分段故障核心在pthread_join之前转储

Question

My wish

I build a funny code which must run as far on linux:

---

$ ./lib

Main: Creating threads

Main: Waiting for threads to finish

Hello #0 from Thread 1

Hello #0 from Thread 2

Hello #1 from Thread 1

Hello #1 from Thread 2

Hello #2 from Thread 1

Hello #2 from Thread 2

Hello #3 from Thread 1

Hello #3 from Thread 2

Hello #4 from Thread 1

Hello #4 from Thread 2

Hello #5 from Thread 1

Hello #5 from Thread 2

Hello #6 from Thread 1

Hello #6 from Thread 2

Hello #7 from Thread 1

Hello #7 from Thread 2

Hello #8 from Thread 1

Hello #8 from Thread 2

Hello #9 from Thread 1

Hello #9 from Thread 2

Thread 1 terminates

Thread 2 terminates

Main: Exiting

---

My problem experiented

but it ends up with that :

---

$ ./lib

Main: Creating threads

Main: Waiting for threads to finish

Hello #0 from Thread 1

Hello #0 from Thread 2

Hello #1 from Thread 1

Hello #1 from Thread 2

Hello #2 from Thread 1

Hello #2 from Thread 2

Hello #3 from Thread 1

Hello #3 from Thread 2

Hello #4 from Thread 1

Hello #4 from Thread 2

Hello #5 from Thread 1

Hello #5 from Thread 2

Hello #6 from Thread 1

Hello #6 from Thread 2

Hello #7 from Thread 1

Hello #7 from Thread 2

Hello #8 from Thread 1

Hello #8 from Thread 2

Hello #9 from Thread 1

Hello #9 from Thread 2

segmentation fault (core dumped)

---

The Way I compiled

in my ubuntu 14.14 machine I just simply typed this:

$ g++ lab.cpp -o lab -lpthread

I have also tryied with -pthread, too

$g++ lab.cpp -o lab -pthread

But, without luck!!

---

My code

this is my code:

#include <iostream>
#include <stdlib.h>
#include <pthread.h>
#include <unistd.h>

using namespace std;

void *print_message_function1(void *ptr);
void *print_message_function2(void *ptr);

int main(){
  cout << "\nMain: Creating threads" << endl;
  cout << "Main: Waiting for threads to finish" << endl << endl;

  pthread_t thread1, thread2;
  char message1[] = " from Thread 1";
  char message2[] = " from Thread 2";
  int iret1, iret2;



  iret1 = pthread_create( &thread1, 0, print_message_function1, (void*) message1);
  iret2 = pthread_create( &thread1, 0, print_message_function2, (void*) message2);



  pthread_join(thread1, 0);
  pthread_join(thread2, 0);

  cout << "Thread 0 terminates" << endl;
  cout << "Thread 1 terminates" << endl;
  cout << "Main: Exiting" << endl;
  exit(0);
}


void *print_message_function1(void *ptr){
 char *message;
 message = (char*) ptr;
 for(int i=0; i<10; i++){
   cout << "Hello #" << i << message << endl;
   sleep(1);
 }  
}

void *print_message_function2(void *ptr){
 char *message;
 message = (char*) ptr;
 for(int i=0; i<10; i++){
   cout << "Hello #" << i << message << endl;
   sleep(1);
 }  
}

Any one can see that problem and i will be very thankfull to solve that problem..

Answer 1

Looks like a typo. You use thread1 in both calls to pthread_create.

iret1 = pthread_create( &thread1, 0, print_message_function1, (void*) message1);
iret2 = pthread_create( &thread1, 0, print_message_function2, (void*) message2);

So pthread_join(thread2, 0); is pretty much doomed.

Answer 2

This is really just relevant information , not an answer as such, but unfortunately SO does not support code in comments.

The problem that you noticed with your code was a simple typo, but I didn't see that until I read the now accepted answer . For, I sat down and rewrote the code to standard C++, and in that process the typo disappeared, completely unnoticed! :) The problem that I did note was that you're using output statements without synchronization in your code, and that might cause output lines to be mixed, and is, I think, formally Undefined Behavior.

To fix that you can use a mutex . In this rewrite of your code I do not abstract up anything. But I think you can readily see the natural abstractions that hide between the code lines here:

#include <chrono>       // operator""ms
#include <iostream>
#include <mutex>
#include <stdlib.h>
#include <thread>

using namespace std;

mutex output_ownership;

void print_message_function1( char const* const message )
{
    for( int i = 0; i < 10; ++i )
    {
        {
            lock_guard<mutex> mux( output_ownership );
            cout << "Hello #" << i << message << endl;
        }
        this_thread::sleep_for( 1ms );
    }
    lock_guard<mutex> mux( output_ownership );
    cout << "Thread 0 terminates" << endl;
}

void print_message_function2( char const* const message )
{
    for( int i = 0; i < 10; ++i )
    {
        {
            lock_guard<mutex> mux( output_ownership );
            cout << "Hello #" << i << message << endl;
        }
        this_thread::sleep_for( 1ms );
    }
    lock_guard<mutex> mux( output_ownership );
    cout << "Thread 1 terminates" << endl;
}

auto main() -> int
{
    cout << "Main: Creating threads" << endl;
    cout << "Main: Waiting for threads to finish" << endl << endl;

    thread thread1( print_message_function1, " from Thread 1" );
    thread thread2( print_message_function2, " from Thread 2" );

    thread1.join();
    thread2.join();

    cout << "Main: Exiting" << endl;
}

This compiles and runs fine with Visual C++ 2015. It does not compile with MinGW-64 g++ 5.1.0, because its std::thread support is very much lacking. I do not know whether it will compile with latest version of g++ in Unix-land.