home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
DOS/V Power Report 1997 May
/
VPR9705A.ISO
/
VPR_DATA
/
PROGRAM
/
CBTRIAL
/
SETUP
/
DATA.Z
/
LIMITS.H
< prev
next >
Wrap
C/C++ Source or Header
|
1997-02-14
|
37KB
|
1,065 lines
/* limits.h
Defines implementation specific limits on type values.
*/
/*
* C/C++ Run Time Library - Version 8.0
*
* Copyright (c) 1987, 1997 by Borland International
* All Rights Reserved.
*
*/
#if !defined(__USING_STD_NAMES__)
#ifndef __LIMITS_H
#define __LIMITS_H
#if !defined(___DEFS_H)
#include <_defs.h>
#endif
#define CHAR_BIT 8 /* number of bits in a char */
#define MB_LEN_MAX 2 /* max. # bytes in multibyte char */
#define SCHAR_MIN (-128) /* minimum signed char value */
#define SCHAR_MAX 127 /* maximum signed char value */
#define UCHAR_MAX 255 /* maximum unsigned char value */
#if ('\x80' < 0)
#define CHAR_MIN SCHAR_MIN /* minimum char value */
#define CHAR_MAX SCHAR_MAX /* maximum char value */
#else
#define CHAR_MIN 0
#define CHAR_MAX UCHAR_MAX
#endif
#define SHRT_MIN (-32767-1) /* minimum signed short value */
#define SHRT_MAX 32767 /* maximum signed short value */
#define USHRT_MAX 65535U /* maximum unsigned short value */
#define LONG_MIN (-2147483647L-1) /* minimum signed long value */
#define LONG_MAX 2147483647L /* maximum signed long value */
#define ULONG_MAX 4294967295UL /* maximum unsigned long value */
#if defined(__FLAT__)
#define _I64_MIN (-9223372036854775807i64-1) /* minimum signed __int64 value */
#define _I64_MAX 9223372036854775807i64 /* maximum signed __int64 value */
#define _UI64_MAX 18446744073709551615ui64 /* maximum unsigned __int64 value */
#endif
#if !defined(__FLAT__)
#define INT_MIN SHRT_MIN /* minimum signed int value */
#define INT_MAX SHRT_MAX /* maximum signed int value */
#define UINT_MAX USHRT_MAX /* maximum unsigned int value */
#else
#define INT_MIN LONG_MIN
#define INT_MAX LONG_MAX
#define UINT_MAX ULONG_MAX
#endif
#if defined(__FLAT__) && defined(_POSIX_)
#define _POSIX_ARG_MAX 4096
#define _POSIX_CHILD_MAX 6
#define _POSIX_LINK_MAX 8
#define _POSIX_MAX_CANON 255
#define _POSIX_MAX_INPUT 255
#define _POSIX_NAME_MAX 14
#define _POSIX_NGROUPS_MAX 0
#define _POSIX_OPEN_MAX 16
#define _POSIX_PATH_MAX 255
#define _POSIX_PIPE_BUF 512
#define _POSIX_SSIZE_MAX 32767
#define _POSIX_STREAM_MAX 8
#define _POSIX_TZNAME_MAX 3
#define ARG_MAX (16384 - 256) /* 16k-(heap overhead+safety) */
#define CHILD_MAX 32
#define LINK_MAX _POSIX_LINK_MAX
#define MAX_CANON _POSIX_MAX_CANON
#define MAX_INPUT _POSIX_MAX_INPUT
#define NAME_MAX 255
#define NGROUPS_MAX 16
#define OPEN_MAX 32
#define PATH_MAX 512
#define PIPE_BUF _POSIX_PIPE_BUF
#define SSIZE_MAX _POSIX_SSIZE_MAX
#define STREAM_MAX 20
#define TZNAME_MAX 10
#endif /* POSIX */
#endif /* __LIMITS_H */
#else /* __USING_STD_NAMES__ */
#ifndef __STD_LIMITS
#define __STD_LIMITS
/***************************************************************************
*
* limits - Declarations for the Standard Library limits class
*
* $Id: limits,v 1.42 1995/09/29 01:06:31 smithey Exp $
*
***************************************************************************
*
* (c) Copyright 1994, 1995 Rogue Wave Software, Inc.
