Setting member pointers to nullptr in destructor may not be good practice

[fcollot]

Following my remarks on #1131 I made a small test.

First I created the following class:

class Foo
{
public:
    int* i;

    Foo()
    {
        i = new int;
    }

    ~Foo()
    {
        delete i;
        i = nullptr;
    }
};

and executed this code:

Foo* foo = new Foo();
qDebug() << foo->i;
delete foo;
qDebug() << foo->i;

In debug build the pointer is being set to nullptr, but in release build it is not.

EDIT: I removed the test on double delete as it is not relevant. The first test proves that setting to nullptr will not work in release, and so will not protect from double delete as well.