C++ faq-lite 11.10 placement new object destruction

[Fei Liu]

The faq states that an explicit call to the destructor must be made,
however, I don't see why a simple 'delete f' wouldn't work in this case.
At least my test code works correctly on FC5 with g++ 4.1.1. 'delete
f' correctly invokes 'f->~Fred()' and 'operator delete'.

void someCode()
{
char memory[sizeof(Fred)];
void* p = memory;
Fred* f = new(p) Fred();
...
f->~Fred(); // Explicitly call the destructor for the placed object
}

any comments?

Fei

 

[Victor Bazarov]

The faq states that an explicit call to the destructor must be made,
however, I don't see why a simple 'delete f' wouldn't work in this
case. At least my test code works correctly on FC5 with g++ 4.1.1.
'delete f' correctly invokes 'f->~Fred()' and 'operator delete'.

void someCode()
{
char memory[sizeof(Fred)];
void* p = memory;
Fred* f = new(p) Fred();
...
f->~Fred(); // Explicitly call the destructor for the placed
object }

any comments?

The actual memory for 'memory' is not allacated using 'new' or 'new[]',
so calling 'delete' (or 'delete[]') for it has undefined behaviour.

V

 

[Fei Liu]

The faq states that an explicit call to the destructor must be made,
however, I don't see why a simple 'delete f' wouldn't work in this
case. At least my test code works correctly on FC5 with g++ 4.1.1.
'delete f' correctly invokes 'f->~Fred()' and 'operator delete'.

void someCode()
{
char memory[sizeof(Fred)];
void* p = memory;
Fred* f = new(p) Fred();
...
f->~Fred(); // Explicitly call the destructor for the placed
object }

any comments?

The actual memory for 'memory' is not allacated using 'new' or 'new[]',
so calling 'delete' (or 'delete[]') for it has undefined behaviour.

The behavior is not undefined if Fred also provides 'operator delete'
that does not actually free any memory. The Faq needs to clarify on this
point instead of a plain statement that an explicit call to the
destructor must be made.

Fei

 

[Roland Pibinger]

The actual memory for 'memory' is not allacated using 'new' or 'new[]',
so calling 'delete' (or 'delete[]') for it has undefined behaviour.

The behavior is not undefined if Fred also provides 'operator delete'
that does not actually free any memory.

That "Fred also provides 'operator delete'" is not obvious from your
posted code. Moreover, I'm not sure that your assertion is backed by
the C++ Standard.

The Faq needs to clarify on this
point instead of a plain statement that an explicit call to the
destructor must be made.

Probably not (see Victor Bazarov's answer)