* ALL RIGHTS RESERVED
*
* The software and information contained herein are proprietary to, and
* comprise valuable trade secrets of, Rogue Wave Software, Inc., which
* intends to preserve as trade secrets such software and information.
* This software is furnished pursuant to a written license agreement and
* may be used, copied, transmitted, and stored only in accordance with
* the terms of such license and with the inclusion of the above copyright
* notice. This software and information or any other copies thereof may
* not be provided or otherwise made available to any other person.
*
* Notwithstanding any other lease or license that may pertain to, or
* accompany the delivery of, this computer software and information, the
* rights of the Government regarding its use, reproduction and disclosure
* are as set forth in Section 52.227-19 of the FARS Computer
* Software-Restricted Rights clause.
*
* Use, duplication, or disclosure by the Government is subject to
* restrictions as set forth in subparagraph (c)(1)(ii) of the Rights in
* Technical Data and Computer Software clause at DFARS 252.227-7013.
* Contractor/Manufacturer is Rogue Wave Software, Inc.,
* P.O. Box 2328, Corvallis, Oregon 97339.
*
* This computer software and information is distributed with "restricted
* rights." Use, duplication or disclosure is subject to restrictions as
* set forth in NASA FAR SUP 18-52.227-79 (April 1985) "Commercial
* Computer Software-Restricted Rights (April 1985)." If the Clause at
* 18-52.227-74 "Rights in Data General" is specified in the contract,
* then the "Alternate III" clause applies.
*
**************************************************************************/
#include <stdcomp.h>
#include <stddefs.h>
#ifndef RWSTD_NO_NEW_HEADER
#include <cfloat>
#include <climits>
#else
#include <float.h>
#include <limits.h>
#endif
//
// Hack to get around some <math.h>s defining an exception structure.
//
#define exception math_exception
#ifndef RWSTD_NO_NEW_HEADER
#include <cmath>
#else
#include <math.h>
#endif
//
// Undo the hack.
//
#undef exception
//
// WARNING: this is a hack to get around MSVC's complex
//
#ifdef _MSC_VER
#ifdef complex
#undef complex
#endif
#endif
#ifndef RWSTD_NO_NAMESPACE
namespace std {
#endif
enum float_round_style
{
round_indeterminable = -1,
round_toward_zero = 0,
round_to_nearest = 1,
round_toward_infinity = 2,
round_toward_neg_infinity = 3
};
template <class T>
class numeric_limits
{
public:
static const bool is_specialized;
static inline T min () { return (T) 0; }
static inline T max () { return (T) 0; }
static const int digits;
static const int digits10;
static const bool is_signed;
static const bool is_integer;
static const bool is_exact;
static const int radix;
static inline T epsilon () { return (T) 0; }
static inline T round_error () { return (T) 0; }
static const int min_exponent;
static const int min_exponent10;
static const int max_exponent;
static const int max_exponent10;
static const bool has_infinity;
static const bool has_quiet_NaN;
static const bool has_signaling_NaN;
static const bool has_denorm;
static inline T infinity () { return (T) 0; }
static inline T quiet_NaN () { return (T) 0; }
static inline T signaling_NaN () { return (T) 0; }
static inline T denorm_min () { return min(); }
static const bool is_iec559;
static const bool is_bounded;
static const bool is_modulo;
static const bool traps;
static const bool tinyness_before;
static const float_round_style round_style;
};
#ifndef RWSTD_NO_TEMPLATE_SPECIALIZATION
template <class T>
const bool numeric_limits<T>::is_specialized = false;
template <class T>
const int numeric_limits<T>::digits = 0;
template <class T>
const int numeric_limits<T>::digits10 = 0;
template <class T>
const bool numeric_limits<T>::is_signed = false;
template <class T>
const bool numeric_limits<T>::is_integer = false;
template <class T>
const bool numeric_limits<T>::is_exact = false;
template <class T>
const int numeric_limits<T>::radix = 0;
template <class T>
const int numeric_limits<T>::min_exponent10 =0;
template <class T>
const int numeric_limits<T>::max_exponent10 = 0;
template <class T>
const int numeric_limits<T>::min_exponent = 0;
template <class T>
const int numeric_limits<T>::max_exponent = 0;
template <class T>
const bool numeric_limits<T>::has_infinity = false;
template <class T>
const bool numeric_limits<T>::has_quiet_NaN = false;
template <class T>
const bool numeric_limits<T>::has_signaling_NaN = false;
template <class T>
const bool numeric_limits<T>::is_iec559 = false;
template <class T>
const bool numeric_limits<T>::is_bounded = false;
template <class T>
const bool numeric_limits<T>::is_modulo = false;
template <class T>
const bool numeric_limits<T>::has_denorm = false;
template <class T>
const bool numeric_limits<T>::traps = false;
template <class T>
const bool numeric_limits<T>::tinyness_before = false;
template <class T>
const float_round_style numeric_limits<T>::round_style = round_toward_zero;
#endif
#ifndef RWSTD_NO_STI_SIMPLE
#define __RW_INIT(n) = n
#else
#define __RW_INIT(n) /**/
#endif
//**********************************************************************
//
// Assumptions made in this implementation:
//
// 1) numeric_limits<int>::radix equals numeric_limits<T>::radix
// for all integral T specialized in this file.
