]> git.deb.at Git - pkg/abook.git/blobdiff - intl/vasnprintf.c
Merge remote-tracking branch 'upstream/master' into upstream
[pkg/abook.git] / intl / vasnprintf.c
index 8a07ca6e72da1331e3bf1a741aaa6efbf80085ac..ad0a0676d26b9b8a6426c7dcec992c78568fcb54 100644 (file)
@@ -1,5 +1,5 @@
 /* vsprintf with automatic memory allocation.
-   Copyright (C) 1999, 2002-2010 Free Software Foundation, Inc.
+   Copyright (C) 1999, 2002-2005 Free Software Foundation, Inc.
 
    This program is free software; you can redistribute it and/or modify it
    under the terms of the GNU Library General Public License as published
 
    You should have received a copy of the GNU Library General Public
    License along with this program; if not, write to the Free Software
-   Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301,
+   Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307,
    USA.  */
 
-/* This file can be parametrized with the following macros:
-     VASNPRINTF         The name of the function being defined.
-     FCHAR_T            The element type of the format string.
-     DCHAR_T            The element type of the destination (result) string.
-     FCHAR_T_ONLY_ASCII Set to 1 to enable verification that all characters
-                        in the format string are ASCII. MUST be set if
-                        FCHAR_T and DCHAR_T are not the same type.
-     DIRECTIVE          Structure denoting a format directive.
-                        Depends on FCHAR_T.
-     DIRECTIVES         Structure denoting the set of format directives of a
-                        format string.  Depends on FCHAR_T.
-     PRINTF_PARSE       Function that parses a format string.
-                        Depends on FCHAR_T.
-     DCHAR_CPY          memcpy like function for DCHAR_T[] arrays.
-     DCHAR_SET          memset like function for DCHAR_T[] arrays.
-     DCHAR_MBSNLEN      mbsnlen like function for DCHAR_T[] arrays.
-     SNPRINTF           The system's snprintf (or similar) function.
-                        This may be either snprintf or swprintf.
-     TCHAR_T            The element type of the argument and result string
-                        of the said SNPRINTF function.  This may be either
-                        char or wchar_t.  The code exploits that
-                        sizeof (TCHAR_T) | sizeof (DCHAR_T) and
-                        alignof (TCHAR_T) <= alignof (DCHAR_T).
-     DCHAR_IS_TCHAR     Set to 1 if DCHAR_T and TCHAR_T are the same type.
-     DCHAR_CONV_FROM_ENCODING A function to convert from char[] to DCHAR[].
-     DCHAR_IS_UINT8_T   Set to 1 if DCHAR_T is uint8_t.
-     DCHAR_IS_UINT16_T  Set to 1 if DCHAR_T is uint16_t.
-     DCHAR_IS_UINT32_T  Set to 1 if DCHAR_T is uint32_t.  */
-
 /* Tell glibc's <stdio.h> to provide a prototype for snprintf().
    This must come before <config.h> because <config.h> may include
    <features.h>, and once <features.h> has been included, it's too late.  */
@@ -52,7 +23,7 @@
 # define _GNU_SOURCE    1
 #endif
 
-#ifndef VASNPRINTF
+#ifdef HAVE_CONFIG_H
 # include <config.h>
 #endif
 #ifndef IN_LIBINTL
 #endif
 
 /* Specification.  */
-#ifndef VASNPRINTF
-# if WIDE_CHAR_VERSION
-#  include "vasnwprintf.h"
-# else
-#  include "vasnprintf.h"
-# endif
-#endif
-
-#include <locale.h>     /* localeconv() */
-#include <stdio.h>      /* snprintf(), sprintf() */
-#include <stdlib.h>     /* abort(), malloc(), realloc(), free() */
-#include <string.h>     /* memcpy(), strlen() */
-#include <errno.h>      /* errno */
-#include <limits.h>     /* CHAR_BIT */
-#include <float.h>      /* DBL_MAX_EXP, LDBL_MAX_EXP */
-#if HAVE_NL_LANGINFO
-# include <langinfo.h>
-#endif
-#ifndef VASNPRINTF
-# if WIDE_CHAR_VERSION
-#  include "wprintf-parse.h"
-# else
-#  include "printf-parse.h"
-# endif
-#endif
-
-/* Checked size_t computations.  */
-#include "xsize.h"
-
-#if (NEED_PRINTF_DOUBLE || NEED_PRINTF_LONG_DOUBLE) && !defined IN_LIBINTL
-# include <math.h>
-# include "float+.h"
-#endif
-
-#if (NEED_PRINTF_DOUBLE || NEED_PRINTF_INFINITE_DOUBLE) && !defined IN_LIBINTL
-# include <math.h>
-# include "isnand-nolibm.h"
-#endif
-
-#if (NEED_PRINTF_LONG_DOUBLE || NEED_PRINTF_INFINITE_LONG_DOUBLE) && !defined IN_LIBINTL
-# include <math.h>
-# include "isnanl-nolibm.h"
-# include "fpucw.h"
-#endif
-
-#if (NEED_PRINTF_DIRECTIVE_A || NEED_PRINTF_DOUBLE) && !defined IN_LIBINTL
-# include <math.h>
-# include "isnand-nolibm.h"
-# include "printf-frexp.h"
-#endif
-
-#if (NEED_PRINTF_DIRECTIVE_A || NEED_PRINTF_LONG_DOUBLE) && !defined IN_LIBINTL
-# include <math.h>
-# include "isnanl-nolibm.h"
-# include "printf-frexpl.h"
-# include "fpucw.h"
-#endif
-
-/* Default parameters.  */
-#ifndef VASNPRINTF
-# if WIDE_CHAR_VERSION
-#  define VASNPRINTF vasnwprintf
-#  define FCHAR_T wchar_t
-#  define DCHAR_T wchar_t
-#  define TCHAR_T wchar_t
-#  define DCHAR_IS_TCHAR 1
-#  define DIRECTIVE wchar_t_directive
-#  define DIRECTIVES wchar_t_directives
-#  define PRINTF_PARSE wprintf_parse
-#  define DCHAR_CPY wmemcpy
-#  define DCHAR_SET wmemset
-# else
-#  define VASNPRINTF vasnprintf
-#  define FCHAR_T char
-#  define DCHAR_T char
-#  define TCHAR_T char
-#  define DCHAR_IS_TCHAR 1
-#  define DIRECTIVE char_directive
-#  define DIRECTIVES char_directives
-#  define PRINTF_PARSE printf_parse
-#  define DCHAR_CPY memcpy
-#  define DCHAR_SET memset
-# endif
-#endif
 #if WIDE_CHAR_VERSION
-  /* TCHAR_T is wchar_t.  */
-# define USE_SNPRINTF 1
-# if HAVE_DECL__SNWPRINTF
-   /* On Windows, the function swprintf() has a different signature than
-      on Unix; we use the function _snwprintf() or - on mingw - snwprintf()
-      instead.  The mingw function snwprintf() has fewer bugs than the
-      MSVCRT function _snwprintf(), so prefer that.  */
-#  if defined __MINGW32__
-#   define SNPRINTF snwprintf
-#  else
-#   define SNPRINTF _snwprintf
-#  endif
-# else
-   /* Unix.  */
-#  define SNPRINTF swprintf
-# endif
+# include "vasnwprintf.h"
 #else
-  /* TCHAR_T is char.  */
-  /* Use snprintf if it exists under the name 'snprintf' or '_snprintf'.
-     But don't use it on BeOS, since BeOS snprintf produces no output if the
-     size argument is >= 0x3000000.
-     Also don't use it on Linux libc5, since there snprintf with size = 1
-     writes any output without bounds, like sprintf.  */
-# if (HAVE_DECL__SNPRINTF || HAVE_SNPRINTF) && !defined __BEOS__ && !(__GNU_LIBRARY__ == 1)
-#  define USE_SNPRINTF 1
-# else
-#  define USE_SNPRINTF 0
-# endif
-# if HAVE_DECL__SNPRINTF
-   /* Windows.  The mingw function snprintf() has fewer bugs than the MSVCRT
-      function _snprintf(), so prefer that.  */
-#  if defined __MINGW32__
-#   define SNPRINTF snprintf
-    /* Here we need to call the native snprintf, not rpl_snprintf.  */
-#   undef snprintf
-#  else
-#   define SNPRINTF _snprintf
-#  endif
-# else
-   /* Unix.  */
-#  define SNPRINTF snprintf
-   /* Here we need to call the native snprintf, not rpl_snprintf.  */
-#  undef snprintf
-# endif
+# include "vasnprintf.h"
 #endif
-/* Here we need to call the native sprintf, not rpl_sprintf.  */
-#undef sprintf
 
-/* GCC >= 4.0 with -Wall emits unjustified "... may be used uninitialized"
-   warnings in this file.  Use -Dlint to suppress them.  */
-#ifdef lint
-# define IF_LINT(Code) Code
+#include <stdio.h>     /* snprintf(), sprintf() */
+#include <stdlib.h>    /* abort(), malloc(), realloc(), free() */
+#include <string.h>    /* memcpy(), strlen() */
+#include <errno.h>     /* errno */
+#include <limits.h>    /* CHAR_BIT, INT_MAX */
+#include <float.h>     /* DBL_MAX_EXP, LDBL_MAX_EXP */
+#if WIDE_CHAR_VERSION
+# include "wprintf-parse.h"
 #else
-# define IF_LINT(Code) /* empty */
+# include "printf-parse.h"
 #endif
 
-/* Avoid some warnings from "gcc -Wshadow".
-   This file doesn't use the exp() and remainder() functions.  */
-#undef exp
-#define exp expo
-#undef remainder
-#define remainder rem
+/* Checked size_t computations.  */
+#include "xsize.h"
 
-#if (!USE_SNPRINTF || !HAVE_SNPRINTF_RETVAL_C99) && !WIDE_CHAR_VERSION
-# if (HAVE_STRNLEN && !defined _AIX)
-#  define local_strnlen strnlen
-# else
-#  ifndef local_strnlen_defined
-#   define local_strnlen_defined 1
-static size_t
-local_strnlen (const char *string, size_t maxlen)
-{
-  const char *end = memchr (string, '\0', maxlen);
-  return end ? (size_t) (end - string) : maxlen;
-}
-#  endif
-# endif
+/* Some systems, like OSF/1 4.0 and Woe32, don't have EOVERFLOW.  */
+#ifndef EOVERFLOW
+# define EOVERFLOW E2BIG
 #endif
 
-#if (((!USE_SNPRINTF || !HAVE_SNPRINTF_RETVAL_C99) && WIDE_CHAR_VERSION) || ((!USE_SNPRINTF || !HAVE_SNPRINTF_RETVAL_C99 || (NEED_PRINTF_DIRECTIVE_LS && !defined IN_LIBINTL)) && !WIDE_CHAR_VERSION && DCHAR_IS_TCHAR)) && HAVE_WCHAR_T
-# if HAVE_WCSLEN
+#ifdef HAVE_WCHAR_T
+# ifdef HAVE_WCSLEN
 #  define local_wcslen wcslen
 # else
    /* Solaris 2.5.1 has wcslen() in a separate library libw.so. To avoid
@@ -247,5298 +80,806 @@ local_wcslen (const wchar_t *s)
 # endif
 #endif
 
-#if (!USE_SNPRINTF || !HAVE_SNPRINTF_RETVAL_C99) && HAVE_WCHAR_T && WIDE_CHAR_VERSION
-# if HAVE_WCSNLEN
-#  define local_wcsnlen wcsnlen
+#if WIDE_CHAR_VERSION
+# define VASNPRINTF vasnwprintf
+# define CHAR_T wchar_t
+# define DIRECTIVE wchar_t_directive
+# define DIRECTIVES wchar_t_directives
+# define PRINTF_PARSE wprintf_parse
+# define USE_SNPRINTF 1
+# if HAVE_DECL__SNWPRINTF
+   /* On Windows, the function swprintf() has a different signature than
+      on Unix; we use the _snwprintf() function instead.  */
+#  define SNPRINTF _snwprintf
 # else
-#  ifndef local_wcsnlen_defined
-#   define local_wcsnlen_defined 1
-static size_t
-local_wcsnlen (const wchar_t *s, size_t maxlen)
-{
-  const wchar_t *ptr;
-
-  for (ptr = s; maxlen > 0 && *ptr != (wchar_t) 0; ptr++, maxlen--)
-    ;
-  return ptr - s;
-}
-#  endif
+   /* Unix.  */
+#  define SNPRINTF swprintf
 # endif
-#endif
-
-#if (NEED_PRINTF_DIRECTIVE_A || NEED_PRINTF_LONG_DOUBLE || NEED_PRINTF_INFINITE_LONG_DOUBLE || NEED_PRINTF_DOUBLE || NEED_PRINTF_INFINITE_DOUBLE) && !defined IN_LIBINTL
-/* Determine the decimal-point character according to the current locale.  */
-# ifndef decimal_point_char_defined
-#  define decimal_point_char_defined 1
-static char
-decimal_point_char (void)
-{
-  const char *point;
-  /* Determine it in a multithread-safe way.  We know nl_langinfo is
-     multithread-safe on glibc systems and MacOS X systems, but is not required
-     to be multithread-safe by POSIX.  sprintf(), however, is multithread-safe.
-     localeconv() is rarely multithread-safe.  */
-#  if HAVE_NL_LANGINFO && (__GLIBC__ || (defined __APPLE__ && defined __MACH__))
-  point = nl_langinfo (RADIXCHAR);
-#  elif 1
-  char pointbuf[5];
-  sprintf (pointbuf, "%#.0f", 1.0);
-  point = &pointbuf[1];
-#  else
-  point = localeconv () -> decimal_point;
-#  endif
-  /* The decimal point is always a single byte: either '.' or ','.  */
-  return (point[0] != '\0' ? point[0] : '.');
-}
+#else
+# define VASNPRINTF vasnprintf
+# define CHAR_T char
+# define DIRECTIVE char_directive
+# define DIRECTIVES char_directives
+# define PRINTF_PARSE printf_parse
+# define USE_SNPRINTF (HAVE_DECL__SNPRINTF || HAVE_SNPRINTF)
+# if HAVE_DECL__SNPRINTF
+   /* Windows.  */
+#  define SNPRINTF _snprintf
+# else
+   /* Unix.  */
+#  define SNPRINTF snprintf
 # endif
 #endif
 
-#if NEED_PRINTF_INFINITE_DOUBLE && !NEED_PRINTF_DOUBLE && !defined IN_LIBINTL
-
-/* Equivalent to !isfinite(x) || x == 0, but does not require libm.  */
-static int
-is_infinite_or_zero (double x)
-{
-  return isnand (x) || x + x == x;
-}
-
-#endif
-
-#if NEED_PRINTF_INFINITE_LONG_DOUBLE && !NEED_PRINTF_LONG_DOUBLE && !defined IN_LIBINTL
-
-/* Equivalent to !isfinite(x) || x == 0, but does not require libm.  */
-static int
-is_infinite_or_zerol (long double x)
-{
-  return isnanl (x) || x + x == x;
-}
-
-#endif
-
-#if (NEED_PRINTF_LONG_DOUBLE || NEED_PRINTF_DOUBLE) && !defined IN_LIBINTL
-
-/* Converting 'long double' to decimal without rare rounding bugs requires
-   real bignums.  We use the naming conventions of GNU gmp, but vastly simpler
-   (and slower) algorithms.  */
-
-typedef unsigned int mp_limb_t;
-# define GMP_LIMB_BITS 32
-typedef int mp_limb_verify[2 * (sizeof (mp_limb_t) * CHAR_BIT == GMP_LIMB_BITS) - 1];
-
-typedef unsigned long long mp_twolimb_t;
-# define GMP_TWOLIMB_BITS 64
-typedef int mp_twolimb_verify[2 * (sizeof (mp_twolimb_t) * CHAR_BIT == GMP_TWOLIMB_BITS) - 1];
-
-/* Representation of a bignum >= 0.  */
-typedef struct
-{
-  size_t nlimbs;
-  mp_limb_t *limbs; /* Bits in little-endian order, allocated with malloc().  */
-} mpn_t;
-
-/* Compute the product of two bignums >= 0.
-   Return the allocated memory in case of success, NULL in case of memory
-   allocation failure.  */
-static void *
-multiply (mpn_t src1, mpn_t src2, mpn_t *dest)
+CHAR_T *
+VASNPRINTF (CHAR_T *resultbuf, size_t *lengthp, const CHAR_T *format, va_list args)
 {
-  const mp_limb_t *p1;
-  const mp_limb_t *p2;
-  size_t len1;
-  size_t len2;
+  DIRECTIVES d;
+  arguments a;
 
