's looking for. What you're passing won't work.

You can implement whatever static function you like to do this, and it can reference an instance of c and execute what you want in the thread. pthread_create is designed to take not only a function pointer, but a pointer to "context". In this case you just pass it a pointer to an instance of c.

For instance:

static void* execute_print(void* ctx) {
    c* cptr = (c*)ctx;
    cptr->print();
    return NULL;
}


void func() {

    ...

    pthread_create(&t1, NULL, execute_print, &c[0]);

    ...
}
Jared Oberhaus
źródło
1
ooo i see what you mean.. pass it the pointer of c, gotcha.. will implement and try it out
Angel.King.47
2

The above answers are good, but in my case, 1st approach that converts the function to be a static didn't work. I was trying to convert exiting code to move into thread function but that code had lots to references to non-static class members already. The second solution of encapsulating into C++ object works, but has 3-level wrappers to run a thread.

I had an alternate solution that uses existing C++ construct - 'friend' function, and it worked perfect for my case. An example of how I used 'friend' (will use the above same example for names showing how it can be converted into a compact form using friend)

    class MyThreadClass
    {
    public:
       MyThreadClass() {/* empty */}
       virtual ~MyThreadClass() {/* empty */}

       bool Init()
       {
          return (pthread_create(&_thread, NULL, &ThreadEntryFunc, this) == 0);
       }

       /** Will not return until the internal thread has exited. */
       void WaitForThreadToExit()
       {
          (void) pthread_join(_thread, NULL);
       }

    private:
       //our friend function that runs the thread task
       friend void* ThreadEntryFunc(void *);

       pthread_t _thread;
    };

    //friend is defined outside of class and without any qualifiers
    void* ThreadEntryFunc(void *obj_param) {
    MyThreadClass *thr  = ((MyThreadClass *)obj_param); 

    //access all the members using thr->

    return NULL;
    }

Ofcourse, we can use boost::thread and avoid all these, but I was trying to modify the C++ code to not use boost (the code was linking against boost just for this purpose)

sanmara
źródło
1

My first answer ever in the hope that it'll be usefull to someone : I now this is an old question but I encountered exactly the same error as the above question as I'm writing a TcpServer class and I was trying to use pthreads. I found this question and I understand now why it was happening. I ended up doing this:

#include <thread>

method to run threaded -> void* TcpServer::sockethandler(void* lp) {/*code here*/}

and I call it with a lambda -> std::thread( [=] { sockethandler((void*)csock); } ).detach();

that seems a clean approach to me.

ZoOl007
źródło
0

Too many times I've found ways to solve what you are asking for that, in my opinion are too complicated. For instance you have to define new class types, link library etc. So I decided to write a few lines of code that allow the end user to basically be able to "thread-ize" a "void ::method(void)" of whatever class. For sure this solution I implemented can be extended, improved etc, so, if you need more specific methods or features, add them and please be so kind to keep me in the loop.

Here are 3 files that show what I did.

    // A basic mutex class, I called this file Mutex.h
#ifndef MUTEXCONDITION_H_
#define MUTEXCONDITION_H_

#include <pthread.h>
#include <stdio.h>

class MutexCondition
{
private:
    bool init() {
        //printf("MutexCondition::init called\n");
        pthread_mutex_init(&m_mut, NULL);
        pthread_cond_init(&m_con, NULL);
        return true;
    }

    bool destroy() {
        pthread_mutex_destroy(&m_mut);
        pthread_cond_destroy(&m_con);
        return true;
    }

public:
    pthread_mutex_t m_mut;
    pthread_cond_t m_con;

    MutexCondition() {
        init();
    }
    virtual ~MutexCondition() {
        destroy();
    }

    bool lock() {
        pthread_mutex_lock(&m_mut);
        return true;
    }

    bool unlock() {
        pthread_mutex_unlock(&m_mut);
        return true;
    }

    bool wait() {
        lock();
        pthread_cond_wait(&m_con, &m_mut);
        unlock();
        return true;
    }

    bool signal() {
        pthread_cond_signal(&m_con);
        return true;
    }
};
#endif
// End of Mutex.h

