Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!ut-emx!walt.cc.utexas.edu!jkuhn
From: jkuhn@walt.cc.utexas.edu (Jeff Kuhn)
Newsgroups: comp.lang.c++
Subject: Re: Derived can't access protected members of base's
Message-ID: <26464@ut-emx.UUCP>
Date: 20 Mar 90 03:04:20 GMT
References: <56.UUL1.3#5109@pantor.UUCP>
Sender: news@ut-emx.UUCP
Reply-To: jkuhn@walt.cc.utexas.edu (Jeff Kuhn)
Organization: The University of Texas at Austin, Austin, Texas
Lines: 54

In article <56.UUL1.3#5109@pantor.UUCP> richard@pantor.UUCP (Richard Sargent) writes:
>We just got bitten by a 2.0 "gotcha". I wonder if anyone "in the know"
>can explain the reasons for the restriction, and whether there is
>some one who can provide an explanation for doing what we are trying
>to do.
>
>C++ 2.0 very clearly states (acc.to Lippman) that "The derived class
>has no special access privilege to objects of its base class. Rather,
>the derived class has access privilege to the nonprivate inherited
>members of a derived class object." (See Lippman, page 314 and 316)
>
>This either was not the case in 1.2, or the compiler was more lax
>in regard to this rule.

The Zortech 2.0 version doesn't seem to catch the error, but the Intek 2.0
compiler and (I believe) the AT&T compiler's won't let you access base member
of another object from within a derived class.

>The way we see the situation, we have to disagree with the restriction.
>However, I am not foolish enough to think that BS does not have good
>reasons for doing so. I sure would like to understand, 'though.

I disagree too.

>To our mind, we have a class derivation tree (root class and a number
>of derived branches). We think that any object of a type from that
>tree of classes should be able to access the protected members of
>other objects also of that tree, but no class not in the tree.
>
>Assume we have an array of objects from that tree (or pointers to,
>it makes no real difference). We wish to traverse the array and
>access a protected member from each object (for whatever reason).
>It seems we should be able to use a "base*" pointer, and just
>access the item. So how can we accomplish this?
>
>Making the member public makes it too public :-).

The only way that I have found to get around this is to typecast your base
class pointer to a pointer to the derived class you want access through. This
is not the safest solution.

example:

derived::f()
{
	base *pb;

	((derived *)pb)->member;
}

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Jeff R. Kuhn	jkuhn@ccwf.cc.utexas.edu
Nth Graphics
Austin, Texas