From mboxrd@z Thu Jan  1 00:00:00 1970
From: "Peter J. Stieber" <pete@toyon.com>
To: nobody@gcc.gnu.org
Cc: gcc-prs@gcc.gnu.org
Subject: Re: libstdc++/3181: Unable to use sqrt,cos,sin,... with int argument.
Date: Wed, 20 Jun 2001 09:26:00 -0000
Message-id: <20010620162603.1225.qmail@sourceware.cygnus.com>
X-SW-Source: 2001-06/msg00836.html
List-Id: <gcc-prs.sourceware.org>

The following reply was made to PR libstdc++/3181; it has been noted by GNATS.

From: "Peter J. Stieber" <pete@toyon.com>
To: <bkoz@gcc.gnu.org>, <nobody@gcc.gnu.org>, <gcc-prs@gcc.gnu.org>,
        <gcc-bugs@gcc.gnu.org>
Cc: <gcc-gnats@gcc.gnu.org>
Subject: Re: libstdc++/3181: Unable to use sqrt,cos,sin,... with int argument.
Date: Wed, 20 Jun 2001 09:16:12 -0700

 The book by Nicolai Josuttis "The C++ Standard Library - A Tutorial and
 Reference" has a good discussion of this topic (pp.581-582 in my copy). The
 fact that standard conforming behavior is defined as overloading the
 functions for all floating-point types (float, double and long double) leads
 to this problem. In the old days, non-standard function names were used
 (sqrtf, sqrtl, cosf, cosl,...) for the other floating point types, so
 implicit conversion would work. It is common in numeric programming
 (although not good practice) to see code like:
 
 double Pi = acos(-1);
 
 This will now cause problems, but it's always nice to work with legacy code.
 Since the old standard function names were defined for double
 FunctionName(double), you could overload for all numeric types and cast:
 
 namespace std
 {
   double sqrt(int a)
   {
     return sqrt(static_cast<double>(a));
   }
   double sqrt(long a)
   {
     return sqrt(static_cast<double>(a));
   }
 .
 .
 .
 };
 
 This behavior could be controlled by compiler switches or using the
 preprocessor.
 
 The work-around I use in my code is to make the following changes
 
 acos(-1)  change to acos(-1.0)
 
 or
 
 int i;
 acos(i) changes to acos(static_cast<double>(i))
 
 but for large projects, this would be a pain.