// The class that incapsulates all the work to thread-ize a method (test.h):

#ifndef __THREAD_HANDLER___
#define __THREAD_HANDLER___

#include <pthread.h>
#include <vector>
#include <iostream>
#include "Mutex.h"

using namespace std;

template <class T> 
class CThreadInfo
{
  public:
    typedef void (T::*MHT_PTR) (void);
    vector<MHT_PTR> _threaded_methods;
    vector<bool> _status_flags;
    T *_data;
    MutexCondition _mutex;
    int _idx;
    bool _status;

    CThreadInfo(T* p1):_data(p1), _idx(0) {}
    void setThreadedMethods(vector<MHT_PTR> & pThreadedMethods)
    {
        _threaded_methods = pThreadedMethods;
      _status_flags.resize(_threaded_methods.size(), false);
    }
};

template <class T> 
class CSThread {
  protected:
    typedef void (T::*MHT_PTR) (void);
    vector<MHT_PTR> _threaded_methods;
    vector<string> _thread_labels;
    MHT_PTR _stop_f_pt;
    vector<T*> _elements;
    vector<T*> _performDelete;
    vector<CThreadInfo<T>*> _threadlds;
    vector<pthread_t*> _threads;
    int _totalRunningThreads;

    static void * gencker_(void * pArg)
    {
      CThreadInfo<T>* vArg = (CThreadInfo<T> *) pArg;
      vArg->_mutex.lock();
      int vIndex = vArg->_idx++;
      vArg->_mutex.unlock();

      vArg->_status_flags[vIndex]=true;

      MHT_PTR mhtCalledOne = vArg->_threaded_methods[vIndex];
      (vArg->_data->*mhtCalledOne)();
      vArg->_status_flags[vIndex]=false;
        return NULL;
    }

  public:
    CSThread ():_stop_f_pt(NULL), _totalRunningThreads(0)  {}
    ~CSThread()
    {
      for (int i=_threads.size() -1; i >= 0; --i)
          pthread_detach(*_threads[i]);

      for (int i=_threadlds.size() -1; i >= 0; --i)
        delete _threadlds[i];

      for (int i=_elements.size() -1; i >= 0; --i)
         if (find (_performDelete.begin(), _performDelete.end(), _elements[i]) != _performDelete.end())
              delete _elements[i];
    }
    int  runningThreadsCount(void) {return _totalRunningThreads;}
    int  elementsCount()        {return _elements.size();}
    void addThread (MHT_PTR p, string pLabel="") { _threaded_methods.push_back(p); _thread_labels.push_back(pLabel);}
    void clearThreadedMethods() { _threaded_methods.clear(); }
    void getThreadedMethodsCount() { return _threaded_methods.size(); }
    void addStopMethod(MHT_PTR p)  { _stop_f_pt  = p; }
    string getStatusStr(unsigned int _elementIndex, unsigned int pMethodIndex)
    {
      char ch[99];

      if (getStatus(_elementIndex, pMethodIndex) == true)
        sprintf (ch, "[%s] - TRUE\n", _thread_labels[pMethodIndex].c_str());
      else 
        sprintf (ch, "[%s] - FALSE\n", _thread_labels[pMethodIndex].c_str());

      return ch;
    }
    bool getStatus(unsigned int _elementIndex, unsigned int pMethodIndex)
    {
      if (_elementIndex > _elements.size()) return false;
      return _threadlds[_elementIndex]->_status_flags[pMethodIndex];
    }

    bool run(unsigned int pIdx) 
    {
      T * myElem = _elements[pIdx];
      _threadlds.push_back(new CThreadInfo<T>(myElem));
      _threadlds[_threadlds.size()-1]->setThreadedMethods(_threaded_methods);

      int vStart = _threads.size();
      for (int hhh=0; hhh<_threaded_methods.size(); ++hhh)
          _threads.push_back(new pthread_t);