//
// 2) numeric_limits<int>::is_modulo equals numeric_limits<T>::is_modulo
// for all signed integral T specialized in this file, except
// numeric_limits<bool>::is_modulo, which is assumed to be false.
//
// 3) numeric_limts<T>::traps == false for all builtin integral T.
//
// Does there exist a machine for which these aren't true?
//
//**********************************************************************
//**********************************************************************
//
// If your compiler allows in-class initialization of static const data
// members of integral type, then look for all lines having a comment of
// the form
//
// // VENDOR
//
// and set the value on that line to the proper one for your environment.
//
// If your compiler does NOT allow in-class initialization of static const
// data members of integral type, then you'll need to set the values in
// stdlib/src/limits/limits.cpp so they're properly archived into the
// Standard Library.
//
//**********************************************************************
//
// Specialization for float.
//
class RWSTDExport numeric_limits<float>
{
public:
static const bool is_specialized __RW_INIT(true);
static inline float min () { return FLT_MIN; }
static inline float max () { return FLT_MAX; }
static const int digits __RW_INIT(FLT_MANT_DIG);
static const int digits10 __RW_INIT(FLT_DIG);
static const bool is_signed __RW_INIT(true);
static const bool is_integer __RW_INIT(false);
static const bool is_exact __RW_INIT(false);
static const int radix __RW_INIT(FLT_RADIX);
static inline float epsilon () { return FLT_EPSILON; }
static float round_error ();
static const int min_exponent __RW_INIT(FLT_MIN_EXP);
static const int min_exponent10 __RW_INIT(FLT_MIN_10_EXP);
static const int max_exponent __RW_INIT(FLT_MAX_EXP);
static const int max_exponent10 __RW_INIT(FLT_MAX_10_EXP);
static const bool has_infinity __RW_INIT(false); // VENDOR
static const bool has_quiet_NaN __RW_INIT(false); // VENDOR
static const bool has_signaling_NaN __RW_INIT(false); // VENDOR
static const bool has_denorm __RW_INIT(false); // VENDOR
static float infinity ();
static float quiet_NaN ();
static float signaling_NaN ();
static float denorm_min ();
static const bool is_iec559 __RW_INIT(false); // VENDOR
static const bool is_bounded __RW_INIT(true);
static const bool is_modulo __RW_INIT(false); // VENDOR
static const bool traps __RW_INIT(false); // VENDOR
static const bool tinyness_before __RW_INIT(false); // VENDOR
static const float_round_style round_style __RW_INIT((float_round_style)FLT_ROUNDS);
};
//
// Specialization for double.
