7 replies to this topic

[Gamer Gamester]

Quick question. So I have a base class, and I make the destructor virtual to make sure any inherited classes are cleaned up properly:

class Base
{
  virtual ~Base() {}
}

Now, if I make a child class that inherits from Base, and it doesn't need to deallocate anything for itself, do I need to explicitly make another empty virtual destructor so that any further subclasses get cleaned up, or does the virtual carry through from the Base. Example:

class Sub : public Base
{
  // Various methods, but nothing to clean up so no destructor
}

class SubSub: public Sub
{
  ~SubSub();  // This has a destructor!!  Now if I call delete on a Base*,
              // will this destructor still get called without Sub also
              // having a virtual destructor?
}

I know I could just be safe and put a virtual destructor in each class, but I like to have minimal code, I find it makes everything easier.

[SiCrane]

Yes, virtual will implied for the version in derived classes. This applies to all functions declared virtual in the base class.

[Gamer Gamester]

Thanks.

To make sure I have a grasp on this: The virtual is implied for all functions in derived classes if these derived classes don't override the functions. Is this true also if the function is overridden? Or do I need to mark the overridden version virtual?

(In other words, from my previous example, should SubSub's destructor be virtual to "continue the virtuality"?)

[sigsegv42]

See C++ FAQ Lite entries [11.12] and [20.7]

[Gamer Gamester]

Hmm... the faq seems to make it seem as if once a method is declared virtual, it will remain virtual in all overridden versions, even if they're not marked virtual.

This comes as a surprise to me, as I recall reading in a recent book that you have to explicitly mark later versions virtual as well, or the "virtualness" ends there.

Perhaps the book I was reading was incorrect/out of date?

[Nitage]

Quote:

Hmm... the faq seems to make it seem as if once a method is declared virtual, it will remain virtual in all overridden versions, even if they're not marked virtual.

That's correct

Quote:

Perhaps the book I was reading was incorrect/out of date?

Very possibly. C++ has existed since the mid 1980s, but was only standardised in 1997. Anything written before then is most liekly useless. It's also a very complicated language - many of the people writing about it on the internet aren't expert and make mistakes. The C++ FAQ Lite is a trustworthy source though.

[Gamer Gamester]

Ah, this site explains this very clearly:
IBM Reference
It's the reference for IBM's AIX C++ Compiler, but I assume (perhaps foolishly[wink]) it follows the standard pretty close.

In particular this line covers it (B is subclass of A):

Quote:

If you declare a function named f in class B with the same name and same parameter list as A::f, then B::f is also virtual (regardless whether or not you declare B::f with the virtual keyword) and it overrides A::f.

Out of curiosity, does anybody recommend a good reference for C++ that complies with the standard? (though the standard will change soon with C++0x) Is Stroustrup's book best for that? I've been meaning to get it regardless but am waiting for the time being as it's a tad pricey. There's also the actual standard which you can download, is that a good idea? I'm concerned it'll be filled with stuff that's mostly of use to people creating/maintaining compilers and not so much for programmers who just want to use the language.

[sigsegv42]

Maybe take a look at one of Herb Sutter's Exceptional C++ books. They might be a little cheaper and will definitely be a good source of standards conformant information (Sutter is the chair of the ISO C++ committee). Also, Andrei Alexandrescu's Modern C++.