/* A GNU-like <math.h>.
- Copyright (C) 2002-2003, 2007-2009 Free Software Foundation, Inc.
+ Copyright (C) 2002-2003, 2007-2010 Free Software Foundation, Inc.
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
/* The definition of GL_LINK_WARNING is copied here. */
+/* The definition of _GL_ARG_NONNULL is copied here. */
+
#ifdef __cplusplus
extern "C" {
#if @GNULIB_FREXP@
# if @REPLACE_FREXP@
# define frexp rpl_frexp
-extern double frexp (double x, int *expptr);
+extern double frexp (double x, int *expptr) _GL_ARG_NONNULL ((2));
# endif
#elif defined GNULIB_POSIXCHECK
# undef frexp
# define frexpl rpl_frexpl
#endif
#if (@GNULIB_FREXPL@ && @REPLACE_FREXPL@) || !@HAVE_DECL_FREXPL@
-extern long double frexpl (long double x, int *expptr);
+extern long double frexpl (long double x, int *expptr) _GL_ARG_NONNULL ((2));
#endif
#if !@GNULIB_FREXPL@ && defined GNULIB_POSIXCHECK
# undef frexpl
# if @HAVE_ISNANF@
/* The original <math.h> included above provides a declaration of isnan macro
or (older) isnanf function. */
-# include <math.h>
# if __GNUC__ >= 4
/* GCC 4.0 and newer provides three built-ins for isnan. */
# undef isnanf
to 'double' numbers earlier but now is a type-generic macro. */
# if @HAVE_ISNAND@
/* The original <math.h> included above provides a declaration of isnan macro. */
-# include <math.h>
# if __GNUC__ >= 4
/* GCC 4.0 and newer provides three built-ins for isnan. */
# undef isnand
/* Test for NaN for 'long double' numbers. */
# if @HAVE_ISNANL@
/* The original <math.h> included above provides a declaration of isnan macro or (older) isnanl function. */
-# include <math.h>
# if __GNUC__ >= 4
/* GCC 4.0 and newer provides three built-ins for isnan. */
# undef isnanl
# if defined FLT_SIGNBIT_WORD && defined FLT_SIGNBIT_BIT && !defined gl_signbitf
# define gl_signbitf_OPTIMIZED_MACRO
# define gl_signbitf(arg) \
- ({ union { float _value; \
+ ({ union { float _value; \
unsigned int _word[(sizeof (float) + sizeof (unsigned int) - 1) / sizeof (unsigned int)]; \
- } _m; \
- _m._value = (arg); \
- (_m._word[FLT_SIGNBIT_WORD] >> FLT_SIGNBIT_BIT) & 1; \
+ } _m; \
+ _m._value = (arg); \
+ (_m._word[FLT_SIGNBIT_WORD] >> FLT_SIGNBIT_BIT) & 1; \
})
# endif
# if defined DBL_SIGNBIT_WORD && defined DBL_SIGNBIT_BIT && !defined gl_signbitd
# define gl_signbitd_OPTIMIZED_MACRO
# define gl_signbitd(arg) \
- ({ union { double _value; \
+ ({ union { double _value; \
unsigned int _word[(sizeof (double) + sizeof (unsigned int) - 1) / sizeof (unsigned int)]; \
- } _m; \
- _m._value = (arg); \
- (_m._word[DBL_SIGNBIT_WORD] >> DBL_SIGNBIT_BIT) & 1; \
+ } _m; \
+ _m._value = (arg); \
+ (_m._word[DBL_SIGNBIT_WORD] >> DBL_SIGNBIT_BIT) & 1; \
})
# endif
# if defined LDBL_SIGNBIT_WORD && defined LDBL_SIGNBIT_BIT && !defined gl_signbitl
# define gl_signbitl_OPTIMIZED_MACRO
# define gl_signbitl(arg) \
- ({ union { long double _value; \
+ ({ union { long double _value; \
unsigned int _word[(sizeof (long double) + sizeof (unsigned int) - 1) / sizeof (unsigned int)]; \
- } _m; \
- _m._value = (arg); \
- (_m._word[LDBL_SIGNBIT_WORD] >> LDBL_SIGNBIT_BIT) & 1; \
+ } _m; \
+ _m._value = (arg); \
+ (_m._word[LDBL_SIGNBIT_WORD] >> LDBL_SIGNBIT_BIT) & 1; \
})
# endif
# endif