//
class RWSTDExport numeric_limits<double>
{
public:
static const bool is_specialized __RW_INIT(true);
static inline double min () { return DBL_MIN; }
static inline double max () { return DBL_MAX; }
static const int digits __RW_INIT(DBL_MANT_DIG);
static const int digits10 __RW_INIT(DBL_DIG);
static const bool is_signed __RW_INIT(true);
static const bool is_integer __RW_INIT(false);
static const bool is_exact __RW_INIT(false);
static const int radix __RW_INIT(FLT_RADIX);
static inline double epsilon () { return DBL_EPSILON; }
static double round_error ();
static const int min_exponent __RW_INIT(DBL_MIN_EXP);
static const int min_exponent10 __RW_INIT(DBL_MIN_10_EXP);
static const int max_exponent __RW_INIT(DBL_MAX_EXP);
static const int max_exponent10 __RW_INIT(DBL_MAX_10_EXP);
static const bool has_infinity __RW_INIT(false); // VENDOR
static const bool has_quiet_NaN __RW_INIT(false); // VENDOR
static const bool has_signaling_NaN __RW_INIT(false); // VENDOR
static const bool has_denorm __RW_INIT(false); // VENDOR
static double infinity ();
static double quiet_NaN ();
static double signaling_NaN ();
static double denorm_min ();
static const bool is_iec559 __RW_INIT(false); // VENDOR
static const bool is_bounded __RW_INIT(true);
static const bool is_modulo __RW_INIT(false); // VENDOR
static const bool traps __RW_INIT(false); // VENDOR
static const bool tinyness_before __RW_INIT(false); // VENDOR
static const float_round_style round_style __RW_INIT((float_round_style)FLT_ROUNDS);
};
//
// Specialization for long double.
//
class RWSTDExport numeric_limits<long double>
{
public:
static const bool is_specialized __RW_INIT(true);
static inline long double min () { return LDBL_MIN; }
static inline long double max () { return LDBL_MAX; }
static const int digits __RW_INIT(LDBL_MANT_DIG);
static const int digits10 __RW_INIT(LDBL_DIG);
static const bool is_signed __RW_INIT(true);
static const bool is_integer __RW_INIT(false);
static const bool is_exact __RW_INIT(false);
static const int radix __RW_INIT(FLT_RADIX);
static inline long double epsilon () { return LDBL_EPSILON; }
static long double round_error ();
static const int min_exponent __RW_INIT(LDBL_MIN_EXP);
static const int min_exponent10 __RW_INIT(LDBL_MIN_10_EXP);
static const int max_exponent __RW_INIT(LDBL_MAX_EXP);
static const int max_exponent10 __RW_INIT(LDBL_MAX_10_EXP);
static const bool has_infinity __RW_INIT(false); // VENDOR
static const bool has_quiet_NaN __RW_INIT(false); // VENDOR
static const bool has_signaling_NaN __RW_INIT(false); // VENDOR
static const bool has_denorm __RW_INIT(false); // VENDOR
static long double infinity ();
static long double quiet_NaN ();
static long double signaling_NaN ();
static long double denorm_min ();
static const bool is_iec559 __RW_INIT(false); // VENDOR
static const bool is_bounded __RW_INIT(true);
static const bool is_modulo __RW_INIT(false); // VENDOR
static const bool traps __RW_INIT(false); // VENDOR
static const bool tinyness_before __RW_INIT(false); // VENDOR
static const float_round_style round_style __RW_INIT((float_round_style)FLT_ROUNDS);
};
//
// Specialization for int.
//
class RWSTDExport numeric_limits<int>
{
public:
static const bool is_specialized __RW_INIT(true);
static inline int min () { return INT_MIN; }
static inline int max () { return INT_MAX; }
static const int digits __RW_INIT(CHAR_BIT*sizeof(int)-1);
static const int digits10 __RW_INIT(int(digits/3.321928095));
static const bool is_signed __RW_INIT(true);
static const bool is_integer __RW_INIT(true);
static const bool is_exact __RW_INIT(true);
static const int radix __RW_INIT(2); // VENDOR
static inline int epsilon () { return 0; }
static inline int round_error () { return 0; }
static const int min_exponent __RW_INIT(0);
static const int min_exponent10 __RW_INIT(0);
static const int max_exponent __RW_INIT(0);
static const int max_exponent10 __RW_INIT(0);
static const bool has_infinity __RW_INIT(false);
static const bool has_quiet_NaN __RW_INIT(false);
