calling delete on the shallow copy not blow up

Question

In the below code, memory is allocated for an integer and later a shallow copy is being made and finally delete is being called on it. How does it still print 23 as the output and why doesn't the delete call on q, cause a run time exception.

#include <iostream>

using namespace std;

int main() {
    int* p = new int(23);
    int* q = p;

    delete p;
    cout << *p << endl;
    delete q;


    return 0;
}

[Undefined behaviour](https://en.cppreference.com/w/cpp/language/ub) is undefined. — George, Sep 12 '18 at 09:44

score 1 · Answer 1 · edited Sep 12 '18 at 09:45

1

Undefined behaviour means anything can happen.

It might crash.

It might crash your car.

It might crash your brain.

It might crash Sagittarius A* into your brain.

It might crash your brain into your car, then crash them both into Sagittarius A*.

It might appear to work.

But it's still undefined.

Don't expect results.

edited Sep 12 '18 at 09:45

melpomene

84,125
8
85
148

answered Sep 12 '18 at 09:44

Lightness Races in Orbit

378,754
76
643
1,055

@melpomene: Dammit – Lightness Races in Orbit Sep 12 '18 at 09:45
2

Can it be so undefined, that it changes the standard to become defined? – George Sep 12 '18 at 09:45
@George ... Dammit! – Lightness Races in Orbit Sep 12 '18 at 09:45
@George: To be fair, by the as-if rule, combined with the notion that the code could "appear to work", I suppose the answer is yes. – Lightness Races in Orbit Sep 12 '18 at 09:47
@George but why does delete q work Should not the memory be marked as unavailable and further call to delete cause a runtime error. Does calling delete again on the same memory location just mark it as unavailable again or is the behavior still undefined – Tousif Sep 12 '18 at 09:59
@Tousif Your assumption that there should be a runtime error has been invented. Who told you that there must be a runtime error? They may not be a runtime error. It's chance. It's undefined. It depends on the implementation. – Lightness Races in Orbit Sep 12 '18 at 10:00
@LightnessRacesinOrbit I am just trying to wrap my head around the concept of memory deallocation. Your comment is really appreciated. – Tousif Sep 12 '18 at 10:00
@Tousif I think the key point you may be missing is that memory "availability" is not checked on every access by the runtime - this would be an unacceptable performance hit. But from the C++ standards POV it _could be_ - it's undefined!! – Lightness Races in Orbit Sep 12 '18 at 10:01
@LightnessRacesinOrbit Thanks a lot. So, to be safe the delete q should not be called. – Tousif Sep 12 '18 at 10:03
@Tousif Absolutely. A double-delete is a serious bug. Ideally you wouldn't be doing manual memory management at all... – Lightness Races in Orbit Sep 12 '18 at 10:17
@Tousif to be safe you shouldnt have used `new` in the first place – 463035818_is_not_an_ai Sep 12 '18 at 10:17
1

@user463035818 it was a toy example to clarify my doubt. – Tousif Sep 12 '18 at 10:26

calling delete on the shallow copy not blow up

1 Answers1