Re: POD destruction
Diego Martins wrote:
kiryazev@gmail.com wrote:
"""alex wrote:
"""
I wonder what is the practical ground for this? :-)
The practical ground may be something like this
#include <iostream>
struct POD
{
int i;
double d;
};
POD get_pod()
{
static POD pod = {10, 3.14};
return pod;
}
struct X
{
~X()
{
//Here POD object is used after its destructor was
called. Do we get
UB?
std::cout << get_pod().i << ' ' << get_pod().d <<
'\n'; }
};
X x;
int main()
{
get_pod();
return 0;
}
Huh?
I can't see any usefulness in your code.
Again: what is the practical ground for this?
I can only be guessing here, but it seems that such code can have
its place. Imagine that 'pod' object in 'get_pod' is actually some
kind of singleton. It's initialised once (there is no need for it
to exist in more than one incarnation). The problem, however, is
in the fact that it's initialised after 'x', and will be destroyed
(have the d-tor invoked for it) *before* 'x', and if the destructor
of 'x' tries to use it (directly or indirectly), is 'pod' contents
still valid or not?
There is a possible work-around here, which is to dummy-construct
the 'pod' object before 'x':
// file scope
POD dummy = get_pod();
X x;
...
which removes the "'pod' destroyed before 'x' that uses it in its
own destructor" problem, but by introducing just one more static
object, one can easily tie them in a knot so that no reordering
would solve "I need to use it in the d-tor and it's been already
destroyed" problem:
struct A {
~A();
void operator();
};
struct B {
~B();
void operator();
};
struct C {
~C();
void operator();
};
A global_A;
B global_B;
C global_C;
A::~A() {
global_B();
}
B::~B() {
global_C();
}
C::~C() {
global_A();
}
Such code represents a design flaw that cannot really be fixed.
Or can it?
V
--
Please remove capital 'A's when replying by e-mail
I do not respond to top-posted replies, please don't ask
[ See http://www.gotw.ca/resources/clcm.htm for info about ]
[ comp.lang.c++.moderated. First time posters: Do this! ]