static const bool has_signaling_NaN __RW_INIT(false);
static const bool has_denorm __RW_INIT(false);
static inline int infinity () { return 0; }
static inline int quiet_NaN () { return 0; }
static inline int signaling_NaN () { return 0; }
static inline int denorm_min () { return min(); }
static const bool is_iec559 __RW_INIT(false);
static const bool is_bounded __RW_INIT(true);
static const bool is_modulo __RW_INIT(false); // VENDOR
static const bool traps __RW_INIT(false);
static const bool tinyness_before __RW_INIT(false);
static const float_round_style round_style __RW_INIT(round_toward_zero);
};
#ifndef RWSTD_NO_OVERLOAD_WCHAR
class RWSTDExport numeric_limits<wchar_t>
{
public:
static const bool is_specialized __RW_INIT(true);
static wchar_t min ();
static wchar_t max ();
static const bool is_signed __RW_INIT(true); // VENDOR
static const int digits __RW_INIT(is_signed ? CHAR_BIT*sizeof(wchar_t) - 1 : CHAR_BIT*sizeof(wchar_t));
static const int digits10 __RW_INIT(int(digits/3.321928095));
static const bool is_integer __RW_INIT(true);
static const bool is_exact __RW_INIT(true);
static const int radix __RW_INIT(numeric_limits<int>::radix);
static inline wchar_t epsilon () { return 0; }
static inline wchar_t round_error () { return 0; }
static const int min_exponent __RW_INIT(0);
static const int min_exponent10 __RW_INIT(0);
static const int max_exponent __RW_INIT(0);
static const int max_exponent10 __RW_INIT(0);
static const bool has_infinity __RW_INIT(false);
static const bool has_quiet_NaN __RW_INIT(false);
static const bool has_signaling_NaN __RW_INIT(false);
static const bool has_denorm __RW_INIT(false);
static inline wchar_t infinity () { return 0; }
static inline wchar_t quiet_NaN () { return 0; }
static inline wchar_t signaling_NaN () { return 0; }
static inline wchar_t denorm_min () { return min(); }
static const bool is_iec559 __RW_INIT(false);
static const bool is_bounded __RW_INIT(true);
static const bool is_modulo __RW_INIT(is_signed ? numeric_limits<int>::is_modulo : true);
static const bool traps __RW_INIT(false);
static const bool tinyness_before __RW_INIT(false);
static const float_round_style round_style __RW_INIT(round_toward_zero);
};
#endif /*!RWSTD_NO_OVERLOAD_WCHAR*/
//**********************************************************************
//
// There are no VENDOR-settable values beyond this point.
//
//**********************************************************************
//
// Specialization for short.
//
class RWSTDExport numeric_limits<short>
{
public:
static const bool is_specialized __RW_INIT(true);
static inline short min () { return SHRT_MIN; }
static inline short max () { return SHRT_MAX; }
static const int digits __RW_INIT(CHAR_BIT*sizeof(short)-1);
static const int digits10 __RW_INIT(int(digits/3.321928095));
static const bool is_signed __RW_INIT(true);
static const bool is_integer __RW_INIT(true);
static const bool is_exact __RW_INIT(true);
static const int radix __RW_INIT(numeric_limits<int>::radix);
static inline short epsilon () { return 0; }
static inline short round_error () { return 0; }
static const int min_exponent __RW_INIT(0);
static const int min_exponent10 __RW_INIT(0);
static const int max_exponent __RW_INIT(0);
static const int max_exponent10 __RW_INIT(0);
static const bool has_infinity __RW_INIT(false);
static const bool has_quiet_NaN __RW_INIT(false);
static const bool has_signaling_NaN __RW_INIT(false);
static const bool has_denorm __RW_INIT(false);
static inline short infinity () { return 0; }
static inline short quiet_NaN () { return 0; }
static inline short signaling_NaN () { return 0; }
static inline short denorm_min () { return min(); }
static const bool is_iec559 __RW_INIT(false);
static const bool is_bounded __RW_INIT(true);
static const bool is_modulo __RW_INIT(numeric_limits<int>::is_modulo);
static const bool traps __RW_INIT(false);
static const bool tinyness_before __RW_INIT(false);
static const float_round_style round_style __RW_INIT(round_toward_zero);
};
//
// Specialization for unsigned short.