      for (int currentCount =0; currentCount < _threaded_methods.size(); ++vStart, ++currentCount)
      {
                if (pthread_create(_threads[vStart], NULL, gencker_, (void*) _threadlds[_threadlds.size()-1]) != 0)
        {
                // cout <<"\t\tThread " << currentCount << " creation FAILED for element: " << pIdx << endl;
                    return false;
                }
        else
        {
            ++_totalRunningThreads;
             // cout <<"\t\tThread " << currentCount << " creation SUCCEDED for element: " << pIdx << endl;
                }
      }
      return true;
    }

    bool run() 
    {
            for (int vI = 0; vI < _elements.size(); ++vI) 
            if (run(vI) == false) return false;
          // cout <<"Number of currently running threads: " << _totalRunningThreads << endl;
        return true;
    }

    T * addElement(void)
    {
      int vId=-1;
      return addElement(vId);
    }

    T * addElement(int & pIdx)
    {
      T * myElem = new T();
      _elements.push_back(myElem);
      pIdx = _elements.size()-1;
      _performDelete.push_back(myElem);
      return _elements[pIdx];
    }

    T * addElement(T *pElem)
    {
      int vId=-1;
      return addElement(pElem, vId);
    }

    T * addElement(T *pElem, int & pIdx)
    {
      _elements.push_back(pElem);
      pIdx = _elements.size()-1;
      return pElem;
    }

    T * getElement(int pId) { return _elements[pId]; }

    void stopThread(int i)  
    {
      if (_stop_f_pt != NULL) 
      {
         ( _elements[i]->*_stop_f_pt)() ;
      }
      pthread_detach(*_threads[i]);
      --_totalRunningThreads;
    }

    void stopAll()  
    {
      if (_stop_f_pt != NULL) 
        for (int i=0; i<_elements.size(); ++i) 
        {
          ( _elements[i]->*_stop_f_pt)() ;
        }
      _totalRunningThreads=0;
    }
};
#endif
// end of test.h

// A usage example file "test.cc" that on linux I've compiled with The class that incapsulates all the work to thread-ize a method: g++ -o mytest.exe test.cc -I. -lpthread -lstdc++

#include <test.h>
#include <vector>
#include <iostream>
#include <Mutex.h>

using namespace std;

// Just a class for which I need to "thread-ize" a some methods
// Given that with OOP the objecs include both "functions" (methods)
// and data (attributes), then there is no need to use function arguments,
// just a "void xxx (void)" method.
// 
class TPuck
{
  public:
   bool _go;
   TPuck(int pVal):_go(true)
   {
     Value = pVal;
   }
   TPuck():_go(true)
   {
   }
   int Value;
   int vc;

   void setValue(int p){Value = p; }

   void super()
   {
     while (_go)
     {
      cout <<"super " << vc << endl;
            sleep(2);
         }
      cout <<"end of super " << vc << endl;
   }

   void vusss()
   {
     while (_go)
     {
      cout <<"vusss " << vc << endl;
      sleep(2);
     }
      cout <<"end of vusss " << vc << endl;
   }

   void fazz()
   {
     static int vcount =0;
     vc = vcount++;
     cout <<"Puck create instance: " << vc << endl;
     while (_go)
     {
       cout <<"fazz " << vc << endl;
       sleep(2);
     }
     cout <<"Completed TPuck..fazz instance "<<  vc << endl;
   }

   void stop()
   {
      _go=false;
      cout << endl << "Stopping TPuck...." << vc << endl;
   }
};


int main(int argc, char* argv[])
{
  // just a number of instances of the class I need to make threads
  int vN = 3;

  // This object will be your threads maker.
  // Just declare an instance for each class
  // you need to create method threads
  //
  CSThread<TPuck> PuckThreadMaker;
  //
  // Hera I'm telling which methods should be threaded
  PuckThreadMaker.addThread(&TPuck::fazz, "fazz1");
  PuckThreadMaker.addThread(&TPuck::fazz, "fazz2");
  PuckThreadMaker.addThread(&TPuck::fazz, "fazz3");
  PuckThreadMaker.addThread(&TPuck::vusss, "vusss");
  PuckThreadMaker.addThread(&TPuck::super, "super");

