upcasting pointer-to-pointer-to-class

[Judith]

Hi there everyone

I've got:

class A
{ ... };

class B : public A
{ ... }

void foo(A** a)
{ ... }

int main(...)
{
B* b = new B();
foo(&b);
}

GCC complains with "passing 'B**' as argument 1 of 'foo(A**)'"

Upcasting through one level of reference is most common, why is this different?

Thanks for any insight,

Judith

 

[Ron Natalie]

GCC complains with "passing 'B**' as argument 1 of 'foo(A**)'"

Upcasting through one level of reference is most common, why is this different?

Didn't we just answer this (look around on the group).

class Base { };
class Derived : public Base { };

Base b, *bp, **bpp;
Derived d, *dp, **dpp;

dp = &d;
bp = dp; // fine, we know how to convert from Derived* to Base* and can do so in this operation.
bpp = &dpp; // we convert from Derived** to Base** ok, but how does the value *dpp get converted
// to *bpp?

It's not a safe conversion because the compiler at the time *bpp is used doesn't know
what conversion might be needed to apply to the value to make it really a Base* value.

 

[tom_usenet]

Hi there everyone

I've got:

class A
{ ... };

class B : public A
{ ... }

void foo(A** a)
{ ... }

int main(...)
{
B* b = new B();
foo(&b);
}

GCC complains with "passing 'B**' as argument 1 of 'foo(A**)'"

Upcasting through one level of reference is most common, why is this different?

Here's why it doesn't work. Imagine that it did work, and then
consider this:

class A{};
class B: public A{};

void foo(A** a)
{
*a = new A;
}

int main()
{
B* bPtr;
foo(&bPtr);
//Oh no, bPtr now points at an A, but no compiler error!
}

The workaround is to change your main method to:

B* b = new B();
A* a = b;
foo(&a);
//a might be pointing to an A rather than a B...

Tom