//
class RWSTDExport numeric_limits<unsigned short>
{
public:
static const bool is_specialized __RW_INIT(true);
static inline unsigned short min () { return 0; }
static inline unsigned short max () { return USHRT_MAX; }
static const int digits __RW_INIT(CHAR_BIT*sizeof(unsigned short));
static const int digits10 __RW_INIT(int(digits/3.321928095));
static const bool is_signed __RW_INIT(false);
static const bool is_integer __RW_INIT(true);
static const bool is_exact __RW_INIT(true);
static const int radix __RW_INIT(numeric_limits<int>::radix);
static inline unsigned short epsilon () { return 0; }
static inline unsigned short round_error () { return 0; }
static const int min_exponent __RW_INIT(0);
static const int min_exponent10 __RW_INIT(0);
static const int max_exponent __RW_INIT(0);
static const int max_exponent10 __RW_INIT(0);
static const bool has_infinity __RW_INIT(false);
static const bool has_quiet_NaN __RW_INIT(false);
static const bool has_signaling_NaN __RW_INIT(false);
static const bool has_denorm __RW_INIT(false);
static inline unsigned short infinity () { return 0; }
static inline unsigned short quiet_NaN () { return 0; }
static inline unsigned short signaling_NaN () { return 0; }
static inline unsigned short denorm_min () { return min(); }
static const bool is_iec559 __RW_INIT(false);
static const bool is_bounded __RW_INIT(true);
static const bool is_modulo __RW_INIT(true);
static const bool traps __RW_INIT(false);
static const bool tinyness_before __RW_INIT(false);
static const float_round_style round_style __RW_INIT(round_toward_zero);
};
//
// Specialization for unsigned int.
//
class RWSTDExport numeric_limits<unsigned int>
{
public:
static const bool is_specialized __RW_INIT(true);
static inline unsigned int min () { return 0; }
static inline unsigned int max () { return UINT_MAX; }
static const int digits __RW_INIT(CHAR_BIT*sizeof(unsigned int));
static const int digits10 __RW_INIT(int(digits/3.321928095));
static const bool is_signed __RW_INIT(false);
static const bool is_integer __RW_INIT(true);
static const bool is_exact __RW_INIT(true);
static const int radix __RW_INIT(numeric_limits<int>::radix);
static inline unsigned int epsilon () { return 0; }
static inline unsigned int round_error () { return 0; }
static const int min_exponent __RW_INIT(0);
static const int min_exponent10 __RW_INIT(0);
static const int max_exponent __RW_INIT(0);
static const int max_exponent10 __RW_INIT(0);
static const bool has_infinity __RW_INIT(false);
static const bool has_quiet_NaN __RW_INIT(false);
static const bool has_signaling_NaN __RW_INIT(false);
static const bool has_denorm __RW_INIT(false);
static inline unsigned int infinity () { return 0; }
static inline unsigned int quiet_NaN () { return 0; }
static inline unsigned int signaling_NaN () { return 0; }
static inline unsigned int denorm_min () { return min(); }
static const bool is_iec559 __RW_INIT(false);
static const bool is_bounded __RW_INIT(true);
static const bool is_modulo __RW_INIT(true);
static const bool traps __RW_INIT(false);
static const bool tinyness_before __RW_INIT(false);
static const float_round_style round_style __RW_INIT(round_toward_zero);
};
//
// Specialization for long.
//
class RWSTDExport numeric_limits<long>
{
public:
static const bool is_specialized __RW_INIT(true);
static inline long min () { return LONG_MIN; }
static inline long max () { return LONG_MAX; }
static const int digits __RW_INIT(CHAR_BIT*sizeof(long)-1);
static const int digits10 __RW_INIT(int(digits/3.321928095));
static const bool is_signed __RW_INIT(true);
static const bool is_integer __RW_INIT(true);
static const bool is_exact __RW_INIT(true);
static const int radix __RW_INIT(numeric_limits<int>::radix);
static inline long epsilon () { return 0; }
static inline long round_error () { return 0; }
static const int min_exponent __RW_INIT(0);
static const int min_exponent10 __RW_INIT(0);
static const int max_exponent __RW_INIT(0);
static const int max_exponent10 __RW_INIT(0);
static const bool has_infinity __RW_INIT(false);
static const bool has_quiet_NaN __RW_INIT(false);
static const bool has_signaling_NaN __RW_INIT(false);
static const bool has_denorm __RW_INIT(false);
static inline long infinity () { return 0; }
static inline long quiet_NaN () { return 0; }
static inline long signaling_NaN () { return 0; }
static inline long denorm_min () { return min(); }
static const bool is_iec559 __RW_INIT(false);
static const bool is_bounded __RW_INIT(true);
static const bool is_modulo __RW_INIT(numeric_limits<int>::is_modulo);
static const bool traps __RW_INIT(false);
static const bool tinyness_before __RW_INIT(false);
static const float_round_style round_style __RW_INIT(round_toward_zero);
};
//
// Specialization for unsigned long.
