[gnu.g++.bug] Is this a bug?

doug@asur.cs.cornell.edu (Douglas Campbell) (12/05/89)

I don't know whether the following is a bug or not.  Help, please?

Suppose a function takes a class instance (called by value), and it is
called with  a  reference  to an instance,  which  is returned from  a
function.  That is, the function is declared as:
		fn (CoolClass i)
and called as:
		fn (rcc())	// rcc returns type CoolClass&

Then shouldn't the object pointed to by rcc()'s return value be copied
into an argument for  fn, using either  bitwise-copy or a user-defined
X(X&) constructor?

If not, what?  If so, why does this fail to compile with g++ 1.36.1 on
both Sun3 and Sun4 os4:

----------------------------------------------------------------------
#include <stream.h>

class IntClass
{
    int i;
  public:
    IntClass() { i = 17; }  // Constructor
    friend IntClass& same (IntClass& ic) { return ic; }
    friend void pr(IntClass ic) { cout << ic.i << "\n"; }
};

main(int argc, char *argv[])
{
    IntClass a;
    pr(same(a));
}
----------------------------------------------------------------------

The compilation goes:

	g++ -g -c  test.cc
	test.cc: In function int main (int, char **):
	test.cc:16: too many arguments for constructor `IntClass'
	test.cc:16: in base initialization for class `IntClass'

where line 16 is:     pr(same(a));

The following adjustments to the source cause the problem to disappear:
	*  Make pr() called by reference
	*  Replace line 16 with lines:  IntClass& b = same(a); pr(b);
	*  Delete constructor for IntClass

I have this uneasy feeling that there is a trivial explanation, and
I'm going to look pretty silly.  Oh well, the price of education.

					Much thanks,
					Doug Campbell
					doug@svax.cs.cornell.edu