-  if (src1.nlimbs <= src2.nlimbs)
-    {
-      len1 = src1.nlimbs;
-      p1 = src1.limbs;
-      len2 = src2.nlimbs;
-      p2 = src2.limbs;
-    }
-  else
-    {
-      len1 = src2.nlimbs;
-      p1 = src2.limbs;
-      len2 = src1.nlimbs;
-      p2 = src1.limbs;
-    }
-  /* Now 0 <= len1 <= len2.  */
-  if (len1 == 0)
-    {
-      /* src1 or src2 is zero.  */
-      dest->nlimbs = 0;
-      dest->limbs = (mp_limb_t *) malloc (1);
-    }
-  else
+  if (PRINTF_PARSE (format, &d, &a) < 0)
     {
-      /* Here 1 <= len1 <= len2.  */
-      size_t dlen;
-      mp_limb_t *dp;
-      size_t k, i, j;
-
-      dlen = len1 + len2;
-      dp = (mp_limb_t *) malloc (dlen * sizeof (mp_limb_t));
-      if (dp == NULL)
-        return NULL;
-      for (k = len2; k > 0; )
-        dp[--k] = 0;
-      for (i = 0; i < len1; i++)
-        {
-          mp_limb_t digit1 = p1[i];
-          mp_twolimb_t carry = 0;
-          for (j = 0; j < len2; j++)
-            {
-              mp_limb_t digit2 = p2[j];
-              carry += (mp_twolimb_t) digit1 * (mp_twolimb_t) digit2;
-              carry += dp[i + j];
-              dp[i + j] = (mp_limb_t) carry;
-              carry = carry >> GMP_LIMB_BITS;
-            }
-          dp[i + len2] = (mp_limb_t) carry;
-        }
-      /* Normalise.  */
-      while (dlen > 0 && dp[dlen - 1] == 0)
-        dlen--;
-      dest->nlimbs = dlen;
-      dest->limbs = dp;
+      errno = EINVAL;
+      return NULL;
     }
-  return dest->limbs;
-}
-
-/* Compute the quotient of a bignum a >= 0 and a bignum b > 0.
-   a is written as  a = q * b + r  with 0 <= r < b.  q is the quotient, r
-   the remainder.
-   Finally, round-to-even is performed: If r > b/2 or if r = b/2 and q is odd,
-   q is incremented.
-   Return the allocated memory in case of success, NULL in case of memory
-   allocation failure.  */
-static void *
-divide (mpn_t a, mpn_t b, mpn_t *q)
-{
-  /* Algorithm:
-     First normalise a and b: a=[a[m-1],...,a[0]], b=[b[n-1],...,b[0]]
-     with m>=0 and n>0 (in base beta = 2^GMP_LIMB_BITS).
-     If m<n, then q:=0 and r:=a.
-     If m>=n=1, perform a single-precision division:
-       r:=0, j:=m,
-       while j>0 do
-         {Here (q[m-1]*beta^(m-1)+...+q[j]*beta^j) * b[0] + r*beta^j =
-               = a[m-1]*beta^(m-1)+...+a[j]*beta^j und 0<=r<b[0]<beta}
-         j:=j-1, r:=r*beta+a[j], q[j]:=floor(r/b[0]), r:=r-b[0]*q[j].
-       Normalise [q[m-1],...,q[0]], yields q.
-     If m>=n>1, perform a multiple-precision division:
-       We have a/b < beta^(m-n+1).
-       s:=intDsize-1-(highest bit in b[n-1]), 0<=s<intDsize.
-       Shift a and b left by s bits, copying them. r:=a.
-       r=[r[m],...,r[0]], b=[b[n-1],...,b[0]] with b[n-1]>=beta/2.
-       For j=m-n,...,0: {Here 0 <= r < b*beta^(j+1).}
-         Compute q* :
-           q* := floor((r[j+n]*beta+r[j+n-1])/b[n-1]).
-           In case of overflow (q* >= beta) set q* := beta-1.
-           Compute c2 := ((r[j+n]*beta+r[j+n-1]) - q* * b[n-1])*beta + r[j+n-2]
-           and c3 := b[n-2] * q*.
-           {We have 0 <= c2 < 2*beta^2, even 0 <= c2 < beta^2 if no overflow
-            occurred.  Furthermore 0 <= c3 < beta^2.
-            If there was overflow and
-            r[j+n]*beta+r[j+n-1] - q* * b[n-1] >= beta, i.e. c2 >= beta^2,
-            the next test can be skipped.}
-           While c3 > c2, {Here 0 <= c2 < c3 < beta^2}
-             Put q* := q* - 1, c2 := c2 + b[n-1]*beta, c3 := c3 - b[n-2].
-           If q* > 0:
-             Put r := r - b * q* * beta^j. In detail:
-               [r[n+j],...,r[j]] := [r[n+j],...,r[j]] - q* * [b[n-1],...,b[0]].
-               hence: u:=0, for i:=0 to n-1 do
-                              u := u + q* * b[i],
-                              r[j+i]:=r[j+i]-(u mod beta) (+ beta, if carry),
-                              u:=u div beta (+ 1, if carry in subtraction)
-                      r[n+j]:=r[n+j]-u.
-               {Since always u = (q* * [b[i-1],...,b[0]] div beta^i) + 1
-                               < q* + 1 <= beta,
-                the carry u does not overflow.}
-             If a negative carry occurs, put q* := q* - 1
-               and [r[n+j],...,r[j]] := [r[n+j],...,r[j]] + [0,b[n-1],...,b[0]].
-         Set q[j] := q*.
-       Normalise [q[m-n],..,q[0]]; this yields the quotient q.
-       Shift [r[n-1],...,r[0]] right by s bits and normalise; this yields the
-       rest r.
-       The room for q[j] can be allocated at the memory location of r[n+j].
-     Finally, round-to-even:
-       Shift r left by 1 bit.
-       If r > b or if r = b and q[0] is odd, q := q+1.
-   */
-  const mp_limb_t *a_ptr = a.limbs;
-  size_t a_len = a.nlimbs;
-  const mp_limb_t *b_ptr = b.limbs;
-  size_t b_len = b.nlimbs;
-  mp_limb_t *roomptr;
-  mp_limb_t *tmp_roomptr = NULL;
-  mp_limb_t *q_ptr;
-  size_t q_len;
-  mp_limb_t *r_ptr;
-  size_t r_len;
 
-  /* Allocate room for a_len+2 digits.
-     (Need a_len+1 digits for the real division and 1 more digit for the
-     final rounding of q.)  */
-  roomptr = (mp_limb_t *) malloc ((a_len + 2) * sizeof (mp_limb_t));
-  if (roomptr == NULL)
-    return NULL;
-
-  /* Normalise a.  */
-  while (a_len > 0 && a_ptr[a_len - 1] == 0)
-    a_len--;
+#define CLEANUP() \
+  free (d.dir);                                                                \
+  if (a.arg)                                                           \
+    free (a.arg);
 
-  /* Normalise b.  */
-  for (;;)
+  if (printf_fetchargs (args, &a) < 0)
     {
-      if (b_len == 0)
-        /* Division by zero.  */
-        abort ();
-      if (b_ptr[b_len - 1] == 0)
-        b_len--;
-      else
-        break;
+      CLEANUP ();
+      errno = EINVAL;
+      return NULL;
     }
 
-  /* Here m = a_len >= 0 and n = b_len > 0.  */
+  {
+    size_t buf_neededlength;
+    CHAR_T *buf;
+    CHAR_T *buf_malloced;
+    const CHAR_T *cp;
+    size_t i;
+    DIRECTIVE *dp;
+    /* Output string accumulator.  */
+    CHAR_T *result;
+    size_t allocated;
+    size_t length;
 
-  if (a_len < b_len)
-    {
-      /* m<n: trivial case.  q=0, r := copy of a.  */
-      r_ptr = roomptr;
-      r_len = a_len;
-      memcpy (r_ptr, a_ptr, a_len * sizeof (mp_limb_t));
-      q_ptr = roomptr + a_len;
-      q_len = 0;
-    }
-  else if (b_len == 1)
-    {
-      /* n=1: single precision division.
-         beta^(m-1) <= a < beta^m  ==>  beta^(m-2) <= a/b < beta^m  */
-      r_ptr = roomptr;
-      q_ptr = roomptr + 1;
+    /* Allocate a small buffer that will hold a directive passed to
+       sprintf or snprintf.  */
+    buf_neededlength =
+      xsum4 (7, d.max_width_length, d.max_precision_length, 6);
+#if HAVE_ALLOCA
+    if (buf_neededlength < 4000 / sizeof (CHAR_T))
       {
-        mp_limb_t den = b_ptr[0];
-        mp_limb_t remainder = 0;
-        const mp_limb_t *sourceptr = a_ptr + a_len;
-        mp_limb_t *destptr = q_ptr + a_len;
-        size_t count;
-        for (count = a_len; count > 0; count--)
-          {
-            mp_twolimb_t num =
-              ((mp_twolimb_t) remainder << GMP_LIMB_BITS) | *--sourceptr;
-            *--destptr = num / den;
-            remainder = num % den;
-          }
-        /* Normalise and store r.  */
-        if (remainder > 0)
-          {
-            r_ptr[0] = remainder;
-            r_len = 1;
-          }
-        else
-          r_len = 0;
-        /* Normalise q.  */
-        q_len = a_len;
-        if (q_ptr[q_len - 1] == 0)
-          q_len--;
+       buf = (CHAR_T *) alloca (buf_neededlength * sizeof (CHAR_T));
+       buf_malloced = NULL;
       }
-    }
-  else
-    {
-      /* n>1: multiple precision division.
-         beta^(m-1) <= a < beta^m, beta^(n-1) <= b < beta^n  ==>
-         beta^(m-n-1) <= a/b < beta^(m-n+1).  */
-      /* Determine s.  */
-      size_t s;
+    else
+#endif
       {
-        mp_limb_t msd = b_ptr[b_len - 1]; /* = b[n-1], > 0 */
-        s = 31;
-        if (msd >= 0x10000)
-          {
-            msd = msd >> 16;
-            s -= 16;
-          }
-        if (msd >= 0x100)
-          {
-            msd = msd >> 8;
-            s -= 8;
-          }
-        if (msd >= 0x10)
-          {
-            msd = msd >> 4;
-            s -= 4;
-          }
-        if (msd >= 0x4)
-          {
-            msd = msd >> 2;
-            s -= 2;
-          }
-        if (msd >= 0x2)
-          {
-            msd = msd >> 1;
-            s -= 1;
-          }
+       size_t buf_memsize = xtimes (buf_neededlength, sizeof (CHAR_T));
+       if (size_overflow_p (buf_memsize))
+         goto out_of_memory_1;
+       buf = (CHAR_T *) malloc (buf_memsize);
+       if (buf == NULL)
+         goto out_of_memory_1;
+       buf_malloced = buf;
       }
-      /* 0 <= s < GMP_LIMB_BITS.
-         Copy b, shifting it left by s bits.  */
-      if (s > 0)
-        {
-          tmp_roomptr = (mp_limb_t *) malloc (b_len * sizeof (mp_limb_t));
-          if (tmp_roomptr == NULL)
-            {
-              free (roomptr);
-              return NULL;
-            }
-          {
-            const mp_limb_t *sourceptr = b_ptr;
-            mp_limb_t *destptr = tmp_roomptr;
-            mp_twolimb_t accu = 0;
-            size_t count;
-            for (count = b_len; count > 0; count--)
-              {
-                accu += (mp_twolimb_t) *sourceptr++ << s;
-                *destptr++ = (mp_limb_t) accu;
-                accu = accu >> GMP_LIMB_BITS;
-              }
-            /* accu must be zero, since that was how s was determined.  */
-            if (accu != 0)
-              abort ();
-          }
-          b_ptr = tmp_roomptr;
-        }
-      /* Copy a, shifting it left by s bits, yields r.
-         Memory layout:
-         At the beginning: r = roomptr[0..a_len],
-         at the end: r = roomptr[0..b_len-1], q = roomptr[b_len..a_len]  */
-      r_ptr = roomptr;
-      if (s == 0)
-        {
-          memcpy (r_ptr, a_ptr, a_len * sizeof (mp_limb_t));
-          r_ptr[a_len] = 0;
-        }
-      else
-        {
-          const mp_limb_t *sourceptr = a_ptr;
-          mp_limb_t *destptr = r_ptr;
-          mp_twolimb_t accu = 0;
-          size_t count;
-          for (count = a_len; count > 0; count--)
-            {
-              accu += (mp_twolimb_t) *sourceptr++ << s;
-              *destptr++ = (mp_limb_t) accu;
-              accu = accu >> GMP_LIMB_BITS;
-            }
-          *destptr++ = (mp_limb_t) accu;
-        }
-      q_ptr = roomptr + b_len;
-      q_len = a_len - b_len + 1; /* q will have m-n+1 limbs */
+
+    if (resultbuf != NULL)
       {
-        size_t j = a_len - b_len; /* m-n */
-        mp_limb_t b_msd = b_ptr[b_len - 1]; /* b[n-1] */
-        mp_limb_t b_2msd = b_ptr[b_len - 2]; /* b[n-2] */
-        mp_twolimb_t b_msdd = /* b[n-1]*beta+b[n-2] */
-          ((mp_twolimb_t) b_msd << GMP_LIMB_BITS) | b_2msd;
-        /* Division loop, traversed m-n+1 times.
-           j counts down, b is unchanged, beta/2 <= b[n-1] < beta.  */
-        for (;;)
-          {
-            mp_limb_t q_star;
-            mp_limb_t c1;
-            if (r_ptr[j + b_len] < b_msd) /* r[j+n] < b[n-1] ? */
-              {
-                /* Divide r[j+n]*beta+r[j+n-1] by b[n-1], no overflow.  */
-                mp_twolimb_t num =
-                  ((mp_twolimb_t) r_ptr[j + b_len] << GMP_LIMB_BITS)
-                  | r_ptr[j + b_len - 1];
-                q_star = num / b_msd;
-                c1 = num % b_msd;
-              }
-            else
-              {
-                /* Overflow, hence r[j+n]*beta+r[j+n-1] >= beta*b[n-1].  */
-                q_star = (mp_limb_t)~(mp_limb_t)0; /* q* = beta-1 */
-                /* Test whether r[j+n]*beta+r[j+n-1] - (beta-1)*b[n-1] >= beta
-                   <==> r[j+n]*beta+r[j+n-1] + b[n-1] >= beta*b[n-1]+beta
-                   <==> b[n-1] < floor((r[j+n]*beta+r[j+n-1]+b[n-1])/beta)
-                        {<= beta !}.
-                   If yes, jump directly to the subtraction loop.
-                   (Otherwise, r[j+n]*beta+r[j+n-1] - (beta-1)*b[n-1] < beta
-                    <==> floor((r[j+n]*beta+r[j+n-1]+b[n-1])/beta) = b[n-1] ) */
-                if (r_ptr[j + b_len] > b_msd
-                    || (c1 = r_ptr[j + b_len - 1] + b_msd) < b_msd)
-                  /* r[j+n] >= b[n-1]+1 or
-                     r[j+n] = b[n-1] and the addition r[j+n-1]+b[n-1] gives a
-                     carry.  */
-                  goto subtract;
-              }
-            /* q_star = q*,
-               c1 = (r[j+n]*beta+r[j+n-1]) - q* * b[n-1] (>=0, <beta).  */
-            {
-              mp_twolimb_t c2 = /* c1*beta+r[j+n-2] */
-                ((mp_twolimb_t) c1 << GMP_LIMB_BITS) | r_ptr[j + b_len - 2];
-              mp_twolimb_t c3 = /* b[n-2] * q* */
-                (mp_twolimb_t) b_2msd * (mp_twolimb_t) q_star;
-              /* While c2 < c3, increase c2 and decrease c3.
-                 Consider c3-c2.  While it is > 0, decrease it by
-                 b[n-1]*beta+b[n-2].  Because of b[n-1]*beta+b[n-2] >= beta^2/2
-                 this can happen only twice.  */
-              if (c3 > c2)
-                {
-                  q_star = q_star - 1; /* q* := q* - 1 */
-                  if (c3 - c2 > b_msdd)
-                    q_star = q_star - 1; /* q* := q* - 1 */
-                }
-            }
-            if (q_star > 0)
-              subtract:
-              {
-                /* Subtract r := r - b * q* * beta^j.  */
-                mp_limb_t cr;
-                {
-                  const mp_limb_t *sourceptr = b_ptr;
-                  mp_limb_t *destptr = r_ptr + j;
-                  mp_twolimb_t carry = 0;
-                  size_t count;
-                  for (count = b_len; count > 0; count--)
-                    {
-                      /* Here 0 <= carry <= q*.  */
-                      carry =
-                        carry
-                        + (mp_twolimb_t) q_star * (mp_twolimb_t) *sourceptr++
-                        + (mp_limb_t) ~(*destptr);
-                      /* Here 0 <= carry <= beta*q* + beta-1.  */
-                      *destptr++ = ~(mp_limb_t) carry;
-                      carry = carry >> GMP_LIMB_BITS; /* <= q* */
-                    }
-                  cr = (mp_limb_t) carry;
-                }
-                /* Subtract cr from r_ptr[j + b_len], then forget about
-                   r_ptr[j + b_len].  */
-                if (cr > r_ptr[j + b_len])
-                  {
-                    /* Subtraction gave a carry.  */
-                    q_star = q_star - 1; /* q* := q* - 1 */
-                    /* Add b back.  */
-                    {
-                      const mp_limb_t *sourceptr = b_ptr;
-                      mp_limb_t *destptr = r_ptr + j;
-                      mp_limb_t carry = 0;
-                      size_t count;
-                      for (count = b_len; count > 0; count--)
-                        {
-                          mp_limb_t source1 = *sourceptr++;
-                          mp_limb_t source2 = *destptr;
-                          *destptr++ = source1 + source2 + carry;
-                          carry =
-                            (carry
-                             ? source1 >= (mp_limb_t) ~source2
-                             : source1 > (mp_limb_t) ~source2);
-                        }
-                    }
-                    /* Forget about the carry and about r[j+n].  */
-                  }
-              }
-            /* q* is determined.  Store it as q[j].  */
-            q_ptr[j] = q_star;
-            if (j == 0)
-              break;
-            j--;
-          }
+       result = resultbuf;
+       allocated = *lengthp;
       }
-      r_len = b_len;
-      /* Normalise q.  */
-      if (q_ptr[q_len - 1] == 0)
-        q_len--;
-# if 0 /* Not needed here, since we need r only to compare it with b/2, and
-          b is shifted left by s bits.  */
-      /* Shift r right by s bits.  */
-      if (s > 0)
-        {
-          mp_limb_t ptr = r_ptr + r_len;
-          mp_twolimb_t accu = 0;
-          size_t count;
-          for (count = r_len; count > 0; count--)
-            {
-              accu = (mp_twolimb_t) (mp_limb_t) accu << GMP_LIMB_BITS;
-              accu += (mp_twolimb_t) *--ptr << (GMP_LIMB_BITS - s);
-              *ptr = (mp_limb_t) (accu >> GMP_LIMB_BITS);
-            }
-        }
-# endif
-      /* Normalise r.  */
-      while (r_len > 0 && r_ptr[r_len - 1] == 0)
-        r_len--;
-    }
-  /* Compare r << 1 with b.  */
-  if (r_len > b_len)
-    goto increment_q;
-  {
-    size_t i;
-    for (i = b_len;;)
+    else
       {
-        mp_limb_t r_i =
-          (i <= r_len && i > 0 ? r_ptr[i - 1] >> (GMP_LIMB_BITS - 1) : 0)
-          | (i < r_len ? r_ptr[i] << 1 : 0);
-        mp_limb_t b_i = (i < b_len ? b_ptr[i] : 0);
-        if (r_i > b_i)
-          goto increment_q;
-        if (r_i < b_i)
-          goto keep_q;
-        if (i == 0)
-          break;
-        i--;
+       result = NULL;
+       allocated = 0;
       }
-  }
-  if (q_len > 0 && ((q_ptr[0] & 1) != 0))
-    /* q is odd.  */
-    increment_q:
-    {
-      size_t i;
-      for (i = 0; i < q_len; i++)
-        if (++(q_ptr[i]) != 0)
-          goto keep_q;
-      q_ptr[q_len++] = 1;
-    }
-  keep_q:
-  if (tmp_roomptr != NULL)
-    free (tmp_roomptr);
-  q->limbs = q_ptr;
-  q->nlimbs = q_len;
-  return roomptr;
-}
-
-/* Convert a bignum a >= 0, multiplied with 10^extra_zeroes, to decimal
-   representation.
-   Destroys the contents of a.
-   Return the allocated memory - containing the decimal digits in low-to-high
-   order, terminated with a NUL character - in case of success, NULL in case
-   of memory allocation failure.  */
-static char *
-convert_to_decimal (mpn_t a, size_t extra_zeroes)
-{
-  mp_limb_t *a_ptr = a.limbs;
-  size_t a_len = a.nlimbs;
-  /* 0.03345 is slightly larger than log(2)/(9*log(10)).  */
-  size_t c_len = 9 * ((size_t)(a_len * (GMP_LIMB_BITS * 0.03345f)) + 1);
-  char *c_ptr = (char *) malloc (xsum (c_len, extra_zeroes));
-  if (c_ptr != NULL)
-    {
-      char *d_ptr = c_ptr;
-      for (; extra_zeroes > 0; extra_zeroes--)
-        *d_ptr++ = '0';
-      while (a_len > 0)
-        {
-          /* Divide a by 10^9, in-place.  */
-          mp_limb_t remainder = 0;
-          mp_limb_t *ptr = a_ptr + a_len;
-          size_t count;
-          for (count = a_len; count > 0; count--)
-            {
-              mp_twolimb_t num =
-                ((mp_twolimb_t) remainder << GMP_LIMB_BITS) | *--ptr;
-              *ptr = num / 1000000000;
-              remainder = num % 1000000000;
-            }
-          /* Store the remainder as 9 decimal digits.  */
-          for (count = 9; count > 0; count--)
-            {
-              *d_ptr++ = '0' + (remainder % 10);
-              remainder = remainder / 10;
-            }
-          /* Normalize a.  */
-          if (a_ptr[a_len - 1] == 0)
-            a_len--;
-        }
-      /* Remove leading zeroes.  */
-      while (d_ptr > c_ptr && d_ptr[-1] == '0')
-        d_ptr--;
-      /* But keep at least one zero.  */
-      if (d_ptr == c_ptr)
-        *d_ptr++ = '0';
-      /* Terminate the string.  */
-      *d_ptr = '\0';
-    }
-  return c_ptr;
-}
+    length = 0;
+    /* Invariants:
+       result is either == resultbuf or == NULL or malloc-allocated.
+       If length > 0, then result != NULL.  */
 