//
class RWSTDExport numeric_limits<unsigned long>
{
public:
static const bool is_specialized __RW_INIT(true);
static inline unsigned long min () { return 0; }
static inline unsigned long max () { return ULONG_MAX; }
static const int digits __RW_INIT(CHAR_BIT*sizeof(unsigned long));
static const int digits10 __RW_INIT(int(digits/3.321928095));
static const bool is_signed __RW_INIT(false);
static const bool is_integer __RW_INIT(true);
static const bool is_exact __RW_INIT(true);
static const int radix __RW_INIT(numeric_limits<int>::radix);
static inline unsigned long epsilon () { return 0; }
static inline unsigned long round_error () { return 0; }
static const int min_exponent __RW_INIT(0);
static const int min_exponent10 __RW_INIT(0);
static const int max_exponent __RW_INIT(0);
static const int max_exponent10 __RW_INIT(0);
static const bool has_infinity __RW_INIT(false);
static const bool has_quiet_NaN __RW_INIT(false);
static const bool has_signaling_NaN __RW_INIT(false);
static const bool has_denorm __RW_INIT(false);
static inline unsigned long infinity () { return 0; }
static inline unsigned long quiet_NaN () { return 0; }
static inline unsigned long signaling_NaN () { return 0; }
static inline unsigned long denorm_min () { return min(); }
static const bool is_iec559 __RW_INIT(false);
static const bool is_bounded __RW_INIT(true);
static const bool is_modulo __RW_INIT(true);
static const bool traps __RW_INIT(false);
static const bool tinyness_before __RW_INIT(false);
static const float_round_style round_style __RW_INIT(round_toward_zero);
};
//
// Specialization for char.
//
class RWSTDExport numeric_limits<char>
{
public:
static const bool is_specialized __RW_INIT(true);
static inline char min () { return CHAR_MIN; }
static inline char max () { return CHAR_MAX; }
static const bool is_signed __RW_INIT(CHAR_MAX == SCHAR_MAX ? true : false);
static const int digits __RW_INIT(is_signed ? CHAR_BIT*sizeof(char) -1 : CHAR_BIT*sizeof(char));
static const int digits10 __RW_INIT(int(digits/3.321928095));
static const bool is_integer __RW_INIT(true);
static const bool is_exact __RW_INIT(true);
static const int radix __RW_INIT(numeric_limits<int>::radix);
static inline char epsilon () { return 0; }
static inline char round_error () { return 0; }
static const int min_exponent __RW_INIT(0);
static const int min_exponent10 __RW_INIT(0);
static const int max_exponent __RW_INIT(0);
static const int max_exponent10 __RW_INIT(0);
static const bool has_infinity __RW_INIT(false);
static const bool has_quiet_NaN __RW_INIT(false);
static const bool has_signaling_NaN __RW_INIT(false);
static const bool has_denorm __RW_INIT(false);
static inline char infinity () { return 0; }
static inline char quiet_NaN () { return 0; }
static inline char signaling_NaN () { return 0; }
static inline char denorm_min () { return min(); }
static const bool is_iec559 __RW_INIT(false);
static const bool is_bounded __RW_INIT(true);
static const bool is_modulo __RW_INIT(is_signed ? numeric_limits<int>::is_modulo : true);
static const bool traps __RW_INIT(false);
static const bool tinyness_before __RW_INIT(false);
static const float_round_style round_style __RW_INIT(round_toward_zero);
};
//
// Specialization for unsigned char.
