I have a program where I start multiple, long running threads (such as a REST-API). On primed signals (e.g SIGHUP) I would like to be able to shut down all threads cleanly (by waiting for them to exit). Below follows some code from a thispointer article that illustrated a good idea on how to do this
#include <thread>
#include <iostream>
#include <assert.h>
#include <chrono>
#include <future>
void threadFunction(std::future<void> futureObj)
{
std::cout << "Thread Start" << std::endl;
while (futureObj.wait_for(std::chrono::milliseconds(1)) ==
std::future_status::timeout)
{
std::cout << "Doing Some Work" << std::endl;
std::this_thread::sleep_for(std::chrono::milliseconds(1000));
}
std::cout << "Thread End" << std::endl;
}
int main()
{
// Create a std::promise object
std::promise<void> exitSignal;
//Fetch std::future object associated with promise
std::future<void> futureObj = exitSignal.get_future();
// Starting Thread & move the future object in lambda function by reference
std::thread th(&threadFunction, std::move(futureObj));
//Wait for 10 sec
std::this_thread::sleep_for(std::chrono::seconds(10));
std::cout << "Asking Thread to Stop" << std::endl;
//Set the value in promise
exitSignal.set_value();
//Wait for thread to join
th.join();
std::cout << "Exiting Main Function" << std::endl;
return 0;
}
However, as one might have noticed this concept has a critical drawback: the exitSignal will have to be emitted before th.join() is called.
In a situation where one wants to listen to a signal, e.g using signal(SIGHUP, callback) this is of course impractical.
My question is: are there better concepts for shutting down multiple threads? How would I go about them? I think using a promise is not a bad idea, I just haven't found a way with it to solve my problem.
You can use std::notify_all_at_thread_exit() on a std::condition_variable.
Here is an example:
#include <mutex>
#include <thread>
#include <condition_variable>
#include <cassert>
#include <string>
std::mutex m;
std::condition_variable cv;
bool ready = false;
std::string result; // some arbitrary type
void thread_func()
{
thread_local std::string thread_local_data = "42";
std::unique_lock<std::mutex> lk(m);
// assign a value to result using thread_local data
result = thread_local_data;
ready = true;
std::notify_all_at_thread_exit(cv, std::move(lk));
} // 1. destroy thread_locals;
// 2. unlock mutex;
// 3. notify cv.
int main()
{
std::thread t(thread_func);
t.detach();
// do other work
// ...
// wait for the detached thread
std::unique_lock<std::mutex> lk(m);
cv.wait(lk, [] { return ready; });
// result is ready and thread_local destructors have finished, no UB
assert(result == "42");
}
Source: cppreference.com
Related
Scenario:
I have a condition_variable based wait and signal mechanism. This works! But I need a little more than just the classic wait and signal mechanism. I need to be able to do a timed wait as well as an infinite wait "on the same condition_variable". Hence, I created a wrapper class around a condition_variable which takes care of the spurious wake up issue as well. Following is the code for that:
Code:
// CondVarWrapper.hpp
#pragma once
#include <mutex>
#include <chrono>
#include <condition_variable>
class CondVarWrapper {
public:
void Signal() {
std::unique_lock<std::mutex> unique_lock(cv_mutex);
cond_var_signalled = true;
timed_out = false;
unique_lock.unlock();
cond_var.notify_one();
}
bool WaitFor(const std::chrono::seconds timeout) {
std::unique_lock<std::mutex> unique_lock(cv_mutex);
timed_out = true;
cond_var.wait_for(unique_lock, timeout, [this] {
return cond_var_signalled;
});
cond_var_signalled = false;
return (timed_out == false);
}
bool Wait() {
std::unique_lock<std::mutex> unique_lock(cv_mutex);
timed_out = true;
cond_var.wait(unique_lock, [this] {
return cond_var_signalled;
});
cond_var_signalled = false;
return (timed_out == false);
}
private:
bool cond_var_signalled = false;
bool timed_out = false;
std::mutex cv_mutex;
std::condition_variable cond_var;
};
// main.cpp
#include "CondVarWrapper.hpp"
#include <iostream>
#include <string>
#include <thread>
int main() {
CondVarWrapper cond_var_wrapper;
std::thread my_thread = std::thread([&cond_var_wrapper]{
std::cout << "Thread started" << std::endl;
if (cond_var_wrapper.WaitFor(std::chrono::seconds(10))) {
std::cout << "Thread stopped by signal from main" << std::endl;
} else {
std::cout << "ERROR: Thread stopping because of timeout" << std::endl;
}
});
std::this_thread::sleep_for(std::chrono::seconds(3));
// Uncomment following line to see the timeout working
cond_var_wrapper.Signal();
my_thread.join();
}
Question:
Above code is good but I think there is one problem? Would I really be able to do a wait as as well do a wait_for on the same condition_variable? What if a thread has acquired cv_mutex by calling CondVarWrapper::Wait() and this one never returned for some reason. And then another thread comes in calling CondVarWrapper::WaitFor(std::chrono::seconds(3)) expecting to return out if it does not succeed in 3 seconds. Now, this second thread would not be able to return out of WaitFor after 3 seconds isnt it? In fact it wouldn't ever return. Because the condition_variable wait is a timed wait but not the lock on cv_mutex. Am I correct or Am I wrong in understanding here?
