Leak with subject?

[diorcety]

I have issue with the following code (reduced for be simpliest as possible).
I'm want to unsubcribe from an observable coming from a subject.
I'm unsubcribed, but the lambda is kept somewhere...

#include <rxcpp/rx.hpp>
#include <memory>

int main(int argc, char *argv[]) {
 {
   ::rxcpp::subjects::subject<int> subject;
   auto s = subject.get_subscriber();
          s.on_next(1);
            s.on_next(2);
   auto ints = subject.get_observable();

 class String {
   private:
   std::string mS;
public:
   String(const std::string &s): mS(s) {
    std::cout << "Create" << std::endl;
   };
   ~String() {
    std::cout << "Destroy" << std::endl;
   }
   const std::string & getStr() const {
     return mS;
   }
 };

 rxcpp::subscription sub;
 {
  std::shared_ptr<String> str = std::make_shared<String>(std::string("Test"));
  sub = ints.subscribe(
        [=](int v) {
     std::cout << str->getStr() << std::endl;
    });
  }
  std::cout << "PU" << std::endl;
  sub.unsubscribe();
  std::cout << "AU" << std::endl;
}
  std::cout << "destroy subject" << std::endl;

 return 0;
}

I was expecting

Create
PU
Destroy
AU
destroy subject

but i have:

Create
PU
AU
Destroy
destroy subject

Is it normal?

[diorcety]

Seems to be a bug.
Change

std::cout << "PU" << std::endl;
sub.unsubscribe();
std::cout << "AU" << std::endl;

by

std::cout << "PU" << std::endl;
sub.unsubscribe();
ints.subscribe([=](int v) {});
std::cout << "AU" << std::endl;

Something miss here https://github.com/Reactive-Extensions/RxCpp/blob/master/Rx/v2/src/rxcpp/subjects/rx-subject.hpp#L67 in order to remove from the vector on unsubscription

Using ints.subscribe([=](int v) {}); truck allow to rework the observers vector an remove my old subscription.

[diorcety]

Juste after the L67 seems to almost work (but certainly ugly) :)

        o.add([=](){
          this->observers.erase(std::remove_if(observers.begin(), observers.end(), [&](auto l){
             return l.get_id() == o.get_id();
          }), observers.end());
        });

[kirkshoop]

There are issues with circular lifetimes and locks that need to be addressed.

I think I was able to fix this in #243.

Thank you for the report and diagnosis! This was very helpful!