From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-return-72487-listarch-gcc=gcc.gnu.org@gcc.gnu.org>
Received: (qmail 11864 invoked by alias); 24 Apr 2003 13:12:14 -0000
Mailing-List: contact gcc-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Archive: <http://gcc.gnu.org/ml/gcc/>
List-Post: <mailto:gcc@gcc.gnu.org>
List-Help: <http://gcc.gnu.org/ml/>
Sender: gcc-owner@gcc.gnu.org
Received: (qmail 11853 invoked from network); 24 Apr 2003 13:12:11 -0000
Received: from unknown (HELO synaptics-uk.com) (194.203.111.210)
  by sources.redhat.com with SMTP; 24 Apr 2003 13:12:11 -0000
Received: from uk_exchange.synaptics-uk.com ([194.203.111.213])
	by synaptics-uk.com (8.9.3/8.9.3) with ESMTP id OAA06945;
	Thu, 24 Apr 2003 14:10:51 +0100 (BST)
Received: from dogbert.synaptics-uk.com ([172.20.11.5]) by uk_exchange.synaptics-uk.com with SMTP (Microsoft Exchange Internet Mail Service Version 5.5.2653.13)
	id JMMLNZF8; Thu, 24 Apr 2003 14:12:00 +0100
From: Gareth McCaughan <gmccaughan@synaptics-uk.com>
To: Ivan Molella <i.molella@reply.it>
Subject: Re: compiler confusion?
Date: Thu, 24 Apr 2003 15:12:00 -0000
User-Agent: KMail/1.5
Cc: gcc@gcc.gnu.org
MIME-Version: 1.0
Content-Type: text/plain;
  charset="us-ascii"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
Message-Id: <200304241411.01100.gmccaughan@synaptics-uk.com>
X-SW-Source: 2003-04/txt/msg01230.txt.bz2

Here's a stripped-down version of your code that exhibits
the same problem.

    struct A {
      A(bool x) {}
    };
    
    struct C {
      void f();
    };
    
    static int g() { return 1; }
    void C::f() {
      A(g() == 100);
    }

Here's what I think is going on: the compiler is seeing
that line "A(g() == 100);" as a declaration of a function
called g, returning a value of type A. The same as if it said
"A g() == 100;". And, of course, when it gets to the "==",
that stops making any sense :-).

I'm not enough of a C++ standards guru to know whether
this behaviour is (1) allowed and/or (2) required. It's
weird, certainly, but there are other almost equally
ludicrous interpretations that *are* required by the
standard. For instance, here's an example from Scott Meyers's
"Effective STL".

    ifstream dataFile("ints.dat");
    list<int> data(istream_iterator<int>(dataFile),
                   istream_iterator<int>());

Perhaps contrary to appearances, what this does is
to declare a function called "data" taking two
arguments: the first is of type istream_iterator<int>
and is named "dataFile", and the second is of type
"pointer to function from void to istream_iterator<int>"
and has no name. :-)

-- 
g