Reputation: 13
I'm working on creating a simple threaded producer/consumer example using separate classes, c++ standard library deques, and boost threads on Linux. I'm passing a shared buffer, lock, and condition variables by reference as member variables to my classes. The threads start up ok, but typically crash at a random time due to a lock assertion error.
main: ../nptl/pthread_mutex_lock.c:80: __pthread_mutex_lock: Assertion `mutex->__data.__owner == 0' failed.
Here's how I'm starting the threads in main
std::deque<double> buf;
boost::condition_variable_any buf_has_space;
boost::condition_variable_any buf_has_data;
boost::mutex buf_lock;
boost::thread producerThread(load_func, &buf, &buf_has_space, &buf_has_data, &buf_lock);
boost::thread consumerThread(consume_func, &buf, &buf_has_space, &buf_has_data, &buf_lock);
producerThread.join();
consumerThread.join();
producerThread
and consumerThread
initialize and run instances of myProducer
and myConsumer
.
Relevant code from myProducer
:
void myProducer::add_to_buffer(){
//Main loop
for (int i=0; i<100000; i++){
boost::mutex::scoped_lock lock(*buffer_lock);
while (buffer->size() == max_size){
buffer_has_space->wait(*buffer_lock);
}
buffer->push_back(i);
buffer_has_data->notify_one();
buffer_lock -> unlock();
}
//Consumer will stop when it sees -1
boost::mutex::scoped_lock lock(*buffer_lock);
while (buffer->size() == max_size){
buffer_has_space->wait(*buffer_lock);
}
buffer->push_back(-1.0);
buffer_has_data->notify_one();
buffer_lock -> unlock();
}
Relevant code from myConsumer
:
void myConsumer::load_from_buffer(){
double current = 0;
while (current != -1.0) {
boost::mutex::scoped_lock lock(*buffer_lock);
while (buffer->size() == 0){
buffer_has_data->wait(*buffer_lock);
}
current = buffer->front();
buffer->pop_front();
buffer_has_space->notify_one();
buffer_lock->unlock();
std::cout << current <<"\n";
}
}
I've looked at these questions:
pthread_mutex_lock.c:62: __pthread_mutex_lock: Assertion `mutex->__data.__owner == 0' failed
However, the program is crashing before my classes are deconstructed, and each lock is matched with an unlock in the same thread.
Upvotes: 1
Views: 1991
Reputation: 392833
If you have the lock owned by a lock guard (scoped_lock) it's an error to manually manipulate the lock.
If you must manipulate the lock before the end of the scope, you can do so on the scoped_lock
itself:
lock.unlock(); // instead of buffer_lock -> unlock();
Upvotes: 2