-# if NEED_PRINTF_LONG_DOUBLE
+    /* Ensures that allocated >= needed.  Aborts through a jump to
+       out_of_memory if needed is SIZE_MAX or otherwise too big.  */
+#define ENSURE_ALLOCATION(needed) \
+    if ((needed) > allocated)                                               \
+      {                                                                             \
+       size_t memory_size;                                                  \
+       CHAR_T *memory;                                                      \
+                                                                            \
+       allocated = (allocated > 0 ? xtimes (allocated, 2) : 12);            \
+       if ((needed) > allocated)                                            \
+         allocated = (needed);                                              \
+       memory_size = xtimes (allocated, sizeof (CHAR_T));                   \
+       if (size_overflow_p (memory_size))                                   \
+         goto out_of_memory;                                                \
+       if (result == resultbuf || result == NULL)                           \
+         memory = (CHAR_T *) malloc (memory_size);                          \
+       else                                                                 \
+         memory = (CHAR_T *) realloc (result, memory_size);                 \
+       if (memory == NULL)                                                  \
+         goto out_of_memory;                                                \
+       if (result == resultbuf && length > 0)                               \
+         memcpy (memory, result, length * sizeof (CHAR_T));                 \
+       result = memory;                                                     \
+      }
 
-/* Assuming x is finite and >= 0:
-   write x as x = 2^e * m, where m is a bignum.
-   Return the allocated memory in case of success, NULL in case of memory
-   allocation failure.  */
-static void *
-decode_long_double (long double x, int *ep, mpn_t *mp)
-{
-  mpn_t m;
-  int exp;
-  long double y;
-  size_t i;
+    for (cp = format, i = 0, dp = &d.dir[0]; ; cp = dp->dir_end, i++, dp++)
+      {
+       if (cp != dp->dir_start)
+         {
+           size_t n = dp->dir_start - cp;
+           size_t augmented_length = xsum (length, n);
+
+           ENSURE_ALLOCATION (augmented_length);
+           memcpy (result + length, cp, n * sizeof (CHAR_T));
+           length = augmented_length;
+         }
+       if (i == d.count)
+         break;
+
+       /* Execute a single directive.  */
+       if (dp->conversion == '%')
+         {
+           size_t augmented_length;
+
+           if (!(dp->arg_index == ARG_NONE))
+             abort ();
+           augmented_length = xsum (length, 1);
+           ENSURE_ALLOCATION (augmented_length);
+           result[length] = '%';
+           length = augmented_length;
+         }
+       else
+         {
+           if (!(dp->arg_index != ARG_NONE))
+             abort ();
+
+           if (dp->conversion == 'n')
+             {
+               switch (a.arg[dp->arg_index].type)
+                 {
+                 case TYPE_COUNT_SCHAR_POINTER:
+                   *a.arg[dp->arg_index].a.a_count_schar_pointer = length;
+                   break;
+                 case TYPE_COUNT_SHORT_POINTER:
+                   *a.arg[dp->arg_index].a.a_count_short_pointer = length;
+                   break;
+                 case TYPE_COUNT_INT_POINTER:
+                   *a.arg[dp->arg_index].a.a_count_int_pointer = length;
+                   break;
+                 case TYPE_COUNT_LONGINT_POINTER:
+                   *a.arg[dp->arg_index].a.a_count_longint_pointer = length;
+                   break;
+#ifdef HAVE_LONG_LONG
+                 case TYPE_COUNT_LONGLONGINT_POINTER:
+                   *a.arg[dp->arg_index].a.a_count_longlongint_pointer = length;
+                   break;
+#endif
+                 default:
+                   abort ();
+                 }
+             }
+           else
+             {
+               arg_type type = a.arg[dp->arg_index].type;
+               CHAR_T *p;
+               unsigned int prefix_count;
+               int prefixes[2];
+#if !USE_SNPRINTF
+               size_t tmp_length;
+               CHAR_T tmpbuf[700];
+               CHAR_T *tmp;
+
+               /* Allocate a temporary buffer of sufficient size for calling
+                  sprintf.  */
+               {
+                 size_t width;
+                 size_t precision;
+
+                 width = 0;
+                 if (dp->width_start != dp->width_end)
+                   {
+                     if (dp->width_arg_index != ARG_NONE)
+                       {
+                         int arg;
+
+                         if (!(a.arg[dp->width_arg_index].type == TYPE_INT))
+                           abort ();
+                         arg = a.arg[dp->width_arg_index].a.a_int;
+                         width = (arg < 0 ? (unsigned int) (-arg) : arg);
+                       }
+                     else
+                       {
+                         const CHAR_T *digitp = dp->width_start;
+
+                         do
+                           width = xsum (xtimes (width, 10), *digitp++ - '0');
+                         while (digitp != dp->width_end);
+                       }
+                   }
+
+                 precision = 6;
+                 if (dp->precision_start != dp->precision_end)
+                   {
+                     if (dp->precision_arg_index != ARG_NONE)
+                       {
+                         int arg;
+
+                         if (!(a.arg[dp->precision_arg_index].type == TYPE_INT))
+                           abort ();
+                         arg = a.arg[dp->precision_arg_index].a.a_int;
+                         precision = (arg < 0 ? 0 : arg);
+                       }
+                     else
+                       {
+                         const CHAR_T *digitp = dp->precision_start + 1;
+
+                         precision = 0;
+                         while (digitp != dp->precision_end)
+                           precision = xsum (xtimes (precision, 10), *digitp++ - '0');
+                       }
+                   }
+
+                 switch (dp->conversion)
+                   {
+
+                   case 'd': case 'i': case 'u':
+# ifdef HAVE_LONG_LONG
+                     if (type == TYPE_LONGLONGINT || type == TYPE_ULONGLONGINT)
+                       tmp_length =
+                         (unsigned int) (sizeof (unsigned long long) * CHAR_BIT
+                                         * 0.30103 /* binary -> decimal */
+                                         * 2 /* estimate for FLAG_GROUP */
+                                        )
+                         + 1 /* turn floor into ceil */
+                         + 1; /* account for leading sign */
+                     else
+# endif
+                     if (type == TYPE_LONGINT || type == TYPE_ULONGINT)
+                       tmp_length =
+                         (unsigned int) (sizeof (unsigned long) * CHAR_BIT
+                                         * 0.30103 /* binary -> decimal */
+                                         * 2 /* estimate for FLAG_GROUP */
+                                        )
+                         + 1 /* turn floor into ceil */
+                         + 1; /* account for leading sign */
+                     else
+                       tmp_length =
+                         (unsigned int) (sizeof (unsigned int) * CHAR_BIT
+                                         * 0.30103 /* binary -> decimal */
+                                         * 2 /* estimate for FLAG_GROUP */
+                                        )
+                         + 1 /* turn floor into ceil */
+                         + 1; /* account for leading sign */
+                     break;
+
+                   case 'o':
+# ifdef HAVE_LONG_LONG
+                     if (type == TYPE_LONGLONGINT || type == TYPE_ULONGLONGINT)
+                       tmp_length =
+                         (unsigned int) (sizeof (unsigned long long) * CHAR_BIT
+                                         * 0.333334 /* binary -> octal */
+                                        )
+                         + 1 /* turn floor into ceil */
+                         + 1; /* account for leading sign */
+                     else
+# endif
+                     if (type == TYPE_LONGINT || type == TYPE_ULONGINT)
+                       tmp_length =
+                         (unsigned int) (sizeof (unsigned long) * CHAR_BIT
+                                         * 0.333334 /* binary -> octal */
+                                        )
+                         + 1 /* turn floor into ceil */
+                         + 1; /* account for leading sign */
+                     else
+                       tmp_length =
+                         (unsigned int) (sizeof (unsigned int) * CHAR_BIT
+                                         * 0.333334 /* binary -> octal */
+                                        )
+                         + 1 /* turn floor into ceil */
+                         + 1; /* account for leading sign */
+                     break;
+
+                   case 'x': case 'X':
+# ifdef HAVE_LONG_LONG
+                     if (type == TYPE_LONGLONGINT || type == TYPE_ULONGLONGINT)
+                       tmp_length =
+                         (unsigned int) (sizeof (unsigned long long) * CHAR_BIT
+                                         * 0.25 /* binary -> hexadecimal */
+                                        )
+                         + 1 /* turn floor into ceil */
+                         + 2; /* account for leading sign or alternate form */
+                     else
+# endif
+                     if (type == TYPE_LONGINT || type == TYPE_ULONGINT)
+                       tmp_length =
+                         (unsigned int) (sizeof (unsigned long) * CHAR_BIT
+                                         * 0.25 /* binary -> hexadecimal */
+                                        )
+                         + 1 /* turn floor into ceil */
+                         + 2; /* account for leading sign or alternate form */
+                     else
+                       tmp_length =
+                         (unsigned int) (sizeof (unsigned int) * CHAR_BIT
+                                         * 0.25 /* binary -> hexadecimal */
+                                        )
+                         + 1 /* turn floor into ceil */
+                         + 2; /* account for leading sign or alternate form */
+                     break;
+
+                   case 'f': case 'F':
+# ifdef HAVE_LONG_DOUBLE
+                     if (type == TYPE_LONGDOUBLE)
+                       tmp_length =
+                         (unsigned int) (LDBL_MAX_EXP
+                                         * 0.30103 /* binary -> decimal */
+                                         * 2 /* estimate for FLAG_GROUP */
+                                        )
+                         + 1 /* turn floor into ceil */
+                         + 10; /* sign, decimal point etc. */
+                     else
+# endif
+                       tmp_length =
+                         (unsigned int) (DBL_MAX_EXP
+                                         * 0.30103 /* binary -> decimal */
+                                         * 2 /* estimate for FLAG_GROUP */
+                                        )
+                         + 1 /* turn floor into ceil */
+                         + 10; /* sign, decimal point etc. */
+                     tmp_length = xsum (tmp_length, precision);
+                     break;
+
+                   case 'e': case 'E': case 'g': case 'G':
+                   case 'a': case 'A':
+                     tmp_length =
+                       12; /* sign, decimal point, exponent etc. */
+                     tmp_length = xsum (tmp_length, precision);
+                     break;
+
+                   case 'c':
+# if defined HAVE_WINT_T && !WIDE_CHAR_VERSION
+                     if (type == TYPE_WIDE_CHAR)
+                       tmp_length = MB_CUR_MAX;
+                     else
+# endif
+                       tmp_length = 1;
+                     break;
+
+                   case 's':
+# ifdef HAVE_WCHAR_T
+                     if (type == TYPE_WIDE_STRING)
+                       {
+                         tmp_length =
+                           local_wcslen (a.arg[dp->arg_index].a.a_wide_string);
+
+#  if !WIDE_CHAR_VERSION
+                         tmp_length = xtimes (tmp_length, MB_CUR_MAX);
+#  endif
+                       }
+                     else
+# endif
+                       tmp_length = strlen (a.arg[dp->arg_index].a.a_string);
+                     break;
+
+                   case 'p':
+                     tmp_length =
+                       (unsigned int) (sizeof (void *) * CHAR_BIT
+                                       * 0.25 /* binary -> hexadecimal */
+                                      )
+                         + 1 /* turn floor into ceil */
+                         + 2; /* account for leading 0x */
+                     break;
+
+                   default:
+                     abort ();
+                   }
+
+                 if (tmp_length < width)
+                   tmp_length = width;
+
+                 tmp_length = xsum (tmp_length, 1); /* account for trailing NUL */
+               }
+
+               if (tmp_length <= sizeof (tmpbuf) / sizeof (CHAR_T))
+                 tmp = tmpbuf;
+               else
+                 {
+                   size_t tmp_memsize = xtimes (tmp_length, sizeof (CHAR_T));
+
+                   if (size_overflow_p (tmp_memsize))
+                     /* Overflow, would lead to out of memory.  */
+                     goto out_of_memory;
+                   tmp = (CHAR_T *) malloc (tmp_memsize);
+                   if (tmp == NULL)
+                     /* Out of memory.  */
+                     goto out_of_memory;
+                 }
+#endif
+
+               /* Construct the format string for calling snprintf or
+                  sprintf.  */
+               p = buf;
+               *p++ = '%';
+               if (dp->flags & FLAG_GROUP)
+                 *p++ = '\'';
+               if (dp->flags & FLAG_LEFT)
+                 *p++ = '-';
+               if (dp->flags & FLAG_SHOWSIGN)
+                 *p++ = '+';
+               if (dp->flags & FLAG_SPACE)
+                 *p++ = ' ';
+               if (dp->flags & FLAG_ALT)
+                 *p++ = '#';
+               if (dp->flags & FLAG_ZERO)
+                 *p++ = '0';
+               if (dp->width_start != dp->width_end)
+                 {
+                   size_t n = dp->width_end - dp->width_start;
+                   memcpy (p, dp->width_start, n * sizeof (CHAR_T));
+                   p += n;
+                 }
+               if (dp->precision_start != dp->precision_end)
+                 {
+                   size_t n = dp->precision_end - dp->precision_start;
+                   memcpy (p, dp->precision_start, n * sizeof (CHAR_T));
+                   p += n;
+                 }
+
+               switch (type)
+                 {
+#ifdef HAVE_LONG_LONG
+                 case TYPE_LONGLONGINT:
+                 case TYPE_ULONGLONGINT:
+                   *p++ = 'l';
+                   /*FALLTHROUGH*/
+#endif
+                 case TYPE_LONGINT:
+                 case TYPE_ULONGINT:
+#ifdef HAVE_WINT_T
+                 case TYPE_WIDE_CHAR:
+#endif
+#ifdef HAVE_WCHAR_T
+                 case TYPE_WIDE_STRING:
+#endif
+                   *p++ = 'l';
+                   break;
+#ifdef HAVE_LONG_DOUBLE
+                 case TYPE_LONGDOUBLE:
+                   *p++ = 'L';
+                   break;
+#endif
+                 default:
+                   break;
+                 }
+               *p = dp->conversion;
+#if USE_SNPRINTF
+               p[1] = '%';
+               p[2] = 'n';
+               p[3] = '\0';
+#else
+               p[1] = '\0';
+#endif
+
+               /* Construct the arguments for calling snprintf or sprintf.  */
+               prefix_count = 0;
+               if (dp->width_arg_index != ARG_NONE)
+                 {
+                   if (!(a.arg[dp->width_arg_index].type == TYPE_INT))
+                     abort ();
+                   prefixes[prefix_count++] = a.arg[dp->width_arg_index].a.a_int;
+                 }
+               if (dp->precision_arg_index != ARG_NONE)
+                 {
+                   if (!(a.arg[dp->precision_arg_index].type == TYPE_INT))
+                     abort ();
+                   prefixes[prefix_count++] = a.arg[dp->precision_arg_index].a.a_int;
+                 }
 