If I am correct above then I need to replace std::mutex cv_mutex with a std::timed_mutex cv_mutex and do a timed_wait in CondVarWrapper::WaitFor and do a infinite wait on CondVarWrapper::Wait? Or are there any better/easier ways of handling it?
The mutex is released when calling std::condition::wait on the condition variable cond_var. Thus, when you call CondVarWrapper::Wait from one thread, it releases the mutex when calling std::condition::wait and it hangs in there forever, the second thread can still call CondVarWrapper::WaitFor and successfully lock the mutex cv_mutex.
I have a simple C++11 thread program like below.
Code:
#include <iostream>
#include <thread>
#include <chrono>
#include <atomic>
int main(int argc, char *argv[]) {
std::cout << "My program starts" << std::endl;
std::atomic<bool> exit_thread(false);
std::thread my_thread = std::thread([&exit_thread]{
do {
std::cout << "Thread is doing something..." << std::endl;
std::this_thread::sleep_for(std::chrono::seconds(5));
} while (!exit_thread);
});
std::this_thread::sleep_for(std::chrono::seconds(12));
exit_thread = true;
std::cout << "Might have to wait to exit thread" << std::endl;
my_thread.join();
return 0;
}
As you can see above, there is a loop which has a sleep_for which makes the thread sleep for 5 seconds and then it wakes and loops again provided that exit_thread is set to false. Main thread waits for 12 seconds and prepares to exit firstly by setting exit_thread to true and then does a join on the thread. All good until now.
Problem:
Above is okay and works for objective. But there is a "potential problem". If the thread has just now started to sleep then it would take it 4 seconds more before it gets out of sleep to discover that it now needs to exit. This delays the exit process and destruction.
Question:
How to can I make the thread sleep in an interruptible way? So that I can interrupt the sleep and make the thread exit right away instead by cancelling out of sleep instead of waiting for the potential 4 or 3 or 2 seconds.
I think that the solution to this might be achievable using a std::condition_variable? Probably? I am looking for a piece of code to show how.
Note that my code runs on both clang and gcc.
We should be waiting on a condition variable or semaphore instead of sleeping. Here's the minimal change to do that:
#include <atomic>
#include <chrono>
#include <condition_variable>
#include <iostream>
#include <mutex>
#include <thread>
int main()
{
std::cout << "My program starts" << std::endl;
std::atomic<bool> exit_thread(false);
std::condition_variable cv;
std::mutex m;
std::thread my_thread = std::thread([&exit_thread,&cv,&m]{
do {
std::cout << "Thread is doing something..." << std::endl;
{
std::unique_lock<std::mutex> lock(m);
cv.wait_for(lock, std::chrono::seconds(5));
}
} while (!exit_thread);
});
std::this_thread::sleep_for(std::chrono::seconds(12));
{
std::lock_guard<std::mutex> guard(m);
exit_thread = true;
}
cv.notify_all();
std::cout << "Thread stops immediately" << std::endl;
my_thread.join();
}
Apparently, we do need the mutex:
Even if the shared variable is atomic, it must be modified under the
mutex in order to correctly publish the modification to the waiting
thread.
I am using code that runs on ARM (not Intel processor). Running c++11 code example (CODE A) from: http://www.cplusplus.com/reference/condition_variable/condition_variable/wait_for/ to test the wait_for() mechanism. This is not working right - looks like the wait_for() does not wait. In Intel works fine. After some research and using pthread library directly and setting MONOTONIC_CLOCK definition, solves the issue (CODE B).
(Running on ARM is not the issue)
My problem is :
How can I force the C++11 API wait_for() to work with MONOTONIC_CLOCK?
Actually I would like to stay with 'CODE A' but with the support or setting of MONOTONIC_CLOCK.
