Delete a class with multiple inheritance through base pointer?

[Ian]

Given

struct A{...};
struct B{...};

struct C : public A, public B {...}

int main()
{
A* a = new C;

delete a;

return 0;
}

Will the B part of a be correctly deleted?

Ian

[John Harrison]

"Ian" <no***@nowhere.com> wrote in message
news:40**************@nowhere.com...

Given

struct A{...};
struct B{...};

struct C : public A, public B {...}

int main()
{
A* a = new C;

delete a;

return 0;
}

Will the B part of a be correctly deleted?

Ian

Its not a question of whether or not the B part will be correctly deleted.
Its a question of whether the code is correct or not, if it is not the
consequences are undefined.

The answer is that the code is correct if A has a virtual destructor,
otherwise you have undefined behaviour.

john

[Gianni Mariani]

Ian wrote:

Given

struct A{...};
struct B{...};

struct C : public A, public B {...}

int main()
{
A* a = new C;

delete a;

return 0;
}

Will the B part of a be correctly deleted?
only if A has a virtual destructor.

i.e.

struct A{ virtual ~A(){} ...}; // not the virtual destructor
struct B{...};

struct C : public A, public B {...}

int main()
{
A* a = new C;

delete a;

return 0;
}

Ian

[Ian]

I left out the virtual destructors in A and B...

Should work, but I am seeing problems where it does not.

Ian wrote:

Given

struct A{...};
struct B{...};

struct C : public A, public B {...}

int main()
{
A* a = new C;

delete a;

return 0;
}

Will the B part of a be correctly deleted?

Ian

[John Harrison]

"Ian" <no***@nowhere.com> wrote in message
news:40**************@nowhere.com...

I left out the virtual destructors in A and B...

Should work, but I am seeing problems where it does not.

What problems are you seeing?

john

[Ian]

John Harrison wrote:

"Ian" <no***@nowhere.com> wrote in message
news:40**************@nowhere.com...

I left out the virtual destructors in A and B...

Should work, but I am seeing problems where it does not.

What problems are you seeing?

What I think is a compiler bug.... Members of one base being invalid in
the destructor of the derived class.

Ian

[John Harrison]

"Ian" <no***@nowhere.com> wrote in message
news:40**************@nowhere.com...

John Harrison wrote:

"Ian" <no***@nowhere.com> wrote in message
news:40**************@nowhere.com...

I left out the virtual destructors in A and B...

Should work, but I am seeing problems where it does not.

What problems are you seeing?

What I think is a compiler bug.... Members of one base being invalid in
the destructor of the derived class.

Ian

Compiler bugs are always a possibility but the code you posted is very
simple (maybe your real code is more complex) so I'd be surprised if a
compiler got that wrong.

john

[Chris Theis]

"Ian" <no***@nowhere.com> wrote in message
news:40**************@nowhere.com...

John Harrison wrote:

"Ian" <no***@nowhere.com> wrote in message
news:40**************@nowhere.com...

I left out the virtual destructors in A and B...

Should work, but I am seeing problems where it does not.

What problems are you seeing?

What I think is a compiler bug.... Members of one base being invalid in
the destructor of the derived class.

Could you elaborate on that (probably with some sample code & which compiler
you're using). From your code I do not see a problem (with the virtual
declartions being added) as the order of destruction is simply the reverse
order of construction. Thus the dtor of the derived class is called before
the one of the base class.

Regards
Chris