-  /* Allocate memory for result.  */
-  m.nlimbs = (LDBL_MANT_BIT + GMP_LIMB_BITS - 1) / GMP_LIMB_BITS;
-  m.limbs = (mp_limb_t *) malloc (m.nlimbs * sizeof (mp_limb_t));
-  if (m.limbs == NULL)
-    return NULL;
-  /* Split into exponential part and mantissa.  */
-  y = frexpl (x, &exp);
-  if (!(y >= 0.0L && y < 1.0L))
-    abort ();
-  /* x = 2^exp * y = 2^(exp - LDBL_MANT_BIT) * (y * LDBL_MANT_BIT), and the
-     latter is an integer.  */
-  /* Convert the mantissa (y * LDBL_MANT_BIT) to a sequence of limbs.
-     I'm not sure whether it's safe to cast a 'long double' value between
-     2^31 and 2^32 to 'unsigned int', therefore play safe and cast only
-     'long double' values between 0 and 2^16 (to 'unsigned int' or 'int',
-     doesn't matter).  */
-#  if (LDBL_MANT_BIT % GMP_LIMB_BITS) != 0
-#   if (LDBL_MANT_BIT % GMP_LIMB_BITS) > GMP_LIMB_BITS / 2
-    {
-      mp_limb_t hi, lo;
-      y *= (mp_limb_t) 1 << (LDBL_MANT_BIT % (GMP_LIMB_BITS / 2));
-      hi = (int) y;
-      y -= hi;
-      if (!(y >= 0.0L && y < 1.0L))
-        abort ();
-      y *= (mp_limb_t) 1 << (GMP_LIMB_BITS / 2);
-      lo = (int) y;
-      y -= lo;
-      if (!(y >= 0.0L && y < 1.0L))
-        abort ();
-      m.limbs[LDBL_MANT_BIT / GMP_LIMB_BITS] = (hi << (GMP_LIMB_BITS / 2)) | lo;
-    }
-#   else
-    {
-      mp_limb_t d;
-      y *= (mp_limb_t) 1 << (LDBL_MANT_BIT % GMP_LIMB_BITS);
-      d = (int) y;
-      y -= d;
-      if (!(y >= 0.0L && y < 1.0L))
-        abort ();
-      m.limbs[LDBL_MANT_BIT / GMP_LIMB_BITS] = d;
-    }
-#   endif
-#  endif
-  for (i = LDBL_MANT_BIT / GMP_LIMB_BITS; i > 0; )
-    {
-      mp_limb_t hi, lo;
-      y *= (mp_limb_t) 1 << (GMP_LIMB_BITS / 2);
-      hi = (int) y;
-      y -= hi;
-      if (!(y >= 0.0L && y < 1.0L))
-        abort ();
-      y *= (mp_limb_t) 1 << (GMP_LIMB_BITS / 2);
-      lo = (int) y;
-      y -= lo;
-      if (!(y >= 0.0L && y < 1.0L))
-        abort ();
-      m.limbs[--i] = (hi << (GMP_LIMB_BITS / 2)) | lo;
-    }
-#if 0 /* On FreeBSD 6.1/x86, 'long double' numbers sometimes have excess
-         precision.  */
-  if (!(y == 0.0L))
-    abort ();
+#if USE_SNPRINTF
+               /* Prepare checking whether snprintf returns the count
+                  via %n.  */
+               ENSURE_ALLOCATION (xsum (length, 1));
+               result[length] = '\0';
 #endif
-  /* Normalise.  */
-  while (m.nlimbs > 0 && m.limbs[m.nlimbs - 1] == 0)
-    m.nlimbs--;
-  *mp = m;
-  *ep = exp - LDBL_MANT_BIT;
-  return m.limbs;
-}
-
-# endif
-
-# if NEED_PRINTF_DOUBLE
-
-/* Assuming x is finite and >= 0:
-   write x as x = 2^e * m, where m is a bignum.
-   Return the allocated memory in case of success, NULL in case of memory
-   allocation failure.  */
-static void *
-decode_double (double x, int *ep, mpn_t *mp)
-{
-  mpn_t m;
-  int exp;
-  double y;
-  size_t i;
-
-  /* Allocate memory for result.  */
-  m.nlimbs = (DBL_MANT_BIT + GMP_LIMB_BITS - 1) / GMP_LIMB_BITS;
-  m.limbs = (mp_limb_t *) malloc (m.nlimbs * sizeof (mp_limb_t));
-  if (m.limbs == NULL)
-    return NULL;
-  /* Split into exponential part and mantissa.  */
-  y = frexp (x, &exp);
-  if (!(y >= 0.0 && y < 1.0))
-    abort ();
-  /* x = 2^exp * y = 2^(exp - DBL_MANT_BIT) * (y * DBL_MANT_BIT), and the
-     latter is an integer.  */
-  /* Convert the mantissa (y * DBL_MANT_BIT) to a sequence of limbs.
-     I'm not sure whether it's safe to cast a 'double' value between
-     2^31 and 2^32 to 'unsigned int', therefore play safe and cast only
-     'double' values between 0 and 2^16 (to 'unsigned int' or 'int',
-     doesn't matter).  */
-#  if (DBL_MANT_BIT % GMP_LIMB_BITS) != 0
-#   if (DBL_MANT_BIT % GMP_LIMB_BITS) > GMP_LIMB_BITS / 2
-    {
-      mp_limb_t hi, lo;
-      y *= (mp_limb_t) 1 << (DBL_MANT_BIT % (GMP_LIMB_BITS / 2));
-      hi = (int) y;
-      y -= hi;
-      if (!(y >= 0.0 && y < 1.0))
-        abort ();
-      y *= (mp_limb_t) 1 << (GMP_LIMB_BITS / 2);
-      lo = (int) y;
-      y -= lo;
-      if (!(y >= 0.0 && y < 1.0))
-        abort ();
-      m.limbs[DBL_MANT_BIT / GMP_LIMB_BITS] = (hi << (GMP_LIMB_BITS / 2)) | lo;
-    }
-#   else
-    {
-      mp_limb_t d;
-      y *= (mp_limb_t) 1 << (DBL_MANT_BIT % GMP_LIMB_BITS);
-      d = (int) y;
-      y -= d;
-      if (!(y >= 0.0 && y < 1.0))
-        abort ();
-      m.limbs[DBL_MANT_BIT / GMP_LIMB_BITS] = d;
-    }
-#   endif
-#  endif
-  for (i = DBL_MANT_BIT / GMP_LIMB_BITS; i > 0; )
-    {
-      mp_limb_t hi, lo;
-      y *= (mp_limb_t) 1 << (GMP_LIMB_BITS / 2);
-      hi = (int) y;
-      y -= hi;
-      if (!(y >= 0.0 && y < 1.0))
-        abort ();
-      y *= (mp_limb_t) 1 << (GMP_LIMB_BITS / 2);
-      lo = (int) y;
-      y -= lo;
-      if (!(y >= 0.0 && y < 1.0))
-        abort ();
-      m.limbs[--i] = (hi << (GMP_LIMB_BITS / 2)) | lo;
-    }
-  if (!(y == 0.0))
-    abort ();
-  /* Normalise.  */
-  while (m.nlimbs > 0 && m.limbs[m.nlimbs - 1] == 0)
-    m.nlimbs--;
-  *mp = m;
-  *ep = exp - DBL_MANT_BIT;
-  return m.limbs;
-}
-
-# endif
-
-/* Assuming x = 2^e * m is finite and >= 0, and n is an integer:
-   Returns the decimal representation of round (x * 10^n).
-   Return the allocated memory - containing the decimal digits in low-to-high
-   order, terminated with a NUL character - in case of success, NULL in case
-   of memory allocation failure.  */
-static char *
-scale10_round_decimal_decoded (int e, mpn_t m, void *memory, int n)
-{
-  int s;
-  size_t extra_zeroes;
-  unsigned int abs_n;
-  unsigned int abs_s;
-  mp_limb_t *pow5_ptr;
-  size_t pow5_len;
-  unsigned int s_limbs;
-  unsigned int s_bits;
-  mpn_t pow5;
-  mpn_t z;
-  void *z_memory;
-  char *digits;
-
-  if (memory == NULL)
-    return NULL;
-  /* x = 2^e * m, hence
-     y = round (2^e * 10^n * m) = round (2^(e+n) * 5^n * m)
-       = round (2^s * 5^n * m).  */
-  s = e + n;
-  extra_zeroes = 0;
-  /* Factor out a common power of 10 if possible.  */
-  if (s > 0 && n > 0)
-    {
-      extra_zeroes = (s < n ? s : n);
-      s -= extra_zeroes;
-      n -= extra_zeroes;
-    }
-  /* Here y = round (2^s * 5^n * m) * 10^extra_zeroes.
-     Before converting to decimal, we need to compute
-     z = round (2^s * 5^n * m).  */
-  /* Compute 5^|n|, possibly shifted by |s| bits if n and s have the same
-     sign.  2.322 is slightly larger than log(5)/log(2).  */
-  abs_n = (n >= 0 ? n : -n);
-  abs_s = (s >= 0 ? s : -s);
-  pow5_ptr = (mp_limb_t *) malloc (((int)(abs_n * (2.322f / GMP_LIMB_BITS)) + 1
-                                    + abs_s / GMP_LIMB_BITS + 1)
-                                   * sizeof (mp_limb_t));
-  if (pow5_ptr == NULL)
-    {
-      free (memory);
-      return NULL;
-    }
-  /* Initialize with 1.  */
-  pow5_ptr[0] = 1;
-  pow5_len = 1;
-  /* Multiply with 5^|n|.  */
-  if (abs_n > 0)
-    {
-      static mp_limb_t const small_pow5[13 + 1] =
-        {
-          1, 5, 25, 125, 625, 3125, 15625, 78125, 390625, 1953125, 9765625,
-          48828125, 244140625, 1220703125
-        };
-      unsigned int n13;
-      for (n13 = 0; n13 <= abs_n; n13 += 13)
-        {
-          mp_limb_t digit1 = small_pow5[n13 + 13 <= abs_n ? 13 : abs_n - n13];
-          size_t j;
-          mp_twolimb_t carry = 0;
-          for (j = 0; j < pow5_len; j++)
-            {
-              mp_limb_t digit2 = pow5_ptr[j];
-              carry += (mp_twolimb_t) digit1 * (mp_twolimb_t) digit2;
-              pow5_ptr[j] = (mp_limb_t) carry;
-              carry = carry >> GMP_LIMB_BITS;
-            }
-          if (carry > 0)
-            pow5_ptr[pow5_len++] = (mp_limb_t) carry;
-        }
-    }
-  s_limbs = abs_s / GMP_LIMB_BITS;
-  s_bits = abs_s % GMP_LIMB_BITS;
-  if (n >= 0 ? s >= 0 : s <= 0)
-    {
-      /* Multiply with 2^|s|.  */
-      if (s_bits > 0)
-        {
-          mp_limb_t *ptr = pow5_ptr;
-          mp_twolimb_t accu = 0;
-          size_t count;
-          for (count = pow5_len; count > 0; count--)
-            {
-              accu += (mp_twolimb_t) *ptr << s_bits;
-              *ptr++ = (mp_limb_t) accu;
-              accu = accu >> GMP_LIMB_BITS;
-            }
-          if (accu > 0)
-            {
-              *ptr = (mp_limb_t) accu;
-              pow5_len++;
-            }
-        }
-      if (s_limbs > 0)
-        {
-          size_t count;
-          for (count = pow5_len; count > 0;)
-            {
-              count--;
-              pow5_ptr[s_limbs + count] = pow5_ptr[count];
-            }
-          for (count = s_limbs; count > 0;)
-            {
-              count--;
-              pow5_ptr[count] = 0;
-            }
-          pow5_len += s_limbs;
-        }
-      pow5.limbs = pow5_ptr;
-      pow5.nlimbs = pow5_len;
-      if (n >= 0)
-        {
-          /* Multiply m with pow5.  No division needed.  */
-          z_memory = multiply (m, pow5, &z);
-        }
-      else
-        {
-          /* Divide m by pow5 and round.  */
-          z_memory = divide (m, pow5, &z);
-        }
-    }
-  else
-    {
-      pow5.limbs = pow5_ptr;
-      pow5.nlimbs = pow5_len;
-      if (n >= 0)
-        {
-          /* n >= 0, s < 0.
-             Multiply m with pow5, then divide by 2^|s|.  */
-          mpn_t numerator;
-          mpn_t denominator;
-          void *tmp_memory;
-          tmp_memory = multiply (m, pow5, &numerator);
-          if (tmp_memory == NULL)
-            {
-              free (pow5_ptr);
-              free (memory);
-              return NULL;
-            }
-          /* Construct 2^|s|.  */
-          {
-            mp_limb_t *ptr = pow5_ptr + pow5_len;
-            size_t i;
-            for (i = 0; i < s_limbs; i++)
-              ptr[i] = 0;
-            ptr[s_limbs] = (mp_limb_t) 1 << s_bits;
-            denominator.limbs = ptr;
-            denominator.nlimbs = s_limbs + 1;
-          }
-          z_memory = divide (numerator, denominator, &z);
-          free (tmp_memory);
-        }
-      else
-        {
-          /* n < 0, s > 0.
-             Multiply m with 2^s, then divide by pow5.  */
-          mpn_t numerator;
-          mp_limb_t *num_ptr;
-          num_ptr = (mp_limb_t *) malloc ((m.nlimbs + s_limbs + 1)
-                                          * sizeof (mp_limb_t));
-          if (num_ptr == NULL)
-            {
-              free (pow5_ptr);
-              free (memory);
-              return NULL;
-            }
-          {
-            mp_limb_t *destptr = num_ptr;
-            {
-              size_t i;
-              for (i = 0; i < s_limbs; i++)
-                *destptr++ = 0;
-            }
-            if (s_bits > 0)
-              {
-                const mp_limb_t *sourceptr = m.limbs;
-                mp_twolimb_t accu = 0;
-                size_t count;
-                for (count = m.nlimbs; count > 0; count--)
-                  {
-                    accu += (mp_twolimb_t) *sourceptr++ << s_bits;
-                    *destptr++ = (mp_limb_t) accu;
-                    accu = accu >> GMP_LIMB_BITS;
-                  }
-                if (accu > 0)
-                  *destptr++ = (mp_limb_t) accu;
-              }
-            else
-              {
-                const mp_limb_t *sourceptr = m.limbs;
-                size_t count;
-                for (count = m.nlimbs; count > 0; count--)
-                  *destptr++ = *sourceptr++;
-              }
-            numerator.limbs = num_ptr;
-            numerator.nlimbs = destptr - num_ptr;
-          }
-          z_memory = divide (numerator, pow5, &z);
-          free (num_ptr);
-        }
-    }
-  free (pow5_ptr);
-  free (memory);
-
-  /* Here y = round (x * 10^n) = z * 10^extra_zeroes.  */
-
-  if (z_memory == NULL)
-    return NULL;
-  digits = convert_to_decimal (z, extra_zeroes);
-  free (z_memory);
-  return digits;
-}
 
-# if NEED_PRINTF_LONG_DOUBLE
-
-/* Assuming x is finite and >= 0, and n is an integer:
-   Returns the decimal representation of round (x * 10^n).
-   Return the allocated memory - containing the decimal digits in low-to-high
-   order, terminated with a NUL character - in case of success, NULL in case
-   of memory allocation failure.  */
-static char *
-scale10_round_decimal_long_double (long double x, int n)
-{
-  int e IF_LINT(= 0);
-  mpn_t m;
-  void *memory = decode_long_double (x, &e, &m);
-  return scale10_round_decimal_decoded (e, m, memory, n);
-}
+               for (;;)
+                 {
+                   size_t maxlen;
+                   int count;
+                   int retcount;
 
-# endif
+                   maxlen = allocated - length;
+                   count = -1;
+                   retcount = 0;
 
-# if NEED_PRINTF_DOUBLE
+#if USE_SNPRINTF
+# define SNPRINTF_BUF(arg) \
+                   switch (prefix_count)                                   \
+                     {                                                     \
+                     case 0:                                               \
+                       retcount = SNPRINTF (result + length, maxlen, buf,  \
+                                            arg, &count);                  \
+                       break;                                              \
+                     case 1:                                               \
+                       retcount = SNPRINTF (result + length, maxlen, buf,  \
+                                            prefixes[0], arg, &count);     \
+                       break;                                              \
+                     case 2:                                               \
+                       retcount = SNPRINTF (result + length, maxlen, buf,  \
+                                            prefixes[0], prefixes[1], arg, \
+                                            &count);                       \
+                       break;                                              \
+                     default:                                              \
+                       abort ();                                           \
+                     }
+#else
+# define SNPRINTF_BUF(arg) \
+                   switch (prefix_count)                                   \
+                     {                                                     \
+                     case 0:                                               \
+                       count = sprintf (tmp, buf, arg);                    \
+                       break;                                              \
+                     case 1:                                               \
+                       count = sprintf (tmp, buf, prefixes[0], arg);       \
+                       break;                                              \
+                     case 2:                                               \
+                       count = sprintf (tmp, buf, prefixes[0], prefixes[1],\
+                                        arg);                              \
+                       break;                                              \
+                     default:                                              \
+                       abort ();                                           \
+                     }
+#endif
+
+                   switch (type)
+                     {
+                     case TYPE_SCHAR:
+                       {
+                         int arg = a.arg[dp->arg_index].a.a_schar;
+                         SNPRINTF_BUF (arg);
+                       }
+                       break;
+                     case TYPE_UCHAR:
+                       {
+                         unsigned int arg = a.arg[dp->arg_index].a.a_uchar;
+                         SNPRINTF_BUF (arg);
+                       }
+                       break;
+                     case TYPE_SHORT:
+                       {
+                         int arg = a.arg[dp->arg_index].a.a_short;
+                         SNPRINTF_BUF (arg);
+                       }
+                       break;
+                     case TYPE_USHORT:
+                       {
+                         unsigned int arg = a.arg[dp->arg_index].a.a_ushort;
+                         SNPRINTF_BUF (arg);
+                       }
+                       break;
+                     case TYPE_INT:
+                       {
+                         int arg = a.arg[dp->arg_index].a.a_int;
+                         SNPRINTF_BUF (arg);
+                       }
+                       break;
+                     case TYPE_UINT:
+                       {
+                         unsigned int arg = a.arg[dp->arg_index].a.a_uint;
+                         SNPRINTF_BUF (arg);
+                       }
+                       break;
+                     case TYPE_LONGINT:
+                       {
+                         long int arg = a.arg[dp->arg_index].a.a_longint;
+                         SNPRINTF_BUF (arg);
+                       }
+                       break;
+                     case TYPE_ULONGINT:
+                       {
+                         unsigned long int arg = a.arg[dp->arg_index].a.a_ulongint;
+                         SNPRINTF_BUF (arg);
+                       }
+                       break;
+#ifdef HAVE_LONG_LONG
+                     case TYPE_LONGLONGINT:
+                       {
+                         long long int arg = a.arg[dp->arg_index].a.a_longlongint;
+                         SNPRINTF_BUF (arg);
+                       }
+                       break;
+                     case TYPE_ULONGLONGINT:
+                       {
+                         unsigned long long int arg = a.arg[dp->arg_index].a.a_ulonglongint;
+                         SNPRINTF_BUF (arg);
+                       }
+                       break;
+#endif
+                     case TYPE_DOUBLE:
+                       {
+                         double arg = a.arg[dp->arg_index].a.a_double;
+                         SNPRINTF_BUF (arg);
+                       }
+                       break;
+#ifdef HAVE_LONG_DOUBLE
+                     case TYPE_LONGDOUBLE:
+                       {
+                         long double arg = a.arg[dp->arg_index].a.a_longdouble;
+                         SNPRINTF_BUF (arg);
+                       }
+                       break;
+#endif
+                     case TYPE_CHAR:
+                       {
+                         int arg = a.arg[dp->arg_index].a.a_char;
+                         SNPRINTF_BUF (arg);
+                       }
+                       break;
+#ifdef HAVE_WINT_T
+                     case TYPE_WIDE_CHAR:
+                       {
+                         wint_t arg = a.arg[dp->arg_index].a.a_wide_char;
+                         SNPRINTF_BUF (arg);
+                       }
+                       break;
+#endif
+                     case TYPE_STRING:
+                       {
+                         const char *arg = a.arg[dp->arg_index].a.a_string;
+                         SNPRINTF_BUF (arg);
+                       }
+                       break;
+#ifdef HAVE_WCHAR_T
+                     case TYPE_WIDE_STRING:
+                       {
+                         const wchar_t *arg = a.arg[dp->arg_index].a.a_wide_string;
+                         SNPRINTF_BUF (arg);
+                       }
+                       break;
+#endif
+                     case TYPE_POINTER:
+                       {
+                         void *arg = a.arg[dp->arg_index].a.a_pointer;
+                         SNPRINTF_BUF (arg);
+                       }
+                       break;
+                     default:
+                       abort ();
+                     }
 
-/* Assuming x is finite and >= 0, and n is an integer:
-   Returns the decimal representation of round (x * 10^n).
-   Return the allocated memory - containing the decimal digits in low-to-high
-   order, terminated with a NUL character - in case of success, NULL in case
-   of memory allocation failure.  */
-static char *
-scale10_round_decimal_double (double x, int n)
-{
-  int e IF_LINT(= 0);
-  mpn_t m;
-  void *memory = decode_double (x, &e, &m);
-  return scale10_round_decimal_decoded (e, m, memory, n);
-}
+#if USE_SNPRINTF
+                   /* Portability: Not all implementations of snprintf()
+                      are ISO C 99 compliant.  Determine the number of
+                      bytes that snprintf() has produced or would have
+                      produced.  */
+                   if (count >= 0)
+                     {
+                       /* Verify that snprintf() has NUL-terminated its
+                          result.  */
+                       if (count < maxlen && result[length + count] != '\0')
+                         abort ();
+                       /* Portability hack.  */
+                       if (retcount > count)
+                         count = retcount;
+                     }
+                   else
+                     {
+                       /* snprintf() doesn't understand the '%n'
+                          directive.  */
+                       if (p[1] != '\0')
+                         {
+                           /* Don't use the '%n' directive; instead, look
+                              at the snprintf() return value.  */
+                           p[1] = '\0';
+                           continue;
+                         }
+                       else
+                         {
+                           /* Look at the snprintf() return value.  */
+                           if (retcount < 0)
+                             {
+                               /* HP-UX 10.20 snprintf() is doubly deficient:
+                                  It doesn't understand the '%n' directive,
+                                  *and* it returns -1 (rather than the length
+                                  that would have been required) when the
+                                  buffer is too small.  */
+                               size_t bigger_need =
+                                 xsum (xtimes (allocated, 2), 12);
+                               ENSURE_ALLOCATION (bigger_need);
+                               continue;
+                             }
+                           else
+                             count = retcount;
+                         }
+                     }
+#endif
+
+                   /* Attempt to handle failure.  */
+                   if (count < 0)
+                     {
+                       if (!(result == resultbuf || result == NULL))
+                         free (result);
+                       if (buf_malloced != NULL)
+                         free (buf_malloced);
+                       CLEANUP ();
+                       errno = EINVAL;
+                       return NULL;
+                     }
 
