Is this template problem in ecgs?

Is this template problem in ecgs?

[Nicholas Lee]

---8<---

#include <iostream>

template<typename T> 
inline
T& 
min(const T& x, const T& y) 
{
  return (x<y)? x:y;
}

int 
main()
{
  int a = 1;
  const int b = 10;

  cout << "a = " << a << ", b = " << b << endl; 
  cout << "min(a,b) = " << min(a,b) << endl;

  min(a,b) = 20;
  cout << "a = " << a << ", b = " << b << endl; 
  cout << "min(a,b) = " << min(a,b) << endl;

  min(a,b) = 30;
  cout << "a = " << a << ", b = " << b << endl; 
  cout << "min(a,b) = " << min(a,b) << endl;
}

/*
10:18am> g++ -Wall test2.cc -otest2
10:18am>./test2
a = 1, b = 10
min(a,b) = 1
a = 20, b = 10
min(a,b) = 10
a = 20, b = 10
min(a,b) = 10

10:18am> gcc -v

gcc version egcs-2.90.29 980515 (egcs-1.0.3 release)
(SunOs 5.6)

=== gcc 2.8.1 version ==

10:33:30 advlab$ gcc -V 2.8.1 test2.cc -o test2 -lstdc++
10:33:38 advlab$ ./test2
a = 1, b = 10
min(a,b) = 1
a = 20, b = 10
min(a,b) = 10
a = 20, b = 10
min(a,b) = 10



==== SGI Version ==

Slightly modified about to
<#include <iostreams>
>#include <iostreams.h>
<template <typename T>
>template <class T>
 
===

10:21am> uname 
IRIX64

10:19am> CC test2.cc -o test2sgi
"test2.cc", line 9: error(3514): initial value of reference has excess
          const/volatile qualifiers
    return (x<y)? x:y;
           ^
          detected during instantiation of
                    "int & min(const int &, const int &)" at line 18

1 error detected in the compilation of "test2.cc".

*/

---8<----
--
	"I reserve the right to contradict myself."

Nicholas Lee (Li Peng Ming)  n.lee with math.auckland.ac in nz
Auckland University, New Zealand	

Re: Is this template problem in ecgs?

[Alexandre Oliva]

Nicholas Lee <nlee011@math.auckland.ac.nz> writes:

> template<typename T>
> inline T& min(const T& x, const T& y) { return (x<y)? x:y; }

Not a template problem, actually.  It would also compile if you
replaced `template<typename T>' with `typedef int T;'.  It seems that
gcc disregards CV-qualification if the return value is a ?:
expression, but I haven't investigated the problem very much.

-- 
Alexandre Oliva
mailto:oliva@dcc.unicamp.br mailto:aoliva@acm.org
http://www.dcc.unicamp.br/~oliva
Universidade Estadual de Campinas, SP, Brasil

Re: Is this template problem in ecgs?

[Nicholas Lee]

On 24 Jun 1998, Alexandre Oliva wrote:

> Nicholas Lee <nlee011@math.auckland.ac.nz> writes:
> 
> > template<typename T>
> > inline T& min(const T& x, const T& y) { return (x<y)? x:y; }
> 
> Not a template problem, actually.  It would also compile if you
> replaced `template<typename T>' with `typedef int T;'.  It seems that
> gcc disregards CV-qualification if the return value is a ?:
> expression, but I haven't investigated the problem very much.

What is the correct behaviour?  

I would assume since x and y are declared as const T&, that the compiler
should issue a warning that there was a attempt to discard constness without
a const_cast.


--
	"I reserve the right to contradict myself."

Nicholas Lee (Li Peng Ming)  n.lee with math.auckland.ac in nz
Auckland University, New Zealand	

Re: Is this template problem in ecgs?

[Alexandre Oliva]

Nicholas Lee <nlee011@math.auckland.ac.nz> writes:

> On 24 Jun 1998, Alexandre Oliva wrote:

>> Nicholas Lee <nlee011@math.auckland.ac.nz> writes:
>> 
>> > template<typename T>
>> > inline T& min(const T& x, const T& y) { return (x<y)? x:y; }
>> 
>> It seems that gcc disregards CV-qualification if the return value
>> is a ?: expression

> I would assume since x and y are declared as const T&, that the compiler
> should issue a warning

You're right, I meant to say that, but I wasn't clear at all.  gcc and
egcs are wrong, they violate const-safety without diagnostic.

-- 
Alexandre Oliva
mailto:oliva@dcc.unicamp.br mailto:aoliva@acm.org
http://www.dcc.unicamp.br/~oliva
Universidade Estadual de Campinas, SP, Brasil

Re: Is this template problem in ecgs?

[Joe Buck]

> I would assume since x and y are declared as const T&, that the compiler
> should issue a warning that there was a attempt to discard constness without
> a const_cast.

const violations should, IMHO, be errors, not warnings.  They certainly
are with every other C++ compiler I've used.

Re: Is this template problem in ecgs?

[Nathan Myers]

Joe Buck wrote:
> 
> > I would assume since x and y are declared as const T&, that the compiler
> > should issue a warning that there was a attempt to discard constness without
> > a const_cast.
> 
> const violations should, IMHO, be errors, not warnings.  They certainly
> are with every other C++ compiler I've used.

Agree, at least for C++.

Nathan Myers
ncm@cantrip.org