Why is unique_ptr::release not defined with [[nodiscard]]?

问题

C++17 added [[nodiscard]].

C++20 added the use of [[nodiscard]] on empty methods, e.g. vector::empty() -- maybe, to avoid user confusion with the method clear (i.e. calling empty() accidentally to clear the vector).

Why didn't C++20 use this opportunity to add [[nodiscard]] to unique_ptr::release?

Is there a valid reasonable scenario in which one would call unique_ptr::release without taking the returned value?

In the same manner of avoiding user confusion (if this was the reason for adding [[nodiscard]] to the empty methods) - the name release was always very confusing, sounds like, well... something is going to be released here.

Adding [[nodiscard]] could fix this name issue, in a way.

回答1:

This is addressed in the paper that added [[nodiscard]] to many of the functions. From P0600R1 this is the remark about adding [[nodiscard]] to unique_ptr::release()

Titus: at Google 3.5% of calls would fail, but analysis showed that it was correct (but weird ownership semantics). See reflector email.

回答2:

Because you've previously retrieved the pointer value and done stuff with it.

Simple approximation:

unique_ptr<someclass> ptr;
// ...
someclass *s = ptr.get();
if (s->are_we_there_yet()) {
    ptr.release();
    // finish up with s...
    s->close_garage_door();
    delete s;
}

回答3:

// returns true on success
bool run_a_thing(void (*)(void*), void* context);

struct state {
    // whatever
};

void runner(void* context) {
    std::unique_ptr<state> s(static_cast<state*>(context));
    // do things
}

void run_thing() {
    auto s = std::make_unique<state>(....);
    if (run_a_thing(runner, s.get())) {
        s.release();
    }
}

This is basically the structure of libstdc++'s std::thread. run_a_thing is pthread_create.

来源：https://stackoverflow.com/questions/60535399/why-is-unique-ptrrelease-not-defined-with-nodiscard

标签

c++

unique-ptr

nodiscard