-# endif
+#if !USE_SNPRINTF
+                   if (count >= tmp_length)
+                     /* tmp_length was incorrectly calculated - fix the
+                        code above!  */
+                     abort ();
+#endif
+
+                   /* Make room for the result.  */
+                   if (count >= maxlen)
+                     {
+                       /* Need at least count bytes.  But allocate
+                          proportionally, to avoid looping eternally if
+                          snprintf() reports a too small count.  */
+                       size_t n =
+                         xmax (xsum (length, count), xtimes (allocated, 2));
+
+                       ENSURE_ALLOCATION (n);
+#if USE_SNPRINTF
+                       continue;
+#endif
+                     }
 
-# if NEED_PRINTF_LONG_DOUBLE
+#if USE_SNPRINTF
+                   /* The snprintf() result did fit.  */
+#else
+                   /* Append the sprintf() result.  */
+                   memcpy (result + length, tmp, count * sizeof (CHAR_T));
+                   if (tmp != tmpbuf)
+                     free (tmp);
+#endif
 
-/* Assuming x is finite and > 0:
-   Return an approximation for n with 10^n <= x < 10^(n+1).
-   The approximation is usually the right n, but may be off by 1 sometimes.  */
-static int
-floorlog10l (long double x)
-{
-  int exp;
-  long double y;
-  double z;
-  double l;
+                   length += count;
+                   break;
+                 }
+             }
+         }
+      }
 
-  /* Split into exponential part and mantissa.  */
-  y = frexpl (x, &exp);
-  if (!(y >= 0.0L && y < 1.0L))
-    abort ();
-  if (y == 0.0L)
-    return INT_MIN;
-  if (y < 0.5L)
-    {
-      while (y < (1.0L / (1 << (GMP_LIMB_BITS / 2)) / (1 << (GMP_LIMB_BITS / 2))))
-        {
-          y *= 1.0L * (1 << (GMP_LIMB_BITS / 2)) * (1 << (GMP_LIMB_BITS / 2));
-          exp -= GMP_LIMB_BITS;
-        }
-      if (y < (1.0L / (1 << 16)))
-        {
-          y *= 1.0L * (1 << 16);
-          exp -= 16;
-        }
-      if (y < (1.0L / (1 << 8)))
-        {
-          y *= 1.0L * (1 << 8);
-          exp -= 8;
-        }
-      if (y < (1.0L / (1 << 4)))
-        {
-          y *= 1.0L * (1 << 4);
-          exp -= 4;
-        }
-      if (y < (1.0L / (1 << 2)))
-        {
-          y *= 1.0L * (1 << 2);
-          exp -= 2;
-        }
-      if (y < (1.0L / (1 << 1)))
-        {
-          y *= 1.0L * (1 << 1);
-          exp -= 1;
-        }
-    }
-  if (!(y >= 0.5L && y < 1.0L))
-    abort ();
-  /* Compute an approximation for l = log2(x) = exp + log2(y).  */
-  l = exp;
-  z = y;
-  if (z < 0.70710678118654752444)
-    {
-      z *= 1.4142135623730950488;
-      l -= 0.5;
-    }
-  if (z < 0.8408964152537145431)
-    {
-      z *= 1.1892071150027210667;
-      l -= 0.25;
-    }
-  if (z < 0.91700404320467123175)
-    {
-      z *= 1.0905077326652576592;
-      l -= 0.125;
-    }
-  if (z < 0.9576032806985736469)
-    {
-      z *= 1.0442737824274138403;
-      l -= 0.0625;
-    }
-  /* Now 0.95 <= z <= 1.01.  */
-  z = 1 - z;
-  /* log2(1-z) = 1/log(2) * (- z - z^2/2 - z^3/3 - z^4/4 - ...)
-     Four terms are enough to get an approximation with error < 10^-7.  */
-  l -= 1.4426950408889634074 * z * (1.0 + z * (0.5 + z * ((1.0 / 3) + z * 0.25)));
-  /* Finally multiply with log(2)/log(10), yields an approximation for
-     log10(x).  */
-  l *= 0.30102999566398119523;
-  /* Round down to the next integer.  */
-  return (int) l + (l < 0 ? -1 : 0);
-}
+    /* Add the final NUL.  */
+    ENSURE_ALLOCATION (xsum (length, 1));
+    result[length] = '\0';
 
-# endif
+    if (result != resultbuf && length + 1 < allocated)
+      {
+       /* Shrink the allocated memory if possible.  */
+       CHAR_T *memory;
 
-# if NEED_PRINTF_DOUBLE
+       memory = (CHAR_T *) realloc (result, (length + 1) * sizeof (CHAR_T));
+       if (memory != NULL)
+         result = memory;
+      }
 
-/* Assuming x is finite and > 0:
-   Return an approximation for n with 10^n <= x < 10^(n+1).
-   The approximation is usually the right n, but may be off by 1 sometimes.  */
-static int
-floorlog10 (double x)
-{
-  int exp;
-  double y;
-  double z;
-  double l;
+    if (buf_malloced != NULL)
+      free (buf_malloced);
+    CLEANUP ();
+    *lengthp = length;
+    if (length > INT_MAX)
+      goto length_overflow;
+    return result;
 