  PuckThreadMaker.addStopMethod(&TPuck::stop);

  for (int ii=0; ii<vN; ++ii)
  {
    // Creating instances of the class that I need to run threads.
    // If you already have your instances, then just pass them as a
    // parameter such "mythreadmaker.addElement(&myinstance);"
    TPuck * vOne = PuckThreadMaker.addElement();
  }

  if (PuckThreadMaker.run() == true)
  {
    cout <<"All running!" << endl;
  }
  else
  {
    cout <<"Error: not all threads running!" << endl;
  }

  sleep(1);
  cout <<"Totale threads creati: " << PuckThreadMaker.runningThreadsCount()  << endl;
  for (unsigned int ii=0; ii<vN; ++ii)
  {
    unsigned int kk=0;
    cout <<"status for element " << ii << " is " << PuckThreadMaker.getStatusStr(ii, kk++) << endl; 
    cout <<"status for element " << ii << " is " << PuckThreadMaker.getStatusStr(ii, kk++) << endl; 
    cout <<"status for element " << ii << " is " << PuckThreadMaker.getStatusStr(ii, kk++) << endl; 
    cout <<"status for element " << ii << " is " << PuckThreadMaker.getStatusStr(ii, kk++) << endl; 
    cout <<"status for element " << ii << " is " << PuckThreadMaker.getStatusStr(ii, kk++) << endl; 
  }

  sleep(2);
  PuckThreadMaker.stopAll();
  cout <<"\n\nAfter the stop!!!!" << endl;
  sleep(2);

  for (int ii=0; ii<vN; ++ii)
  {
    int kk=0;
    cout <<"status for element " << ii << " is " << PuckThreadMaker.getStatusStr(ii, kk++) << endl; 
    cout <<"status for element " << ii << " is " << PuckThreadMaker.getStatusStr(ii, kk++) << endl; 
    cout <<"status for element " << ii << " is " << PuckThreadMaker.getStatusStr(ii, kk++) << endl; 
    cout <<"status for element " << ii << " is " << PuckThreadMaker.getStatusStr(ii, kk++) << endl; 
    cout <<"status for element " << ii << " is " << PuckThreadMaker.getStatusStr(ii, kk++) << endl; 
  }

  sleep(5);
  return 0;
}

// End of test.cc
Dodo
źródło
0

This is a bit old question but a very common issue which many face. Following is a simple and elegant way to handle this by using std::thread

#include <iostream>
#include <utility>
#include <thread>
#include <chrono>

class foo
{
    public:
        void bar(int j)
        {
            n = j;
            for (int i = 0; i < 5; ++i) {
                std::cout << "Child thread executing\n";
                ++n;
                std::this_thread::sleep_for(std::chrono::milliseconds(10));
            }
        }
        int n = 0;
};

int main()
{
    int n = 5;
    foo f;
    std::thread class_thread(&foo::bar, &f, n); // t5 runs foo::bar() on object f
    std::this_thread::sleep_for(std::chrono::milliseconds(20));
    std::cout << "Main Thread running as usual";
    class_thread.join();
    std::cout << "Final value of foo::n is " << f.n << '\n';
}

Above code also takes care of passing argument to the thread function.

Refer std::thread document for more details.

pankaj
źródło
-1

My guess would be this is b/c its getting mangled up a bit by C++ b/c your sending it a C++ pointer, not a C function pointer. There is a difference apparently. Try doing a

(void)(*p)(void) = ((void) *(void)) &c[0].print; //(check my syntax on that cast)

and then sending p.

I've done what your doing with a member function also, but i did it in the class that was using it, and with a static function - which i think made the difference.

EdH
źródło
I tried the above but its giving me syntax errors.. Tried to change it around as well... If you would be kind enough to show that using the pthread_create(...) it might be helpfull
Angel.King.47