Thanks
CODE A
// condition_variable::wait_for example
#include <iostream> // std::cout
#include <thread> // std::thread
#include <chrono> // std::chrono::seconds
#include <mutex> // std::mutex, std::unique_lock
#include <condition_variable> // std::condition_variable, std::cv_status
std::condition_variable cv;
int value;
void read_value() {
std::cin >> value;
cv.notify_one();
}
int main ()
{
std::cout << "Please, enter an integer (I'll be printing dots): \n";
std::thread th (read_value);
std::mutex mtx;
std::unique_lock<std::mutex> lck(mtx);
while
(cv.wait_for(lck,std::chrono::seconds(1))==std::cv_status::timeout)
{
std::cout << '.' << std::endl;
}
std::cout << "You entered: " << value << '\n';
th.join();
return 0;
}
CODE B
#include <sys/time.h>
#include <unistd.h>
#include <iostream> // std::cout
#include <thread> // std::thread
#include <chrono> // std::chrono::seconds
#include <mutex> // std::mutex, std::unique_lock
#include <condition_variable> // std::condition_variable, std::cv_status
const size_t NUMTHREADS = 1;
pthread_mutex_t mutex;
pthread_cond_t cond;
int value;
bool done = false;
void* read_value( void* id )
{
const int myid = (long)id; // force the pointer to be a 64bit integer
std::cin >> value;
done = true;
printf( "[thread %d] done is now %d. Signalling cond.\n", myid, done
);
pthread_cond_signal( &cond );
}
int main ()
{
struct timeval now;
pthread_mutexattr_t Attr;
pthread_mutexattr_init(&Attr);
pthread_mutexattr_settype(&Attr, PTHREAD_MUTEX_RECURSIVE);
pthread_mutex_init(&mutex, &Attr);
pthread_condattr_t CaAttr;
pthread_condattr_init(&CaAttr);
pthread_condattr_setclock(&CaAttr, CLOCK_MONOTONIC);
pthread_cond_init(&cond, &CaAttr);
std::cout << "Please, enter an integer:\n";
pthread_t threads[NUMTHREADS];
int t = 0;
pthread_create( &threads[t], NULL, read_value, (void*)(long)t );
struct timespec ts;
pthread_mutex_lock( &mutex );
int rt = 0;
while( !done )
{
clock_gettime(CLOCK_MONOTONIC, &ts);
ts.tv_sec += 1;
rt = pthread_cond_timedwait( & cond, & mutex, &ts );
std::cout << "..." << std::endl;
}
pthread_mutex_unlock( & mutex );
std::cout << "You entered: " << value << '\n';
return 0;
}
The documentation for std::condition_variable::wait_for says:
A steady clock is used to measure the duration.
std::chrono::steady_clock:
Class std::chrono::steady_clock represents a monotonic clock. The time points of this clock cannot decrease as physical time moves forward.
Unfortunately, this is gcc Bug 41861 (DR887) - (DR 887)(C++0x) does not use monotonic_clock that it uses system_clock instead of steady_clock for condition variables.
One solution is to use wait_until (be sure to read Notes section) function that allows to specify durations relative to a specific clock. E.g.:
cv.wait_until(lck, std::chrono::steady_clock::now() + std::chrono::seconds(1))
I am trying to teach myself C++11 threading, and I would like to start a background producer thread at the beginning of the application, and have it run until application exit. I would also like to have consumer thread (which also runs for the life of the application).
A real-world example would be a producer thread listening on a Com port for incoming GPS data. Once a full message had been accumulated, it could be parsed to see if it was a message of interest, then converted into a string (say), and 'delivered back' to be consumed (update current location, for example).
My issue is I haven't been able to figure out how to do this without blocking the rest of the application when I 'join()' on the consumer thread.
Here is my very simplified example that hopefully shows my issues:
#include <QCoreApplication>
#include <QDebug>
#include <thread>
#include <atomic>
#include <iostream>
#include <queue>
#include <mutex>
#include <chrono>
#include "threadsafequeuetwo.h"
ThreadSafeQueueTwo<int> goods;
std::mutex mainMutex;
std::atomic<bool> isApplicationRunning = false;
void theProducer ()
{
std::atomic<int> itr = 0;
while(isApplicationRunning)
{
// Simulate this taking some time...
std::this_thread::sleep_for(std::chrono::milliseconds(60));
// Push the "produced" value onto the queue...
goods.push(++itr);
// Diagnostic printout only...
if ((itr % 10) == 0)
{
std::unique_lock<std::mutex> lock(mainMutex);
std::cout << "PUSH " << itr << " on thread ID: "
<< std::this_thread::get_id() << std::endl;
}
// Thread ending logic.
if (itr > 100) isApplicationRunning = false;
}
}
void theConsumer ()
{
while(isApplicationRunning || !goods.empty())
{
int val;
// Wait on new values, and 'pop' when available...
goods.waitAndPop(val);
// Here, we would 'do something' with the new values...