-  /* Split into exponential part and mantissa.  */
-  y = frexp (x, &exp);
-  if (!(y >= 0.0 && y < 1.0))
-    abort ();
-  if (y == 0.0)
-    return INT_MIN;
-  if (y < 0.5)
-    {
-      while (y < (1.0 / (1 << (GMP_LIMB_BITS / 2)) / (1 << (GMP_LIMB_BITS / 2))))
-        {
-          y *= 1.0 * (1 << (GMP_LIMB_BITS / 2)) * (1 << (GMP_LIMB_BITS / 2));
-          exp -= GMP_LIMB_BITS;
-        }
-      if (y < (1.0 / (1 << 16)))
-        {
-          y *= 1.0 * (1 << 16);
-          exp -= 16;
-        }
-      if (y < (1.0 / (1 << 8)))
-        {
-          y *= 1.0 * (1 << 8);
-          exp -= 8;
-        }
-      if (y < (1.0 / (1 << 4)))
-        {
-          y *= 1.0 * (1 << 4);
-          exp -= 4;
-        }
-      if (y < (1.0 / (1 << 2)))
-        {
-          y *= 1.0 * (1 << 2);
-          exp -= 2;
-        }
-      if (y < (1.0 / (1 << 1)))
-        {
-          y *= 1.0 * (1 << 1);
-          exp -= 1;
-        }
-    }
-  if (!(y >= 0.5 && y < 1.0))
-    abort ();
-  /* Compute an approximation for l = log2(x) = exp + log2(y).  */
-  l = exp;
-  z = y;
-  if (z < 0.70710678118654752444)
-    {
-      z *= 1.4142135623730950488;
-      l -= 0.5;
-    }
-  if (z < 0.8408964152537145431)
-    {
-      z *= 1.1892071150027210667;
-      l -= 0.25;
-    }
-  if (z < 0.91700404320467123175)
-    {
-      z *= 1.0905077326652576592;
-      l -= 0.125;
-    }
-  if (z < 0.9576032806985736469)
-    {
-      z *= 1.0442737824274138403;
-      l -= 0.0625;
-    }
-  /* Now 0.95 <= z <= 1.01.  */
-  z = 1 - z;
-  /* log2(1-z) = 1/log(2) * (- z - z^2/2 - z^3/3 - z^4/4 - ...)
-     Four terms are enough to get an approximation with error < 10^-7.  */
-  l -= 1.4426950408889634074 * z * (1.0 + z * (0.5 + z * ((1.0 / 3) + z * 0.25)));
-  /* Finally multiply with log(2)/log(10), yields an approximation for
-     log10(x).  */
-  l *= 0.30102999566398119523;
-  /* Round down to the next integer.  */
-  return (int) l + (l < 0 ? -1 : 0);
-}
-
-# endif
-
-/* Tests whether a string of digits consists of exactly PRECISION zeroes and
-   a single '1' digit.  */
-static int
-is_borderline (const char *digits, size_t precision)
-{
-  for (; precision > 0; precision--, digits++)
-    if (*digits != '0')
-      return 0;
-  if (*digits != '1')
-    return 0;
-  digits++;
-  return *digits == '\0';
-}
-
-#endif
-
-#if !USE_SNPRINTF || !HAVE_SNPRINTF_RETVAL_C99
-
-/* Use a different function name, to make it possible that the 'wchar_t'
-   parametrization and the 'char' parametrization get compiled in the same
-   translation unit.  */
-# if WIDE_CHAR_VERSION
-#  define MAX_ROOM_NEEDED wmax_room_needed
-# else
-#  define MAX_ROOM_NEEDED max_room_needed
-# endif
-
-/* Returns the number of TCHAR_T units needed as temporary space for the result
-   of sprintf or SNPRINTF of a single conversion directive.  */
-static inline size_t
-MAX_ROOM_NEEDED (const arguments *ap, size_t arg_index, FCHAR_T conversion,
-                 arg_type type, int flags, size_t width, int has_precision,
-                 size_t precision, int pad_ourselves)
-{
-  size_t tmp_length;
-
-  switch (conversion)
-    {
-    case 'd': case 'i': case 'u':
-# if HAVE_LONG_LONG_INT
-      if (type == TYPE_LONGLONGINT || type == TYPE_ULONGLONGINT)
-        tmp_length =
-          (unsigned int) (sizeof (unsigned long long) * CHAR_BIT
-                          * 0.30103 /* binary -> decimal */
-                         )
-          + 1; /* turn floor into ceil */
-      else
-# endif
-      if (type == TYPE_LONGINT || type == TYPE_ULONGINT)
-        tmp_length =
-          (unsigned int) (sizeof (unsigned long) * CHAR_BIT
-                          * 0.30103 /* binary -> decimal */
-                         )
-          + 1; /* turn floor into ceil */
-      else
-        tmp_length =
-          (unsigned int) (sizeof (unsigned int) * CHAR_BIT
-                          * 0.30103 /* binary -> decimal */
-                         )
-          + 1; /* turn floor into ceil */
-      if (tmp_length < precision)
-        tmp_length = precision;
-      /* Multiply by 2, as an estimate for FLAG_GROUP.  */
-      tmp_length = xsum (tmp_length, tmp_length);
-      /* Add 1, to account for a leading sign.  */
-      tmp_length = xsum (tmp_length, 1);
-      break;
-
-    case 'o':
-# if HAVE_LONG_LONG_INT
-      if (type == TYPE_LONGLONGINT || type == TYPE_ULONGLONGINT)
-        tmp_length =
-          (unsigned int) (sizeof (unsigned long long) * CHAR_BIT
-                          * 0.333334 /* binary -> octal */
-                         )
-          + 1; /* turn floor into ceil */
-      else
-# endif
-      if (type == TYPE_LONGINT || type == TYPE_ULONGINT)
-        tmp_length =
-          (unsigned int) (sizeof (unsigned long) * CHAR_BIT
-                          * 0.333334 /* binary -> octal */
-                         )
-          + 1; /* turn floor into ceil */
-      else
-        tmp_length =
-          (unsigned int) (sizeof (unsigned int) * CHAR_BIT
-                          * 0.333334 /* binary -> octal */
-                         )
-          + 1; /* turn floor into ceil */
-      if (tmp_length < precision)
-        tmp_length = precision;
-      /* Add 1, to account for a leading sign.  */
-      tmp_length = xsum (tmp_length, 1);
-      break;
-
-    case 'x': case 'X':
-# if HAVE_LONG_LONG_INT
-      if (type == TYPE_LONGLONGINT || type == TYPE_ULONGLONGINT)
-        tmp_length =
-          (unsigned int) (sizeof (unsigned long long) * CHAR_BIT
-                          * 0.25 /* binary -> hexadecimal */
-                         )
-          + 1; /* turn floor into ceil */
-      else
-# endif
-      if (type == TYPE_LONGINT || type == TYPE_ULONGINT)
-        tmp_length =
-          (unsigned int) (sizeof (unsigned long) * CHAR_BIT
-                          * 0.25 /* binary -> hexadecimal */
-                         )
-          + 1; /* turn floor into ceil */
-      else
-        tmp_length =
-          (unsigned int) (sizeof (unsigned int) * CHAR_BIT
-                          * 0.25 /* binary -> hexadecimal */
-                         )
-          + 1; /* turn floor into ceil */
-      if (tmp_length < precision)
-        tmp_length = precision;
-      /* Add 2, to account for a leading sign or alternate form.  */
-      tmp_length = xsum (tmp_length, 2);
-      break;
-
-    case 'f': case 'F':
-      if (type == TYPE_LONGDOUBLE)
-        tmp_length =
-          (unsigned int) (LDBL_MAX_EXP
-                          * 0.30103 /* binary -> decimal */
-                          * 2 /* estimate for FLAG_GROUP */
-                         )
-          + 1 /* turn floor into ceil */
-          + 10; /* sign, decimal point etc. */
-      else
-        tmp_length =
-          (unsigned int) (DBL_MAX_EXP
-                          * 0.30103 /* binary -> decimal */
-                          * 2 /* estimate for FLAG_GROUP */
-                         )
-          + 1 /* turn floor into ceil */
-          + 10; /* sign, decimal point etc. */
-      tmp_length = xsum (tmp_length, precision);
-      break;
-
-    case 'e': case 'E': case 'g': case 'G':
-      tmp_length =
-        12; /* sign, decimal point, exponent etc. */
-      tmp_length = xsum (tmp_length, precision);
-      break;
-
-    case 'a': case 'A':
-      if (type == TYPE_LONGDOUBLE)
-        tmp_length =
-          (unsigned int) (LDBL_DIG
-                          * 0.831 /* decimal -> hexadecimal */
-                         )
-          + 1; /* turn floor into ceil */
-      else
-        tmp_length =
-          (unsigned int) (DBL_DIG
-                          * 0.831 /* decimal -> hexadecimal */
-                         )
-          + 1; /* turn floor into ceil */
-      if (tmp_length < precision)
-        tmp_length = precision;
-      /* Account for sign, decimal point etc. */
-      tmp_length = xsum (tmp_length, 12);
-      break;
-
-    case 'c':
-# if HAVE_WINT_T && !WIDE_CHAR_VERSION
-      if (type == TYPE_WIDE_CHAR)
-        tmp_length = MB_CUR_MAX;
-      else
-# endif
-        tmp_length = 1;
-      break;
-
-    case 's':
-# if HAVE_WCHAR_T
-      if (type == TYPE_WIDE_STRING)
-        {
-#  if WIDE_CHAR_VERSION
-          /* ISO C says about %ls in fwprintf:
-               "If the precision is not specified or is greater than the size
-                of the array, the array shall contain a null wide character."
-             So if there is a precision, we must not use wcslen.  */
-          const wchar_t *arg = ap->arg[arg_index].a.a_wide_string;
-
-          if (has_precision)
-            tmp_length = local_wcsnlen (arg, precision);
-          else
-            tmp_length = local_wcslen (arg);
-#  else
-          /* ISO C says about %ls in fprintf:
-               "If a precision is specified, no more than that many bytes are
-                written (including shift sequences, if any), and the array
-                shall contain a null wide character if, to equal the multibyte
-                character sequence length given by the precision, the function
-                would need to access a wide character one past the end of the
-                array."
-             So if there is a precision, we must not use wcslen.  */
-          /* This case has already been handled separately in VASNPRINTF.  */
-          abort ();
-#  endif
-        }
-      else
-# endif
-        {
-# if WIDE_CHAR_VERSION
-          /* ISO C says about %s in fwprintf:
-               "If the precision is not specified or is greater than the size
-                of the converted array, the converted array shall contain a
-                null wide character."
-             So if there is a precision, we must not use strlen.  */
-          /* This case has already been handled separately in VASNPRINTF.  */
-          abort ();
-# else
-          /* ISO C says about %s in fprintf:
-               "If the precision is not specified or greater than the size of
-                the array, the array shall contain a null character."
-             So if there is a precision, we must not use strlen.  */
-          const char *arg = ap->arg[arg_index].a.a_string;
-
-          if (has_precision)
-            tmp_length = local_strnlen (arg, precision);
-          else
-            tmp_length = strlen (arg);
-# endif
-        }
-      break;
-
-    case 'p':
-      tmp_length =
-        (unsigned int) (sizeof (void *) * CHAR_BIT
-                        * 0.25 /* binary -> hexadecimal */
-                       )
-          + 1 /* turn floor into ceil */
-          + 2; /* account for leading 0x */
-      break;
-
-    default:
-      abort ();
-    }
-
-  if (!pad_ourselves)
-    {
-# if ENABLE_UNISTDIO
-      /* Padding considers the number of characters, therefore the number of
-         elements after padding may be
-           > max (tmp_length, width)
-         but is certainly
-           <= tmp_length + width.  */
-      tmp_length = xsum (tmp_length, width);
-# else
-      /* Padding considers the number of elements, says POSIX.  */
-      if (tmp_length < width)
-        tmp_length = width;
-# endif
-    }
-
-  tmp_length = xsum (tmp_length, 1); /* account for trailing NUL */
-
-  return tmp_length;
-}
-
-#endif
-
-DCHAR_T *
-VASNPRINTF (DCHAR_T *resultbuf, size_t *lengthp,
-            const FCHAR_T *format, va_list args)
-{
-  DIRECTIVES d;
-  arguments a;
-
-  if (PRINTF_PARSE (format, &d, &a) < 0)
-    /* errno is already set.  */
-    return NULL;
-
-#define CLEANUP() \
-  free (d.dir);                                                         \
-  if (a.arg)                                                            \
-    free (a.arg);
-
-  if (PRINTF_FETCHARGS (args, &a) < 0)
-    {
-      CLEANUP ();
-      errno = EINVAL;
-      return NULL;
-    }
-
-  {
-    size_t buf_neededlength;
-    TCHAR_T *buf;
-    TCHAR_T *buf_malloced;
-    const FCHAR_T *cp;
-    size_t i;
-    DIRECTIVE *dp;
-    /* Output string accumulator.  */
-    DCHAR_T *result;
-    size_t allocated;
-    size_t length;
-
-    /* Allocate a small buffer that will hold a directive passed to
-       sprintf or snprintf.  */
-    buf_neededlength =
-      xsum4 (7, d.max_width_length, d.max_precision_length, 6);
-#if HAVE_ALLOCA
-    if (buf_neededlength < 4000 / sizeof (TCHAR_T))
-      {
-        buf = (TCHAR_T *) alloca (buf_neededlength * sizeof (TCHAR_T));
-        buf_malloced = NULL;
-      }
-    else
-#endif
-      {
-        size_t buf_memsize = xtimes (buf_neededlength, sizeof (TCHAR_T));
-        if (size_overflow_p (buf_memsize))
-          goto out_of_memory_1;
-        buf = (TCHAR_T *) malloc (buf_memsize);
-        if (buf == NULL)
-          goto out_of_memory_1;
-        buf_malloced = buf;
-      }
-
-    if (resultbuf != NULL)
-      {
-        result = resultbuf;
-        allocated = *lengthp;
-      }
-    else
-      {
-        result = NULL;
-        allocated = 0;
-      }
-    length = 0;
-    /* Invariants:
-       result is either == resultbuf or == NULL or malloc-allocated.
-       If length > 0, then result != NULL.  */
-
-    /* Ensures that allocated >= needed.  Aborts through a jump to
-       out_of_memory if needed is SIZE_MAX or otherwise too big.  */
-#define ENSURE_ALLOCATION(needed) \
-    if ((needed) > allocated)                                                \
-      {                                                                      \
-        size_t memory_size;                                                  \
-        DCHAR_T *memory;                                                     \
-                                                                             \
-        allocated = (allocated > 0 ? xtimes (allocated, 2) : 12);            \
-        if ((needed) > allocated)                                            \
-          allocated = (needed);                                              \
-        memory_size = xtimes (allocated, sizeof (DCHAR_T));                  \
-        if (size_overflow_p (memory_size))                                   \
-          goto out_of_memory;                                                \
-        if (result == resultbuf || result == NULL)                           \
-          memory = (DCHAR_T *) malloc (memory_size);                         \
-        else                                                                 \
-          memory = (DCHAR_T *) realloc (result, memory_size);                \
-        if (memory == NULL)                                                  \
-          goto out_of_memory;                                                \
-        if (result == resultbuf && length > 0)                               \
-          DCHAR_CPY (memory, result, length);                                \
-        result = memory;                                                     \
-      }
-
-    for (cp = format, i = 0, dp = &d.dir[0]; ; cp = dp->dir_end, i++, dp++)
-      {
-        if (cp != dp->dir_start)
-          {
-            size_t n = dp->dir_start - cp;
-            size_t augmented_length = xsum (length, n);
-
-            ENSURE_ALLOCATION (augmented_length);
-            /* This copies a piece of FCHAR_T[] into a DCHAR_T[].  Here we
-               need that the format string contains only ASCII characters
-               if FCHAR_T and DCHAR_T are not the same type.  */
-            if (sizeof (FCHAR_T) == sizeof (DCHAR_T))
-              {
-                DCHAR_CPY (result + length, (const DCHAR_T *) cp, n);
-                length = augmented_length;
-              }
-            else
-              {
-                do
-                  result[length++] = (unsigned char) *cp++;
-                while (--n > 0);
-              }
-          }
-        if (i == d.count)
-          break;
-
-        /* Execute a single directive.  */
-        if (dp->conversion == '%')
-          {
-            size_t augmented_length;
-
-            if (!(dp->arg_index == ARG_NONE))
-              abort ();
-            augmented_length = xsum (length, 1);
-            ENSURE_ALLOCATION (augmented_length);
-            result[length] = '%';
-            length = augmented_length;
-          }
-        else
-          {
-            if (!(dp->arg_index != ARG_NONE))
-              abort ();
-
-            if (dp->conversion == 'n')
-              {
-                switch (a.arg[dp->arg_index].type)
-                  {
-                  case TYPE_COUNT_SCHAR_POINTER:
-                    *a.arg[dp->arg_index].a.a_count_schar_pointer = length;
-                    break;
-                  case TYPE_COUNT_SHORT_POINTER:
-                    *a.arg[dp->arg_index].a.a_count_short_pointer = length;
-                    break;
-                  case TYPE_COUNT_INT_POINTER:
-                    *a.arg[dp->arg_index].a.a_count_int_pointer = length;
-                    break;
-                  case TYPE_COUNT_LONGINT_POINTER:
-                    *a.arg[dp->arg_index].a.a_count_longint_pointer = length;
-                    break;
-#if HAVE_LONG_LONG_INT
-                  case TYPE_COUNT_LONGLONGINT_POINTER:
-                    *a.arg[dp->arg_index].a.a_count_longlongint_pointer = length;
-                    break;
-#endif
-                  default:
-                    abort ();
-                  }
-              }
-#if ENABLE_UNISTDIO
-            /* The unistdio extensions.  */
-            else if (dp->conversion == 'U')
-              {
-                arg_type type = a.arg[dp->arg_index].type;
-                int flags = dp->flags;
-                int has_width;
-                size_t width;
-                int has_precision;
-                size_t precision;
-
-                has_width = 0;
-                width = 0;
-                if (dp->width_start != dp->width_end)
-                  {
-                    if (dp->width_arg_index != ARG_NONE)
-                      {
-                        int arg;
-
-                        if (!(a.arg[dp->width_arg_index].type == TYPE_INT))
-                          abort ();
-                        arg = a.arg[dp->width_arg_index].a.a_int;
-                        if (arg < 0)
-                          {
-                            /* "A negative field width is taken as a '-' flag
-                                followed by a positive field width."  */
-                            flags |= FLAG_LEFT;
-                            width = (unsigned int) (-arg);
-                          }
-                        else
-                          width = arg;
-                      }
-                    else
-                      {
-                        const FCHAR_T *digitp = dp->width_start;
-
-                        do
-                          width = xsum (xtimes (width, 10), *digitp++ - '0');
-                        while (digitp != dp->width_end);
-                      }
-                    has_width = 1;
-                  }
-
-                has_precision = 0;
-                precision = 0;
-                if (dp->precision_start != dp->precision_end)
-                  {
-                    if (dp->precision_arg_index != ARG_NONE)
-                      {
-                        int arg;
-
-                        if (!(a.arg[dp->precision_arg_index].type == TYPE_INT))
-                          abort ();
-                        arg = a.arg[dp->precision_arg_index].a.a_int;
-                        /* "A negative precision is taken as if the precision
-                            were omitted."  */
-                        if (arg >= 0)
-                          {
-                            precision = arg;
-                            has_precision = 1;
-                          }
-                      }
-                    else
-                      {
-                        const FCHAR_T *digitp = dp->precision_start + 1;
-
-                        precision = 0;
-                        while (digitp != dp->precision_end)
-                          precision = xsum (xtimes (precision, 10), *digitp++ - '0');
-                        has_precision = 1;
-                      }
-                  }
-
-                switch (type)
-                  {
-                  case TYPE_U8_STRING:
-                    {
-                      const uint8_t *arg = a.arg[dp->arg_index].a.a_u8_string;
-                      const uint8_t *arg_end;
-                      size_t characters;
-
-                      if (has_precision)
-                        {
-                          /* Use only PRECISION characters, from the left.  */
-                          arg_end = arg;
-                          characters = 0;
-                          for (; precision > 0; precision--)
-                            {
-                              int count = u8_strmblen (arg_end);
-                              if (count == 0)
-                                break;
-                              if (count < 0)
-                                {
-                                  if (!(result == resultbuf || result == NULL))
-                                    free (result);
-                                  if (buf_malloced != NULL)
-                                    free (buf_malloced);
-                                  CLEANUP ();
-                                  errno = EILSEQ;
-                                  return NULL;
-                                }
-                              arg_end += count;
-                              characters++;
-                            }
-                        }
-                      else if (has_width)
-                        {
-                          /* Use the entire string, and count the number of
-                             characters.  */
-                          arg_end = arg;
-                          characters = 0;
-                          for (;;)
-                            {
-                              int count = u8_strmblen (arg_end);
-                              if (count == 0)
-                                break;
-                              if (count < 0)
-                                {
-                                  if (!(result == resultbuf || result == NULL))
-                                    free (result);
-                                  if (buf_malloced != NULL)
-                                    free (buf_malloced);
-                                  CLEANUP ();
-                                  errno = EILSEQ;
-                                  return NULL;
-                                }
-                              arg_end += count;
-                              characters++;
-                            }
-                        }
-                      else
-                        {
-                          /* Use the entire string.  */
-                          arg_end = arg + u8_strlen (arg);
-                          /* The number of characters doesn't matter.  */
-                          characters = 0;
-                        }
-
-                      if (has_width && width > characters
-                          && !(dp->flags & FLAG_LEFT))
-                        {
-                          size_t n = width - characters;
-                          ENSURE_ALLOCATION (xsum (length, n));
-                          DCHAR_SET (result + length, ' ', n);
-                          length += n;
-                        }
-
-# if DCHAR_IS_UINT8_T
-                      {
-                        size_t n = arg_end - arg;
-                        ENSURE_ALLOCATION (xsum (length, n));
-                        DCHAR_CPY (result + length, arg, n);
-                        length += n;
-                      }
-# else
-                      { /* Convert.  */
-                        DCHAR_T *converted = result + length;
-                        size_t converted_len = allocated - length;
-#  if DCHAR_IS_TCHAR
-                        /* Convert from UTF-8 to locale encoding.  */
-                        converted =
-                          u8_conv_to_encoding (locale_charset (),
-                                               iconveh_question_mark,
-                                               arg, arg_end - arg, NULL,
-                                               converted, &converted_len);
-#  else
-                        /* Convert from UTF-8 to UTF-16/UTF-32.  */
-                        converted =
-                          U8_TO_DCHAR (arg, arg_end - arg,
-                                       converted, &converted_len);
-#  endif
-                        if (converted == NULL)
-                          {
-                            int saved_errno = errno;
-                            if (!(result == resultbuf || result == NULL))
-                              free (result);
-                            if (buf_malloced != NULL)
-                              free (buf_malloced);
-                            CLEANUP ();
-                            errno = saved_errno;
-                            return NULL;
-                          }
-                        if (converted != result + length)
-                          {
-                            ENSURE_ALLOCATION (xsum (length, converted_len));
-                            DCHAR_CPY (result + length, converted, converted_len);
-                            free (converted);
-                          }
-                        length += converted_len;
-                      }
-# endif
-
-                      if (has_width && width > characters
-                          && (dp->flags & FLAG_LEFT))
-                        {
-                          size_t n = width - characters;
-                          ENSURE_ALLOCATION (xsum (length, n));
-                          DCHAR_SET (result + length, ' ', n);
-                          length += n;
-                        }
-                    }
-                    break;
-
-                  case TYPE_U16_STRING:
-                    {
-                      const uint16_t *arg = a.arg[dp->arg_index].a.a_u16_string;
-                      const uint16_t *arg_end;
-                      size_t characters;
-
-                      if (has_precision)
-                        {
-                          /* Use only PRECISION characters, from the left.  */
-                          arg_end = arg;
-                          characters = 0;
-                          for (; precision > 0; precision--)
-                            {
-                              int count = u16_strmblen (arg_end);
-                              if (count == 0)
-                                break;
-                              if (count < 0)
-                                {
-                                  if (!(result == resultbuf || result == NULL))
-                                    free (result);
-                                  if (buf_malloced != NULL)
-                                    free (buf_malloced);
-                                  CLEANUP ();
-                                  errno = EILSEQ;
-                                  return NULL;
-                                }
-                              arg_end += count;
-                              characters++;
-                            }
-                        }
-                      else if (has_width)
-                        {
-                          /* Use the entire string, and count the number of
-                             characters.  */
-                          arg_end = arg;
-                          characters = 0;
-                          for (;;)
-                            {
-                              int count = u16_strmblen (arg_end);
-                              if (count == 0)
-                                break;
-                              if (count < 0)
-                                {
-                                  if (!(result == resultbuf || result == NULL))
-                                    free (result);
-                                  if (buf_malloced != NULL)
-                                    free (buf_malloced);
-                                  CLEANUP ();
-                                  errno = EILSEQ;
-                                  return NULL;
-                                }
-                              arg_end += count;
-                              characters++;
-                            }
-                        }
-                      else
-                        {
-                          /* Use the entire string.  */
-                          arg_end = arg + u16_strlen (arg);
-                          /* The number of characters doesn't matter.  */
-                          characters = 0;
-                        }
-
-                      if (has_width && width > characters
-                          && !(dp->flags & FLAG_LEFT))