//
class RWSTDExport numeric_limits<unsigned char>
{
public:
static const bool is_specialized __RW_INIT(true);
static inline unsigned char min () { return 0; }
static inline unsigned char max () { return UCHAR_MAX; }
static const int digits __RW_INIT(CHAR_BIT*sizeof(unsigned char));
static const int digits10 __RW_INIT(int(digits/3.321928095));
static const bool is_signed __RW_INIT(false);
static const bool is_integer __RW_INIT(true);
static const bool is_exact __RW_INIT(true);
static const int radix __RW_INIT(numeric_limits<int>::radix);
static inline unsigned char epsilon () { return 0; }
static inline unsigned char round_error () { return 0; }
static const int min_exponent __RW_INIT(0);
static const int min_exponent10 __RW_INIT(0);
static const int max_exponent __RW_INIT(0);
static const int max_exponent10 __RW_INIT(0);
static const bool has_infinity __RW_INIT(false);
static const bool has_quiet_NaN __RW_INIT(false);
static const bool has_signaling_NaN __RW_INIT(false);
static const bool has_denorm __RW_INIT(false);
static inline unsigned char infinity () { return 0; }
static inline unsigned char quiet_NaN () { return 0; }
static inline unsigned char signaling_NaN () { return 0; }
static inline unsigned char denorm_min () { return min(); }
static const bool is_iec559 __RW_INIT(false);
static const bool is_bounded __RW_INIT(true);
static const bool is_modulo __RW_INIT(true);
static const bool traps __RW_INIT(false);
static const bool tinyness_before __RW_INIT(false);
static const float_round_style round_style __RW_INIT(round_toward_zero);
};
//
// Specialization for signed char.
//
class RWSTDExport numeric_limits<signed char>
{
public:
static const bool is_specialized __RW_INIT(true);
static inline signed char min () { return SCHAR_MIN; }
static inline signed char max () { return SCHAR_MAX; }
static const int digits __RW_INIT(CHAR_BIT*sizeof(signed char)-1);
static const int digits10 __RW_INIT(int(digits/3.321928095));
static const bool is_signed __RW_INIT(true);
static const bool is_integer __RW_INIT(true);
static const bool is_exact __RW_INIT(true);
static const int radix __RW_INIT(numeric_limits<int>::radix);
static inline signed char epsilon () { return 0; }
static inline signed char round_error () { return 0; }
static const int min_exponent __RW_INIT(0);
static const int min_exponent10 __RW_INIT(0);
static const int max_exponent __RW_INIT(0);
static const int max_exponent10 __RW_INIT(0);
static const bool has_infinity __RW_INIT(false);
static const bool has_quiet_NaN __RW_INIT(false);
static const bool has_signaling_NaN __RW_INIT(false);
static const bool has_denorm __RW_INIT(false);
static inline signed char infinity () { return 0; }
static inline signed char quiet_NaN () { return 0; }
static inline signed char signaling_NaN () { return 0; }
static inline signed char denorm_min () { return min(); }
static const bool is_iec559 __RW_INIT(false);
static const bool is_bounded __RW_INIT(true);
static const bool is_modulo __RW_INIT(numeric_limits<int>::is_modulo);
static const bool traps __RW_INIT(false);
static const bool tinyness_before __RW_INIT(false);
static const float_round_style round_style __RW_INIT(round_toward_zero);
};
#ifndef RWSTD_NO_BOOL
//
// Specialization for bool.
//
class RWSTDExport numeric_limits<bool>
{
public:
static const bool is_specialized __RW_INIT(true);
static inline bool min () { return false; }
static inline bool max () { return true; }
static const int digits __RW_INIT(1);
static const int digits10 __RW_INIT(0);
static const bool is_signed __RW_INIT(false);
static const bool is_integer __RW_INIT(true);
static const bool is_exact __RW_INIT(true);
static const int radix __RW_INIT(numeric_limits<int>::radix);
static inline bool epsilon () { return 0; }
static inline bool round_error () { return 0; }
static const int min_exponent __RW_INIT(0);
static const int min_exponent10 __RW_INIT(0);
static const int max_exponent __RW_INIT(0);
static const int max_exponent10 __RW_INIT(0);
static const bool has_infinity __RW_INIT(false);
static const bool has_quiet_NaN __RW_INIT(false);
static const bool has_signaling_NaN __RW_INIT(false);
static const bool has_denorm __RW_INIT(false);
static inline int infinity () { return 0; }
static inline int quiet_NaN () { return 0; }
static inline int signaling_NaN () { return 0; }
static inline int denorm_min () { return min(); }
static const bool is_iec559 __RW_INIT(false);
static const bool is_bounded __RW_INIT(true);
static const bool is_modulo __RW_INIT(false);
static const bool traps __RW_INIT(false);
static const bool tinyness_before __RW_INIT(false);
static const float_round_style round_style __RW_INIT(round_toward_zero);
};
#endif /*!RWSTD_NO_BOOL*/
#undef __RW_INIT
#ifndef RWSTD_NO_NAMESPACE
}
#endif
#endif /*__STD_LIMITS*/
#endif /* __USING_STD_NAMES__ */