...one of the most highly
regarded and expertly designed C++ library projects in the
world.
— Herb Sutter and Andrei
Alexandrescu, C++
Coding Standards
Home | Libraries | People | FAQ | More |
This page gives portability hints on some language features of the Borland C++ version 5.5.1 compiler. Furthermore, the appendix presents additional problems with Borland C++ version 5.5. Borland C++ 5.5.1 is a freely available command-line compiler for Win32 available at http://www.borland.com/.
Each entry in the following list describes a particular issue, complete with sample source code to demonstrate the effect. Most sample code herein has been verified to compile with gcc 2.95.2 and Comeau C++ 4.2.44.
__BORLANDC__
is defined for all
Borland C++ compilers. Its value is the version number of the
compiler interpreted as a hexadecimal number. The following table
lists some known values.
Compiler | __BORLANDC__ value |
---|---|
Borland C++ Builder 4 | 0x0540 |
Borland C++ Builder 5 | 0x0550 |
Borland C++ 5.5 | 0x0550 |
Borland C++ 5.5.1 | 0x0551 |
Borland C++ Builder 6 | 0x0560 |
using
-declarations and
using
-directivesusing
-directives (which refer to whole namespaces)
and namespace-level using
-declarations (which refer to
individual identifiers within foreign namespaces) causes ambiguities
where there are none. The following code fragment illustrates this:
namespace N { int x(); } using N::x; using namespace N; int main() { &x; // Ambiguous overload }
using
-declarations for class
templatesusing
-declarations as any other identifier. However, the
following code fails to compile with Borland C++:
template<class T> class X { }; namespace N { // "cannot use template 'X<T>' without specifying specialization parameters" using ::X; };
template<class T> void f(T x) { x = 1; // works (void) &x; T y = 17; y = 20; // "Cannot modify a const object in function f<const int>(int)" (void) &y; } int main() { const int i = 17; f(i); }The boost/rational.hpp header exhibits this problem in connection with the gcd() function.
template<class Arg> void f( void(*g)(Arg) ); void h(int); void h(double); template<class T> void h2(T); int main() { void (*p)(int) = h; // this works (std:13.4-1.1) void (*p2)(unsigned char) = h2; // this works as well (std:13.4-1.1) f<int>(h2); // this also works (std:13.4-1.3) // "Cannot generate template specialization from h(int)", // "Could not find a match for f<Arg>(void (*)(int))" f<double>(h); // should work (std:13.4-1.3) f( (void(*)(double))h); // C-style cast works (std:13.4-1.6 with 5.4) // "Overloaded 'h' ambiguous in this context" f(static_cast<void(*)(double)>(h)); // should work (std:13.4-1.6 with 5.2.9) }Workaround: Always use C-style casts when determining addresses of (potentially) overloaded functions.
const char *
to
std::string
const char *
parameters to
std::string
arguments fails if template functions are
explicitly instantiated (it works in the usual cases, though):
#include <string> template<class T> void f(const std::string & s) {} int main() { f<double>("hello"); // "Could not find a match for f<T>(char *)" }Workaround: Avoid explicit template function instantiations (they have significant problems with Microsoft Visual C++) and pass default-constructed unused dummy arguments with the appropriate type. Alternatively, if you wish to keep to the explicit instantiation, you could use an explicit conversion to
std::string
or declare the template function as taking a
const char *
parameter.
template<class T> struct A { static const bool value = true; }; // "Templates must be classes or functions", "Declaration syntax error" template<class T, bool v = A<T>::value> struct B {}; int main() { B<int> x; }Workaround: If the relevant non-type template parameter is an implementation detail, use inheritance and a fully qualified identifier (for example, ::N::A<T>::value).
#include <iostream> template<class T> struct A {}; template<class T1> void f(const A<T1> &) { std::cout << "f(const A<T1>&)\n"; } template<class T> void f(T) { std::cout << "f(T)\n"; } int main() { A<double> a; f(a); // output: f(T) (wrong) f(1); // output: f(T) (correct) }Workaround: Declare all such functions uniformly as either taking a value or a reference parameter.
template<class U> class C { }; template<class T> class A { static const int v = C<void (T::*)()>::value; };Workaround: Use an intermediate
typedef
:
template<class U> class C { }; template<class T> class A { typedef void (T::*my_type)(); static const int v = C<my_type>::value; };(Extracted from e-mail exchange of David Abrahams, Fernando Cacciola, and Peter Dimov; not actually tested.)
double std::abs(double)
missingdouble std::abs(double)
should be defined
(std:26.5-5 [lib.c.math]), but it is not:
#include <cmath> int main() { double (*p)(double) = std::abs; // error }Note that
int std::abs(int)
will be used without warning
if you write std::abs(5.1)
.
Similar remarks apply to seemingly all of the other standard math
functions, where Borland C++ fails to provide float
and
long double
overloads.
Workaround: Use std::fabs
instead if
type genericity is not required.
namespace N { template<class T> class A { // "f is not a member of 'N' in function main()" friend bool f(T x, T y) { return x < y; } }; } int main() { N::A<int> a; }This technique is extensively used in boost/operators.hpp. Giving in to the wish of the compiler doesn't work in this case, because then the "instantiate one template, get lots of helper functions at namespace scope" approach doesn't work anymore. Defining BOOST_NO_OPERATORS_IN_NAMESPACE (a define BOOST_NO_INLINE_FRIENDS_IN_CLASS_TEMPLATES would match this case better) works around this problem and leads to another one, see [using-template].