-                        {
-                          size_t n = width - characters;
-                          ENSURE_ALLOCATION (xsum (length, n));
-                          DCHAR_SET (result + length, ' ', n);
-                          length += n;
-                        }
-
-# if DCHAR_IS_UINT16_T
-                      {
-                        size_t n = arg_end - arg;
-                        ENSURE_ALLOCATION (xsum (length, n));
-                        DCHAR_CPY (result + length, arg, n);
-                        length += n;
-                      }
-# else
-                      { /* Convert.  */
-                        DCHAR_T *converted = result + length;
-                        size_t converted_len = allocated - length;
-#  if DCHAR_IS_TCHAR
-                        /* Convert from UTF-16 to locale encoding.  */
-                        converted =
-                          u16_conv_to_encoding (locale_charset (),
-                                                iconveh_question_mark,
-                                                arg, arg_end - arg, NULL,
-                                                converted, &converted_len);
-#  else
-                        /* Convert from UTF-16 to UTF-8/UTF-32.  */
-                        converted =
-                          U16_TO_DCHAR (arg, arg_end - arg,
-                                        converted, &converted_len);
-#  endif
-                        if (converted == NULL)
-                          {
-                            int saved_errno = errno;
-                            if (!(result == resultbuf || result == NULL))
-                              free (result);
-                            if (buf_malloced != NULL)
-                              free (buf_malloced);
-                            CLEANUP ();
-                            errno = saved_errno;
-                            return NULL;
-                          }
-                        if (converted != result + length)
-                          {
-                            ENSURE_ALLOCATION (xsum (length, converted_len));
-                            DCHAR_CPY (result + length, converted, converted_len);
-                            free (converted);
-                          }
-                        length += converted_len;
-                      }
-# endif
-
-                      if (has_width && width > characters
-                          && (dp->flags & FLAG_LEFT))
-                        {
-                          size_t n = width - characters;
-                          ENSURE_ALLOCATION (xsum (length, n));
-                          DCHAR_SET (result + length, ' ', n);
-                          length += n;
-                        }
-                    }
-                    break;
-
-                  case TYPE_U32_STRING:
-                    {
-                      const uint32_t *arg = a.arg[dp->arg_index].a.a_u32_string;
-                      const uint32_t *arg_end;
-                      size_t characters;
-
-                      if (has_precision)
-                        {
-                          /* Use only PRECISION characters, from the left.  */
-                          arg_end = arg;
-                          characters = 0;
-                          for (; precision > 0; precision--)
-                            {
-                              int count = u32_strmblen (arg_end);
-                              if (count == 0)
-                                break;
-                              if (count < 0)
-                                {
-                                  if (!(result == resultbuf || result == NULL))
-                                    free (result);
-                                  if (buf_malloced != NULL)
-                                    free (buf_malloced);
-                                  CLEANUP ();
-                                  errno = EILSEQ;
-                                  return NULL;
-                                }
-                              arg_end += count;
-                              characters++;
-                            }
-                        }
-                      else if (has_width)
-                        {
-                          /* Use the entire string, and count the number of
-                             characters.  */
-                          arg_end = arg;
-                          characters = 0;
-                          for (;;)
-                            {
-                              int count = u32_strmblen (arg_end);
-                              if (count == 0)
-                                break;
-                              if (count < 0)
-                                {
-                                  if (!(result == resultbuf || result == NULL))
-                                    free (result);
-                                  if (buf_malloced != NULL)
-                                    free (buf_malloced);
-                                  CLEANUP ();
-                                  errno = EILSEQ;
-                                  return NULL;
-                                }
-                              arg_end += count;
-                              characters++;
-                            }
-                        }
-                      else
-                        {
-                          /* Use the entire string.  */
-                          arg_end = arg + u32_strlen (arg);
-                          /* The number of characters doesn't matter.  */
-                          characters = 0;
-                        }
-
-                      if (has_width && width > characters
-                          && !(dp->flags & FLAG_LEFT))
-                        {
-                          size_t n = width - characters;
-                          ENSURE_ALLOCATION (xsum (length, n));
-                          DCHAR_SET (result + length, ' ', n);
-                          length += n;
-                        }
-
-# if DCHAR_IS_UINT32_T
-                      {
-                        size_t n = arg_end - arg;
-                        ENSURE_ALLOCATION (xsum (length, n));
-                        DCHAR_CPY (result + length, arg, n);
-                        length += n;
-                      }
-# else
-                      { /* Convert.  */
-                        DCHAR_T *converted = result + length;
-                        size_t converted_len = allocated - length;
-#  if DCHAR_IS_TCHAR
-                        /* Convert from UTF-32 to locale encoding.  */
-                        converted =
-                          u32_conv_to_encoding (locale_charset (),
-                                                iconveh_question_mark,
-                                                arg, arg_end - arg, NULL,
-                                                converted, &converted_len);
-#  else
-                        /* Convert from UTF-32 to UTF-8/UTF-16.  */
-                        converted =
-                          U32_TO_DCHAR (arg, arg_end - arg,
-                                        converted, &converted_len);
-#  endif
-                        if (converted == NULL)
-                          {
-                            int saved_errno = errno;
-                            if (!(result == resultbuf || result == NULL))
-                              free (result);
-                            if (buf_malloced != NULL)
-                              free (buf_malloced);
-                            CLEANUP ();
-                            errno = saved_errno;
-                            return NULL;
-                          }
-                        if (converted != result + length)
-                          {
-                            ENSURE_ALLOCATION (xsum (length, converted_len));
-                            DCHAR_CPY (result + length, converted, converted_len);
-                            free (converted);
-                          }
-                        length += converted_len;
-                      }
-# endif
-
-                      if (has_width && width > characters
-                          && (dp->flags & FLAG_LEFT))
-                        {
-                          size_t n = width - characters;
-                          ENSURE_ALLOCATION (xsum (length, n));
-                          DCHAR_SET (result + length, ' ', n);
-                          length += n;
-                        }
-                    }
-                    break;
-
-                  default:
-                    abort ();
-                  }
-              }
-#endif
-#if (!USE_SNPRINTF || !HAVE_SNPRINTF_RETVAL_C99 || (NEED_PRINTF_DIRECTIVE_LS && !defined IN_LIBINTL)) && HAVE_WCHAR_T
-            else if (dp->conversion == 's'
-# if WIDE_CHAR_VERSION
-                     && a.arg[dp->arg_index].type != TYPE_WIDE_STRING
-# else
-                     && a.arg[dp->arg_index].type == TYPE_WIDE_STRING
-# endif
-                    )
-              {
-                /* The normal handling of the 's' directive below requires
-                   allocating a temporary buffer.  The determination of its
-                   length (tmp_length), in the case when a precision is
-                   specified, below requires a conversion between a char[]
-                   string and a wchar_t[] wide string.  It could be done, but
-                   we have no guarantee that the implementation of sprintf will
-                   use the exactly same algorithm.  Without this guarantee, it
-                   is possible to have buffer overrun bugs.  In order to avoid
-                   such bugs, we implement the entire processing of the 's'
-                   directive ourselves.  */
-                int flags = dp->flags;
-                int has_width;
-                size_t width;
-                int has_precision;
-                size_t precision;
-
-                has_width = 0;
-                width = 0;
-                if (dp->width_start != dp->width_end)
-                  {
-                    if (dp->width_arg_index != ARG_NONE)
-                      {
-                        int arg;
-
-                        if (!(a.arg[dp->width_arg_index].type == TYPE_INT))
-                          abort ();
-                        arg = a.arg[dp->width_arg_index].a.a_int;
-                        if (arg < 0)
-                          {
-                            /* "A negative field width is taken as a '-' flag
-                                followed by a positive field width."  */
-                            flags |= FLAG_LEFT;
-                            width = (unsigned int) (-arg);
-                          }
-                        else
-                          width = arg;
-                      }
-                    else
-                      {
-                        const FCHAR_T *digitp = dp->width_start;
-
-                        do
-                          width = xsum (xtimes (width, 10), *digitp++ - '0');
-                        while (digitp != dp->width_end);
-                      }
-                    has_width = 1;
-                  }
-
-                has_precision = 0;
-                precision = 6;
-                if (dp->precision_start != dp->precision_end)
-                  {
-                    if (dp->precision_arg_index != ARG_NONE)
-                      {
-                        int arg;
-
-                        if (!(a.arg[dp->precision_arg_index].type == TYPE_INT))
-                          abort ();
-                        arg = a.arg[dp->precision_arg_index].a.a_int;
-                        /* "A negative precision is taken as if the precision
-                            were omitted."  */
-                        if (arg >= 0)
-                          {
-                            precision = arg;
-                            has_precision = 1;
-                          }
-                      }
-                    else
-                      {
-                        const FCHAR_T *digitp = dp->precision_start + 1;
-
-                        precision = 0;
-                        while (digitp != dp->precision_end)
-                          precision = xsum (xtimes (precision, 10), *digitp++ - '0');
-                        has_precision = 1;
-                      }
-                  }
-
-# if WIDE_CHAR_VERSION
-                /* %s in vasnwprintf.  See the specification of fwprintf.  */
-                {
-                  const char *arg = a.arg[dp->arg_index].a.a_string;
-                  const char *arg_end;
-                  size_t characters;
-
-                  if (has_precision)
-                    {
-                      /* Use only as many bytes as needed to produce PRECISION
-                         wide characters, from the left.  */
-#  if HAVE_MBRTOWC
-                      mbstate_t state;
-                      memset (&state, '\0', sizeof (mbstate_t));
-#  endif
-                      arg_end = arg;
-                      characters = 0;
-                      for (; precision > 0; precision--)
-                        {
-                          int count;
-#  if HAVE_MBRTOWC
-                          count = mbrlen (arg_end, MB_CUR_MAX, &state);
-#  else
-                          count = mblen (arg_end, MB_CUR_MAX);
-#  endif
-                          if (count == 0)
-                            /* Found the terminating NUL.  */
-                            break;
-                          if (count < 0)
-                            {
-                              /* Invalid or incomplete multibyte character.  */
-                              if (!(result == resultbuf || result == NULL))
-                                free (result);
-                              if (buf_malloced != NULL)
-                                free (buf_malloced);
-                              CLEANUP ();
-                              errno = EILSEQ;
-                              return NULL;
-                            }
-                          arg_end += count;
-                          characters++;
-                        }
-                    }
-                  else if (has_width)
-                    {
-                      /* Use the entire string, and count the number of wide
-                         characters.  */
-#  if HAVE_MBRTOWC
-                      mbstate_t state;
-                      memset (&state, '\0', sizeof (mbstate_t));
-#  endif
-                      arg_end = arg;
-                      characters = 0;
-                      for (;;)
-                        {
-                          int count;
-#  if HAVE_MBRTOWC
-                          count = mbrlen (arg_end, MB_CUR_MAX, &state);
-#  else
-                          count = mblen (arg_end, MB_CUR_MAX);
-#  endif
-                          if (count == 0)
-                            /* Found the terminating NUL.  */
-                            break;
-                          if (count < 0)
-                            {
-                              /* Invalid or incomplete multibyte character.  */
-                              if (!(result == resultbuf || result == NULL))
-                                free (result);
-                              if (buf_malloced != NULL)
-                                free (buf_malloced);
-                              CLEANUP ();
-                              errno = EILSEQ;
-                              return NULL;
-                            }
-                          arg_end += count;
-                          characters++;
-                        }
-                    }
-                  else
-                    {
-                      /* Use the entire string.  */
-                      arg_end = arg + strlen (arg);
-                      /* The number of characters doesn't matter.  */
-                      characters = 0;
-                    }
-
-                  if (has_width && width > characters
-                      && !(dp->flags & FLAG_LEFT))
-                    {
-                      size_t n = width - characters;
-                      ENSURE_ALLOCATION (xsum (length, n));
-                      DCHAR_SET (result + length, ' ', n);
-                      length += n;
-                    }
-
-                  if (has_precision || has_width)
-                    {
-                      /* We know the number of wide characters in advance.  */
-                      size_t remaining;
-#  if HAVE_MBRTOWC
-                      mbstate_t state;
-                      memset (&state, '\0', sizeof (mbstate_t));
-#  endif
-                      ENSURE_ALLOCATION (xsum (length, characters));
-                      for (remaining = characters; remaining > 0; remaining--)
-                        {
-                          wchar_t wc;
-                          int count;
-#  if HAVE_MBRTOWC
-                          count = mbrtowc (&wc, arg, arg_end - arg, &state);
-#  else
-                          count = mbtowc (&wc, arg, arg_end - arg);
-#  endif
-                          if (count <= 0)
-                            /* mbrtowc not consistent with mbrlen, or mbtowc
-                               not consistent with mblen.  */
-                            abort ();
-                          result[length++] = wc;
-                          arg += count;
-                        }
-                      if (!(arg == arg_end))
-                        abort ();
-                    }
-                  else
-                    {
-#  if HAVE_MBRTOWC
-                      mbstate_t state;
-                      memset (&state, '\0', sizeof (mbstate_t));
-#  endif
-                      while (arg < arg_end)
-                        {
-                          wchar_t wc;
-                          int count;
-#  if HAVE_MBRTOWC
-                          count = mbrtowc (&wc, arg, arg_end - arg, &state);
-#  else
-                          count = mbtowc (&wc, arg, arg_end - arg);
-#  endif
-                          if (count <= 0)
-                            /* mbrtowc not consistent with mbrlen, or mbtowc
-                               not consistent with mblen.  */
-                            abort ();
-                          ENSURE_ALLOCATION (xsum (length, 1));
-                          result[length++] = wc;
-                          arg += count;
-                        }
-                    }
-
-                  if (has_width && width > characters
-                      && (dp->flags & FLAG_LEFT))
-                    {
-                      size_t n = width - characters;
-                      ENSURE_ALLOCATION (xsum (length, n));
-                      DCHAR_SET (result + length, ' ', n);
-                      length += n;
-                    }
-                }
-# else
-                /* %ls in vasnprintf.  See the specification of fprintf.  */
-                {
-                  const wchar_t *arg = a.arg[dp->arg_index].a.a_wide_string;
-                  const wchar_t *arg_end;
-                  size_t characters;
-#  if !DCHAR_IS_TCHAR
-                  /* This code assumes that TCHAR_T is 'char'.  */
-                  typedef int TCHAR_T_verify[2 * (sizeof (TCHAR_T) == 1) - 1];
-                  TCHAR_T *tmpsrc;
-                  DCHAR_T *tmpdst;
-                  size_t tmpdst_len;
-#  endif
-                  size_t w;
-
-                  if (has_precision)
-                    {
-                      /* Use only as many wide characters as needed to produce
-                         at most PRECISION bytes, from the left.  */
-#  if HAVE_WCRTOMB && !defined GNULIB_defined_mbstate_t
-                      mbstate_t state;
-                      memset (&state, '\0', sizeof (mbstate_t));
-#  endif
-                      arg_end = arg;
-                      characters = 0;
-                      while (precision > 0)
-                        {
-                          char cbuf[64]; /* Assume MB_CUR_MAX <= 64.  */
-                          int count;
-
-                          if (*arg_end == 0)
-                            /* Found the terminating null wide character.  */
-                            break;
-#  if HAVE_WCRTOMB && !defined GNULIB_defined_mbstate_t
-                          count = wcrtomb (cbuf, *arg_end, &state);
-#  else
-                          count = wctomb (cbuf, *arg_end);
-#  endif
-                          if (count < 0)
-                            {
-                              /* Cannot convert.  */
-                              if (!(result == resultbuf || result == NULL))
-                                free (result);
-                              if (buf_malloced != NULL)
-                                free (buf_malloced);
-                              CLEANUP ();
-                              errno = EILSEQ;
-                              return NULL;
-                            }
-                          if (precision < count)
-                            break;
-                          arg_end++;
-                          characters += count;
-                          precision -= count;
-                        }
-                    }
-#  if DCHAR_IS_TCHAR
-                  else if (has_width)
-#  else
-                  else
-#  endif
-                    {
-                      /* Use the entire string, and count the number of
-                         bytes.  */
-#  if HAVE_WCRTOMB && !defined GNULIB_defined_mbstate_t
-                      mbstate_t state;
-                      memset (&state, '\0', sizeof (mbstate_t));
-#  endif
-                      arg_end = arg;
-                      characters = 0;
-                      for (;;)
-                        {
-                          char cbuf[64]; /* Assume MB_CUR_MAX <= 64.  */
-                          int count;
-
-                          if (*arg_end == 0)
-                            /* Found the terminating null wide character.  */
-                            break;
-#  if HAVE_WCRTOMB && !defined GNULIB_defined_mbstate_t
-                          count = wcrtomb (cbuf, *arg_end, &state);
-#  else
-                          count = wctomb (cbuf, *arg_end);
-#  endif
-                          if (count < 0)
-                            {
-                              /* Cannot convert.  */
-                              if (!(result == resultbuf || result == NULL))
-                                free (result);
-                              if (buf_malloced != NULL)
-                                free (buf_malloced);
-                              CLEANUP ();
-                              errno = EILSEQ;
-                              return NULL;
-                            }
-                          arg_end++;
-                          characters += count;
-                        }
-                    }
-#  if DCHAR_IS_TCHAR
-                  else
-                    {
-                      /* Use the entire string.  */
-                      arg_end = arg + local_wcslen (arg);
-                      /* The number of bytes doesn't matter.  */
-                      characters = 0;
-                    }
-#  endif
-
-#  if !DCHAR_IS_TCHAR
-                  /* Convert the string into a piece of temporary memory.  */
-                  tmpsrc = (TCHAR_T *) malloc (characters * sizeof (TCHAR_T));
-                  if (tmpsrc == NULL)
-                    goto out_of_memory;
-                  {
-                    TCHAR_T *tmpptr = tmpsrc;
-                    size_t remaining;
-#   if HAVE_WCRTOMB && !defined GNULIB_defined_mbstate_t
-                    mbstate_t state;
-                    memset (&state, '\0', sizeof (mbstate_t));
-#   endif
-                    for (remaining = characters; remaining > 0; )
-                      {
-                        char cbuf[64]; /* Assume MB_CUR_MAX <= 64.  */
-                        int count;
-
-                        if (*arg == 0)
-                          abort ();
-#   if HAVE_WCRTOMB && !defined GNULIB_defined_mbstate_t
-                        count = wcrtomb (cbuf, *arg, &state);
-#   else
-                        count = wctomb (cbuf, *arg);
-#   endif
-                        if (count <= 0)
-                          /* Inconsistency.  */
-                          abort ();
-                        memcpy (tmpptr, cbuf, count);
-                        tmpptr += count;
-                        arg++;
-                        remaining -= count;
-                      }
-                    if (!(arg == arg_end))
-                      abort ();
-                  }
-
-                  /* Convert from TCHAR_T[] to DCHAR_T[].  */
-                  tmpdst =
-                    DCHAR_CONV_FROM_ENCODING (locale_charset (),
-                                              iconveh_question_mark,
-                                              tmpsrc, characters,
-                                              NULL,
-                                              NULL, &tmpdst_len);
-                  if (tmpdst == NULL)
-                    {
-                      int saved_errno = errno;
-                      free (tmpsrc);
-                      if (!(result == resultbuf || result == NULL))
-                        free (result);
-                      if (buf_malloced != NULL)
-                        free (buf_malloced);
-                      CLEANUP ();
-                      errno = saved_errno;
-                      return NULL;
-                    }
-                  free (tmpsrc);
-#  endif
-
-                  if (has_width)
-                    {
-#  if ENABLE_UNISTDIO
-                      /* Outside POSIX, it's preferrable to compare the width
-                         against the number of _characters_ of the converted
-                         value.  */
-                      w = DCHAR_MBSNLEN (result + length, characters);
-#  else
-                      /* The width is compared against the number of _bytes_
-                         of the converted value, says POSIX.  */
-                      w = characters;
-#  endif
-                    }
-                  else
-                    /* w doesn't matter.  */
-                    w = 0;
-
-                  if (has_width && width > w
-                      && !(dp->flags & FLAG_LEFT))
-                    {
-                      size_t n = width - w;
-                      ENSURE_ALLOCATION (xsum (length, n));
-                      DCHAR_SET (result + length, ' ', n);
-                      length += n;
-                    }
-
-#  if DCHAR_IS_TCHAR
-                  if (has_precision || has_width)
-                    {
-                      /* We know the number of bytes in advance.  */
-                      size_t remaining;
-#   if HAVE_WCRTOMB && !defined GNULIB_defined_mbstate_t
-                      mbstate_t state;
-                      memset (&state, '\0', sizeof (mbstate_t));
-#   endif
-                      ENSURE_ALLOCATION (xsum (length, characters));
-                      for (remaining = characters; remaining > 0; )
-                        {
-                          char cbuf[64]; /* Assume MB_CUR_MAX <= 64.  */
-                          int count;
-
-                          if (*arg == 0)
-                            abort ();
-#   if HAVE_WCRTOMB && !defined GNULIB_defined_mbstate_t
-                          count = wcrtomb (cbuf, *arg, &state);
-#   else
-                          count = wctomb (cbuf, *arg);
-#   endif
-                          if (count <= 0)
-                            /* Inconsistency.  */
-                            abort ();
-                          memcpy (result + length, cbuf, count);
-                          length += count;
-                          arg++;
-                          remaining -= count;
-                        }
-                      if (!(arg == arg_end))
-                        abort ();
-                    }
-                  else
-                    {
-#   if HAVE_WCRTOMB && !defined GNULIB_defined_mbstate_t
-                      mbstate_t state;
-                      memset (&state, '\0', sizeof (mbstate_t));
-#   endif
-                      while (arg < arg_end)
-                        {
-                          char cbuf[64]; /* Assume MB_CUR_MAX <= 64.  */
-                          int count;
-
-                          if (*arg == 0)
-                            abort ();
-#   if HAVE_WCRTOMB && !defined GNULIB_defined_mbstate_t
-                          count = wcrtomb (cbuf, *arg, &state);
-#   else
-                          count = wctomb (cbuf, *arg);
-#   endif
-                          if (count <= 0)
-                            {
