Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!dali.cs.montana.edu!caen!zaphod.mps.ohio-state.edu!wuarchive!news.uu.net!bywater!arnor!watson!blinn.watson.ibm.com!mittle
From: mittle@blinn.watson.ibm.com (Josh Mittleman)
Newsgroups: comp.lang.c++
Subject: Re: Derived/Base Class constructor relationship
Message-ID: <1991May22.153541.24133@watson.ibm.com>
Date: 22 May 91 15:35:41 GMT
References: <29534@hydra.gatech.EDU>
Sender: @watson.ibm.com
Distribution: na
Organization: IBM T. J. Watson Research
Lines: 31

In article <29534@hydra.gatech.EDU>, gt5595c@prism.gatech.EDU (BARTZ) writes:
> class A
> {
>    A(int f1,int f2,double first ...); // Constructor of base class
>    ...
> };
> 
> class B : public A
> {
> 	B(int g1,double first ...) : Matrix(g1,g1,first) {};
> 		 // Constructor for derived class
> 	...
> };

I assume that instead of "Matrix" you meant "A".

Ellipsis is not an argument, it is a compiler directive.  It means that the
compler may match a call with any number of trailing arguments with this
function declaration.  In particular, it is *not* a list of values.  When
you write A(g1, g1, first), all you are passing is the single value, first.

The only way to access the additional values which may follow first is to
use the va_list macros, or some similar method.  These only work within the
function which has the ellipsis signature.  I have been able to pass a
pointer to the list of values to second function by initializing a va_list
in the calling function, and then passing a reference to it.  In the
recipient function, va_arg works just fine with this reference.

===========================================================================
Josh Mittleman (mittle@watson.ibm.com or joshua@paul.rutgers.edu)
J2-C28 T.J. Watson Research Center, PO Box 704, Yorktown Heights, NY  10598