>> Destructor Question

Hey all,

Code:

---

class Test {
    int m_Val;

    public:

    void printVal() { cout << m_Val << endl; }

    ~Test() { m_Val = 5;}
    Test(int val) { m_Val = val; printVal(); this->~Test(); }
};

int main()
{
    Test(5);

    Test a(10);

    a.printVal();

    return 0;
}

---

I was playing around a little, and found out that this not only compiles, but the last call produces the result "5". I though a destructor destroys the class, so the program shouldn't be able to access the variable. Or is this accessing something out of bounds?

Cheers,

Gabe

Quote:

---

Originally Posted by G4B3

I was playing around a little, and found out that this not only compiles, but the last call produces the result "5". I though a destructor destroys the class, so the program shouldn't be able to access the variable. Or is this accessing something out of bounds?

---

I uh, have never considered implementing a miscarriage instead of a constructor. Your program has undefined behaviour for two reasons:

• The explicit destructor call followed by the construction of the object on the stack means that the destructor will be called on a destroyed object (i.e., it will be called a second time, if that even makes sense).
• The member function printVal() is called on a destroyed object.

I would venture to guess that your explicit call to the destructor causes the destructor to be executed, setting m_Val to 5, but the destruction has little effect until something else overwrites the memory used by the object, which doesn't happen in your small program. Explicitly calling the destructor for an object created on the stack can cause serious problems, because the compiler will call the destructor again when the object goes out of scope.

There is only one case where calling the destructor explicitly is valid: when the instance was created with placement new.

Code:

---

#include <iostream>
using namespace std;

class Test {
    int m_Val;

    public:

    void printVal() { cout << m_Val << endl; }

    ~Test() { cout << "~Test\n"; }
    Test(int val) { m_Val = val; }
};

int main()
{
    char buffer[sizeof(Test)];

    //put a Test object into the buffer of bytes
    Test* p_test = new (buffer) Test(42);
    p_test->printVal();

    //now call the destructor, which wouldn't be called otherwise
    p_test->~Test();
    return 0;
}

---

Thanks for the responses. I know you're not supposed to do that even when mommy's not watching, as I said I was just playing around.

Anon: I'm not sure I understand what's going on there...I've never seen new used in such a way. Do you think you could give a more detailed explanation?

Quote:

---

Originally Posted by G4B3

Anon: I'm not sure I understand what's going on there...I've never seen new used in such a way. Do you think you could give a more detailed explanation?

---

You might want to search the Web for "placement new".

Wow, never heard of that before. Thanks laserlight ;-)

In particular, the article from C++ FAQ Lite would be a good starting point. The Wikipedia article also looks trustworthy. Danny Kalev's devx article is too short, but if you search for array placement new instead, his relevant devx article would be a good read.