// Simulate this taking some time...
std::this_thread::sleep_for(std::chrono::milliseconds(10));
// Diagnostic printout only...
if ((val % 10) == 0)
{
std::unique_lock<std::mutex> lock(mainMutex);
std::cout << "POP " << val << " on thread ID: "
<< std::this_thread::get_id() << std::endl;
}
}
}
int main(int argc, char *argv[])
{
std::cout << "MAIN running on thread ID: "
<< std::this_thread::get_id() << std::endl;
// This varaiable gets set to true at startup, and,
// would only get set to false when the application
// wants to exit.
isApplicationRunning = true;
std::thread producerThread (theProducer);
std::thread consumerThread (theConsumer);
producerThread.detach();
consumerThread.join(); // BLOCKS!!! - how to get around this???
std::cout << "MAIN ending on thread ID: "
<< std::this_thread::get_id() << std::endl;
}
The ThreadSafeQueueTwo class is the thread safe queue implementation taken almost exactly as is from the "C++ Concurrency In Action" book. This seems to work just fine. Here it is if anybody is interested:
#ifndef THREADSAFEQUEUETWO_H
#define THREADSAFEQUEUETWO_H
#include <queue>
#include <memory>
#include <mutex>
#include <condition_variable>
template<typename T>
class ThreadSafeQueueTwo
{
public:
ThreadSafeQueueTwo()
{}
ThreadSafeQueueTwo(ThreadSafeQueueTwo const& rhs)
{
std::lock_guard<std::mutex> lock(myMutex);
myQueue = rhs.myQueue;
}
void push(T newValue)
{
std::lock_guard<std::mutex> lock(myMutex);
myQueue.push(newValue);
myCondVar.notify_one();
}
void waitAndPop(T& value)
{
std::unique_lock<std::mutex> lock(myMutex);
myCondVar.wait(lock, [this]{return !myQueue.empty(); });
value = myQueue.front();
myQueue.pop();
}
std::shared_ptr<T> waitAndPop()
{
std::unique_lock<std::mutex> lock(myMutex);
myCondVar.wait(lock, [this]{return !myQueue.empty(); });
std::shared_ptr<T> sharedPtrToT (std::make_shared<T>(myQueue.front()));
myQueue.pop();
return sharedPtrToT;
}
bool tryPop(T& value)
{
std::lock_guard<std::mutex> lock(myMutex);
if (myQueue.empty())
return false;
value = myQueue.front();
myQueue.pop();
return true;
}
std::shared_ptr<T> tryPop()
{
std::lock_guard<std::mutex> lock(myMutex);
if (myQueue.empty())
return std::shared_ptr<T>();
std::shared_ptr<T> sharedPtrToT (std::make_shared<T>(myQueue.front()));
myQueue.pop();
return sharedPtrToT;
}
bool empty()
{
std::lock_guard<std::mutex> lock(myMutex);
return myQueue.empty();
}
private:
mutable std::mutex myMutex;
std::queue<T> myQueue;
std::condition_variable myCondVar;
};
#endif // THREADSAFEQUEUETWO_H
Here's the output:
I know there are obvious issues with my example, but my main question is how would I run something like this in the background, without blocking the main thread?
Perhaps an even better way of trying to solve this is, is there a way that every time the producer has 'produced' some new data, could I simply call a method in the main thread, passing in the new data? This would be similar to queued signal/slots it Qt, which I am big fan of.
I'm trying to send multiple notifications to a running thread from another thread (main thread) using std::condition_variable. Sending it once works however doing it the second or multiple times doesn't seem to work. This is what I did (without unnecessary details of the actual events):
#include <iostream>
#include <thread>
#include <mutex>
#include <condition_variable>
#include <future>
bool keep_running=true;
bool condition_reached=false;
std::mutex cond_mtx;
std::condition_variable cond;
void thread_waiting_to_be_notified(){
while(keep_running){
std::unique_lock<std::mutex> lk(cond_mtx);
cond.wait(lk,[]()->bool{return condition_reached;});
std::cout << "got notitication" << std::endl;
condition_reached=false;
}
}
void some_event(){
/*some event happens here*/
}
void another_event(){
/*another event happens here*/
}
int main(){
std::thread thr(thread_waiting_to_be_notified);
some_event();//first event
std::cout << "some event happened" << std::endl;
condition_reached=true;
cond.notify_one();
another_event();//second event
std::cout << "another event happened" << std::endl;
condition_reached=true;
cond.notify_one();
keep_running=false;
thr.join();
return 0;
}
and the output I got
some event happened
another event happened
got notitication
However, I'd expect
some event happened
another event happened
got notitication
got notitication
Any advice would be appreciated.
Try inserting lk.unlock(); after
condition_reached=false;