CODEWITHSUNDEEP
c++c++11
Irbis
Irbis

Reputation: 1491

Different ways to destructing unique_ptr

I analyze a class with unique_ptr member. That member may be used from different threads. There is also a method which destroys that member:

void uninitialize()
{
    std::unique_ptr<Worker> worker;
    {
        std::lock_guard<std::mutex> guard(mtx_);
        worker = std::move(worker_);
    }
}

I wonder what is purpose of that impl. Is there any difference between the above and the following implementation? :

void uninitialize()
{ 
    std::lock_guard<std::mutex> guard(mtx_);
    worker_.reset();
}

Worker doesn't define move constructor.

Upvotes: 4

Views: 305

Answers (2)

Jesper Juhl
Jesper Juhl

Reputation: 31465

In the first example, worker ends up owning the unique_ptr (until it goes out of scope, without the lock held) In the second it is just destroyed (while the lock is held).

Upvotes: 0

Brian Bi
Brian Bi

Reputation: 119184

The difference is that in the first snippet, the mutex will only be held for long enough for worker to take ownership of the object that worker_ owns; the destruction of that object will be unguarded. In the second snippet, the mutex will not be released until .reset() completes, which means it is held while the object is being destroyed.

Upvotes: 9

Related Questions