Is it possible to overload new at global scope and still call the original new from within that function?

Something like:

[code]
void *operator new(size_t Size)
{
   void *Ptr = ::operator new(Size);

   if(!Ptr)
   {
      // Memory allocation failed
      Output(MSG_FATAL, "Insufficient memory!\n");

      ExitProcess(0);
   }

   return Ptr;
}
[/code]

...except that doesn't recurse.

What are you trying to do? Provide extra handling if allocation fails? If so, see std::set_new_handler().

Ah, excellent. Thanks.

I'd use something else, like malloc or HeapAlloc. No scoping tricks needed there.

You could also use the throwing new with exception handlers, as a possible alternative to set_new_handler if this fits your program better.

is it safe to [color=blue]delete[/color] memory allocated with malloc? Pretty sure it is, but would like to confirm :)

[quote author=Arta[vL] link=board=23;threadid=5351;start=0#msg45012 date=1077180870]
is it safe to [color=blue]delete[/color] memory allocated with malloc? Pretty sure it is, but would like to confirm :)
[/quote]
No :)
Standard C++ says you can't mix them.

Maybe you can in specific compilers as an extension.

I still remember a classic:

HashTable::~HashTable()
{
delete table;
free(position);
}

.. or something very similar to that. It was about there that we scrapped it and started over :)

[quote author=Arta[vL] link=board=23;threadid=5351;start=0#msg45012 date=1077180870]
is it safe to [color=blue]delete[/color] memory allocated with malloc? Pretty sure it is, but would like to confirm :)
[/quote]
No. It's also not safe to delete memory allocated with new[]. Note that this is even true with VC - usually, if you mix the two, you'll cause heap corruption (at best) or crash immediately (at worst).

If you are overloading operator new, you should probably also overload operator delete.

In which case, my original question stands. How do I avoid the scoping issue? Just curious now :)