-                              /* Cannot convert.  */
-                              if (!(result == resultbuf || result == NULL))
-                                free (result);
-                              if (buf_malloced != NULL)
-                                free (buf_malloced);
-                              CLEANUP ();
-                              errno = EILSEQ;
-                              return NULL;
-                            }
-                          ENSURE_ALLOCATION (xsum (length, count));
-                          memcpy (result + length, cbuf, count);
-                          length += count;
-                          arg++;
-                        }
-                    }
-#  else
-                  ENSURE_ALLOCATION (xsum (length, tmpdst_len));
-                  DCHAR_CPY (result + length, tmpdst, tmpdst_len);
-                  free (tmpdst);
-                  length += tmpdst_len;
-#  endif
-
-                  if (has_width && width > w
-                      && (dp->flags & FLAG_LEFT))
-                    {
-                      size_t n = width - w;
-                      ENSURE_ALLOCATION (xsum (length, n));
-                      DCHAR_SET (result + length, ' ', n);
-                      length += n;
-                    }
-                }
-# endif
-              }
-#endif
-#if (NEED_PRINTF_DIRECTIVE_A || NEED_PRINTF_LONG_DOUBLE || NEED_PRINTF_DOUBLE) && !defined IN_LIBINTL
-            else if ((dp->conversion == 'a' || dp->conversion == 'A')
-# if !(NEED_PRINTF_DIRECTIVE_A || (NEED_PRINTF_LONG_DOUBLE && NEED_PRINTF_DOUBLE))
-                     && (0
-#  if NEED_PRINTF_DOUBLE
-                         || a.arg[dp->arg_index].type == TYPE_DOUBLE
-#  endif
-#  if NEED_PRINTF_LONG_DOUBLE
-                         || a.arg[dp->arg_index].type == TYPE_LONGDOUBLE
-#  endif
-                        )
-# endif
-                    )
-              {
-                arg_type type = a.arg[dp->arg_index].type;
-                int flags = dp->flags;
-                int has_width;
-                size_t width;
-                int has_precision;
-                size_t precision;
-                size_t tmp_length;
-                DCHAR_T tmpbuf[700];
-                DCHAR_T *tmp;
-                DCHAR_T *pad_ptr;
-                DCHAR_T *p;
-
-                has_width = 0;
-                width = 0;
-                if (dp->width_start != dp->width_end)
-                  {
-                    if (dp->width_arg_index != ARG_NONE)
-                      {
-                        int arg;
-
-                        if (!(a.arg[dp->width_arg_index].type == TYPE_INT))
-                          abort ();
-                        arg = a.arg[dp->width_arg_index].a.a_int;
-                        if (arg < 0)
-                          {
-                            /* "A negative field width is taken as a '-' flag
-                                followed by a positive field width."  */
-                            flags |= FLAG_LEFT;
-                            width = (unsigned int) (-arg);
-                          }
-                        else
-                          width = arg;
-                      }
-                    else
-                      {
-                        const FCHAR_T *digitp = dp->width_start;
-
-                        do
-                          width = xsum (xtimes (width, 10), *digitp++ - '0');
-                        while (digitp != dp->width_end);
-                      }
-                    has_width = 1;
-                  }
-
-                has_precision = 0;
-                precision = 0;
-                if (dp->precision_start != dp->precision_end)
-                  {
-                    if (dp->precision_arg_index != ARG_NONE)
-                      {
-                        int arg;
-
-                        if (!(a.arg[dp->precision_arg_index].type == TYPE_INT))
-                          abort ();
-                        arg = a.arg[dp->precision_arg_index].a.a_int;
-                        /* "A negative precision is taken as if the precision
-                            were omitted."  */
-                        if (arg >= 0)
-                          {
-                            precision = arg;
-                            has_precision = 1;
-                          }
-                      }
-                    else
-                      {
-                        const FCHAR_T *digitp = dp->precision_start + 1;
-
-                        precision = 0;
-                        while (digitp != dp->precision_end)
-                          precision = xsum (xtimes (precision, 10), *digitp++ - '0');
-                        has_precision = 1;
-                      }
-                  }
-
-                /* Allocate a temporary buffer of sufficient size.  */
-                if (type == TYPE_LONGDOUBLE)
-                  tmp_length =
-                    (unsigned int) ((LDBL_DIG + 1)
-                                    * 0.831 /* decimal -> hexadecimal */
-                                   )
-                    + 1; /* turn floor into ceil */
-                else
-                  tmp_length =
-                    (unsigned int) ((DBL_DIG + 1)
-                                    * 0.831 /* decimal -> hexadecimal */
-                                   )
-                    + 1; /* turn floor into ceil */
-                if (tmp_length < precision)
-                  tmp_length = precision;
-                /* Account for sign, decimal point etc. */
-                tmp_length = xsum (tmp_length, 12);
-
-                if (tmp_length < width)
-                  tmp_length = width;
-
-                tmp_length = xsum (tmp_length, 1); /* account for trailing NUL */
-
-                if (tmp_length <= sizeof (tmpbuf) / sizeof (DCHAR_T))
-                  tmp = tmpbuf;
-                else
-                  {
-                    size_t tmp_memsize = xtimes (tmp_length, sizeof (DCHAR_T));
-
-                    if (size_overflow_p (tmp_memsize))
-                      /* Overflow, would lead to out of memory.  */
-                      goto out_of_memory;
-                    tmp = (DCHAR_T *) malloc (tmp_memsize);
-                    if (tmp == NULL)
-                      /* Out of memory.  */
-                      goto out_of_memory;
-                  }
-
-                pad_ptr = NULL;
-                p = tmp;
-                if (type == TYPE_LONGDOUBLE)
-                  {
-# if NEED_PRINTF_DIRECTIVE_A || NEED_PRINTF_LONG_DOUBLE
-                    long double arg = a.arg[dp->arg_index].a.a_longdouble;
-
-                    if (isnanl (arg))
-                      {
-                        if (dp->conversion == 'A')
-                          {
-                            *p++ = 'N'; *p++ = 'A'; *p++ = 'N';
-                          }
-                        else
-                          {
-                            *p++ = 'n'; *p++ = 'a'; *p++ = 'n';
-                          }
-                      }
-                    else
-                      {
-                        int sign = 0;
-                        DECL_LONG_DOUBLE_ROUNDING
-
-                        BEGIN_LONG_DOUBLE_ROUNDING ();
-
-                        if (signbit (arg)) /* arg < 0.0L or negative zero */
-                          {
-                            sign = -1;
-                            arg = -arg;
-                          }
-
-                        if (sign < 0)
-                          *p++ = '-';
-                        else if (flags & FLAG_SHOWSIGN)
-                          *p++ = '+';
-                        else if (flags & FLAG_SPACE)
-                          *p++ = ' ';
-
-                        if (arg > 0.0L && arg + arg == arg)
-                          {
-                            if (dp->conversion == 'A')
-                              {
-                                *p++ = 'I'; *p++ = 'N'; *p++ = 'F';
-                              }
-                            else
-                              {
-                                *p++ = 'i'; *p++ = 'n'; *p++ = 'f';
-                              }
-                          }
-                        else
-                          {
-                            int exponent;
-                            long double mantissa;
-
-                            if (arg > 0.0L)
-                              mantissa = printf_frexpl (arg, &exponent);
-                            else
-                              {
-                                exponent = 0;
-                                mantissa = 0.0L;
-                              }
-
-                            if (has_precision
-                                && precision < (unsigned int) ((LDBL_DIG + 1) * 0.831) + 1)
-                              {
-                                /* Round the mantissa.  */
-                                long double tail = mantissa;
-                                size_t q;
-
-                                for (q = precision; ; q--)
-                                  {
-                                    int digit = (int) tail;
-                                    tail -= digit;
-                                    if (q == 0)
-                                      {
-                                        if (digit & 1 ? tail >= 0.5L : tail > 0.5L)
-                                          tail = 1 - tail;
-                                        else
-                                          tail = - tail;
-                                        break;
-                                      }
-                                    tail *= 16.0L;
-                                  }
-                                if (tail != 0.0L)
-                                  for (q = precision; q > 0; q--)
-                                    tail *= 0.0625L;
-                                mantissa += tail;
-                              }
-
-                            *p++ = '0';
-                            *p++ = dp->conversion - 'A' + 'X';
-                            pad_ptr = p;
-                            {
-                              int digit;
-
-                              digit = (int) mantissa;
-                              mantissa -= digit;
-                              *p++ = '0' + digit;
-                              if ((flags & FLAG_ALT)
-                                  || mantissa > 0.0L || precision > 0)
-                                {
-                                  *p++ = decimal_point_char ();
-                                  /* This loop terminates because we assume
-                                     that FLT_RADIX is a power of 2.  */
-                                  while (mantissa > 0.0L)
-                                    {
-                                      mantissa *= 16.0L;
-                                      digit = (int) mantissa;
-                                      mantissa -= digit;
-                                      *p++ = digit
-                                             + (digit < 10
-                                                ? '0'
-                                                : dp->conversion - 10);
-                                      if (precision > 0)
-                                        precision--;
-                                    }
-                                  while (precision > 0)
-                                    {
-                                      *p++ = '0';
-                                      precision--;
-                                    }
-                                }
-                              }
-                              *p++ = dp->conversion - 'A' + 'P';
-#  if WIDE_CHAR_VERSION
-                              {
-                                static const wchar_t decimal_format[] =
-                                  { '%', '+', 'd', '\0' };
-                                SNPRINTF (p, 6 + 1, decimal_format, exponent);
-                              }
-                              while (*p != '\0')
-                                p++;
-#  else
-                              if (sizeof (DCHAR_T) == 1)
-                                {
-                                  sprintf ((char *) p, "%+d", exponent);
-                                  while (*p != '\0')
-                                    p++;
-                                }
-                              else
-                                {
-                                  char expbuf[6 + 1];
-                                  const char *ep;
-                                  sprintf (expbuf, "%+d", exponent);
-                                  for (ep = expbuf; (*p = *ep) != '\0'; ep++)
-                                    p++;
-                                }
-#  endif
-                          }
-
-                        END_LONG_DOUBLE_ROUNDING ();
-                      }
-# else
-                    abort ();
-# endif
-                  }
-                else
-                  {
-# if NEED_PRINTF_DIRECTIVE_A || NEED_PRINTF_DOUBLE
-                    double arg = a.arg[dp->arg_index].a.a_double;
-
-                    if (isnand (arg))
-                      {
-                        if (dp->conversion == 'A')
-                          {
-                            *p++ = 'N'; *p++ = 'A'; *p++ = 'N';
-                          }
-                        else
-                          {
-                            *p++ = 'n'; *p++ = 'a'; *p++ = 'n';
-                          }
-                      }
-                    else
-                      {
-                        int sign = 0;
-
-                        if (signbit (arg)) /* arg < 0.0 or negative zero */
-                          {
-                            sign = -1;
-                            arg = -arg;
-                          }
-
-                        if (sign < 0)
-                          *p++ = '-';
-                        else if (flags & FLAG_SHOWSIGN)
-                          *p++ = '+';
-                        else if (flags & FLAG_SPACE)
-                          *p++ = ' ';
-
-                        if (arg > 0.0 && arg + arg == arg)
-                          {
-                            if (dp->conversion == 'A')
-                              {
-                                *p++ = 'I'; *p++ = 'N'; *p++ = 'F';
-                              }
-                            else
-                              {
-                                *p++ = 'i'; *p++ = 'n'; *p++ = 'f';
-                              }
-                          }
-                        else
-                          {
-                            int exponent;
-                            double mantissa;
-
-                            if (arg > 0.0)
-                              mantissa = printf_frexp (arg, &exponent);
-                            else
-                              {
-                                exponent = 0;
-                                mantissa = 0.0;
-                              }
-
-                            if (has_precision
-                                && precision < (unsigned int) ((DBL_DIG + 1) * 0.831) + 1)
-                              {
-                                /* Round the mantissa.  */
-                                double tail = mantissa;
-                                size_t q;
-
-                                for (q = precision; ; q--)
-                                  {
-                                    int digit = (int) tail;
-                                    tail -= digit;
-                                    if (q == 0)
-                                      {
-                                        if (digit & 1 ? tail >= 0.5 : tail > 0.5)
-                                          tail = 1 - tail;
-                                        else
-                                          tail = - tail;
-                                        break;
-                                      }
-                                    tail *= 16.0;
-                                  }
-                                if (tail != 0.0)
-                                  for (q = precision; q > 0; q--)
-                                    tail *= 0.0625;
-                                mantissa += tail;
-                              }
-
-                            *p++ = '0';
-                            *p++ = dp->conversion - 'A' + 'X';
-                            pad_ptr = p;
-                            {
-                              int digit;
-
-                              digit = (int) mantissa;
-                              mantissa -= digit;
-                              *p++ = '0' + digit;
-                              if ((flags & FLAG_ALT)
-                                  || mantissa > 0.0 || precision > 0)
-                                {
-                                  *p++ = decimal_point_char ();
-                                  /* This loop terminates because we assume
-                                     that FLT_RADIX is a power of 2.  */
-                                  while (mantissa > 0.0)
-                                    {
-                                      mantissa *= 16.0;
-                                      digit = (int) mantissa;
-                                      mantissa -= digit;
-                                      *p++ = digit
-                                             + (digit < 10
-                                                ? '0'
-                                                : dp->conversion - 10);
-                                      if (precision > 0)
-                                        precision--;
-                                    }
-                                  while (precision > 0)
-                                    {
-                                      *p++ = '0';
-                                      precision--;
-                                    }
-                                }
-                              }
-                              *p++ = dp->conversion - 'A' + 'P';
-#  if WIDE_CHAR_VERSION
-                              {
-                                static const wchar_t decimal_format[] =
-                                  { '%', '+', 'd', '\0' };
-                                SNPRINTF (p, 6 + 1, decimal_format, exponent);
-                              }
-                              while (*p != '\0')
-                                p++;
-#  else
-                              if (sizeof (DCHAR_T) == 1)
-                                {
-                                  sprintf ((char *) p, "%+d", exponent);
-                                  while (*p != '\0')
-                                    p++;
-                                }
-                              else
-                                {
-                                  char expbuf[6 + 1];
-                                  const char *ep;
-                                  sprintf (expbuf, "%+d", exponent);
-                                  for (ep = expbuf; (*p = *ep) != '\0'; ep++)
-                                    p++;
-                                }
-#  endif
-                          }
-                      }
-# else
-                    abort ();
-# endif
-                  }
-                /* The generated string now extends from tmp to p, with the
-                   zero padding insertion point being at pad_ptr.  */
-                if (has_width && p - tmp < width)
-                  {
-                    size_t pad = width - (p - tmp);
-                    DCHAR_T *end = p + pad;
-
-                    if (flags & FLAG_LEFT)
-                      {
-                        /* Pad with spaces on the right.  */
-                        for (; pad > 0; pad--)
-                          *p++ = ' ';
-                      }
-                    else if ((flags & FLAG_ZERO) && pad_ptr != NULL)
-                      {
-                        /* Pad with zeroes.  */
-                        DCHAR_T *q = end;
-
-                        while (p > pad_ptr)
-                          *--q = *--p;
-                        for (; pad > 0; pad--)
-                          *p++ = '0';
-                      }
-                    else
-                      {
-                        /* Pad with spaces on the left.  */
-                        DCHAR_T *q = end;
-
-                        while (p > tmp)
-                          *--q = *--p;
-                        for (; pad > 0; pad--)
-                          *p++ = ' ';
-                      }
-
-                    p = end;
-                  }
-
-                {
-                  size_t count = p - tmp;
-
-                  if (count >= tmp_length)
-                    /* tmp_length was incorrectly calculated - fix the
-                       code above!  */
-                    abort ();
-
-                  /* Make room for the result.  */
-                  if (count >= allocated - length)
-                    {
-                      size_t n = xsum (length, count);
-
-                      ENSURE_ALLOCATION (n);
-                    }
-
-                  /* Append the result.  */
-                  memcpy (result + length, tmp, count * sizeof (DCHAR_T));
-                  if (tmp != tmpbuf)
-                    free (tmp);
-                  length += count;
-                }
-              }
-#endif
-#if (NEED_PRINTF_INFINITE_DOUBLE || NEED_PRINTF_DOUBLE || NEED_PRINTF_INFINITE_LONG_DOUBLE || NEED_PRINTF_LONG_DOUBLE) && !defined IN_LIBINTL
-            else if ((dp->conversion == 'f' || dp->conversion == 'F'
-                      || dp->conversion == 'e' || dp->conversion == 'E'
-                      || dp->conversion == 'g' || dp->conversion == 'G'
-                      || dp->conversion == 'a' || dp->conversion == 'A')
-                     && (0
-# if NEED_PRINTF_DOUBLE
-                         || a.arg[dp->arg_index].type == TYPE_DOUBLE
-# elif NEED_PRINTF_INFINITE_DOUBLE
-                         || (a.arg[dp->arg_index].type == TYPE_DOUBLE
-                             /* The systems (mingw) which produce wrong output
-                                for Inf, -Inf, and NaN also do so for -0.0.
-                                Therefore we treat this case here as well.  */
-                             && is_infinite_or_zero (a.arg[dp->arg_index].a.a_double))
-# endif
-# if NEED_PRINTF_LONG_DOUBLE
-                         || a.arg[dp->arg_index].type == TYPE_LONGDOUBLE
-# elif NEED_PRINTF_INFINITE_LONG_DOUBLE
-                         || (a.arg[dp->arg_index].type == TYPE_LONGDOUBLE
-                             /* Some systems produce wrong output for Inf,
-                                -Inf, and NaN.  Some systems in this category
-                                (IRIX 5.3) also do so for -0.0.  Therefore we
-                                treat this case here as well.  */
-                             && is_infinite_or_zerol (a.arg[dp->arg_index].a.a_longdouble))
-# endif
-                        ))
-              {
-# if (NEED_PRINTF_DOUBLE || NEED_PRINTF_INFINITE_DOUBLE) && (NEED_PRINTF_LONG_DOUBLE || NEED_PRINTF_INFINITE_LONG_DOUBLE)
-                arg_type type = a.arg[dp->arg_index].type;
-# endif
-                int flags = dp->flags;
-                int has_width;
-                size_t width;
-                int has_precision;
-                size_t precision;
-                size_t tmp_length;
-                DCHAR_T tmpbuf[700];
-                DCHAR_T *tmp;
-                DCHAR_T *pad_ptr;
-                DCHAR_T *p;
-
-                has_width = 0;
-                width = 0;
-                if (dp->width_start != dp->width_end)
-                  {
-                    if (dp->width_arg_index != ARG_NONE)
-                      {
-                        int arg;
-
-                        if (!(a.arg[dp->width_arg_index].type == TYPE_INT))
-                          abort ();
-                        arg = a.arg[dp->width_arg_index].a.a_int;
-                        if (arg < 0)
-                          {
-                            /* "A negative field width is taken as a '-' flag
-                                followed by a positive field width."  */
-                            flags |= FLAG_LEFT;
-                            width = (unsigned int) (-arg);
-                          }
-                        else
-                          width = arg;
-                      }
-                    else
-                      {
-                        const FCHAR_T *digitp = dp->width_start;
-
-                        do
-                          width = xsum (xtimes (width, 10), *digitp++ - '0');
-                        while (digitp != dp->width_end);
-                      }
-                    has_width = 1;
-                  }
-
-                has_precision = 0;
-                precision = 0;
-                if (dp->precision_start != dp->precision_end)
-                  {
-                    if (dp->precision_arg_index != ARG_NONE)
-                      {
-                        int arg;
-
-                        if (!(a.arg[dp->precision_arg_index].type == TYPE_INT))
-                          abort ();
-                        arg = a.arg[dp->precision_arg_index].a.a_int;
-                        /* "A negative precision is taken as if the precision
-                            were omitted."  */
-                        if (arg >= 0)
-                          {
-                            precision = arg;
-                            has_precision = 1;
-                          }
-                      }
-                    else
-                      {
-                        const FCHAR_T *digitp = dp->precision_start + 1;
-
-                        precision = 0;
-                        while (digitp != dp->precision_end)
-                          precision = xsum (xtimes (precision, 10), *digitp++ - '0');
-                        has_precision = 1;
-                      }
-                  }
-
-                /* POSIX specifies the default precision to be 6 for %f, %F,
-                   %e, %E, but not for %g, %G.  Implementations appear to use
-                   the same default precision also for %g, %G.  But for %a, %A,
-                   the default precision is 0.  */
-                if (!has_precision)
-                  if (!(dp->conversion == 'a' || dp->conversion == 'A'))
-                    precision = 6;
-
-                /* Allocate a temporary buffer of sufficient size.  */
-# if NEED_PRINTF_DOUBLE && NEED_PRINTF_LONG_DOUBLE
-                tmp_length = (type == TYPE_LONGDOUBLE ? LDBL_DIG + 1 : DBL_DIG + 1);
-# elif NEED_PRINTF_INFINITE_DOUBLE && NEED_PRINTF_LONG_DOUBLE
-                tmp_length = (type == TYPE_LONGDOUBLE ? LDBL_DIG + 1 : 0);
-# elif NEED_PRINTF_LONG_DOUBLE
-                tmp_length = LDBL_DIG + 1;
-# elif NEED_PRINTF_DOUBLE
-                tmp_length = DBL_DIG + 1;
-# else
-                tmp_length = 0;
-# endif
-                if (tmp_length < precision)
-                  tmp_length = precision;
-# if NEED_PRINTF_LONG_DOUBLE
-#  if NEED_PRINTF_DOUBLE || NEED_PRINTF_INFINITE_DOUBLE
-                if (type == TYPE_LONGDOUBLE)
-#  endif
-                  if (dp->conversion == 'f' || dp->conversion == 'F')
-                    {
-                      long double arg = a.arg[dp->arg_index].a.a_longdouble;
-                      if (!(isnanl (arg) || arg + arg == arg))
-                        {
-                          /* arg is finite and nonzero.  */
-                          int exponent = floorlog10l (arg < 0 ? -arg : arg);
-                          if (exponent >= 0 && tmp_length < exponent + precision)
-                            tmp_length = exponent + precision;
-                        }
-                    }
-# endif
-# if NEED_PRINTF_DOUBLE
-#  if NEED_PRINTF_LONG_DOUBLE || NEED_PRINTF_INFINITE_LONG_DOUBLE
-                if (type == TYPE_DOUBLE)
-#  endif
-                  if (dp->conversion == 'f' || dp->conversion == 'F')
-                    {
-                      double arg = a.arg[dp->arg_index].a.a_double;
-                      if (!(isnand (arg) || arg + arg == arg))
-                        {
-                          /* arg is finite and nonzero.  */
-                          int exponent = floorlog10 (arg < 0 ? -arg : arg);
-                          if (exponent >= 0 && tmp_length < exponent + precision)
-                            tmp_length = exponent + precision;
-                        }
-                    }
-# endif
-                /* Account for sign, decimal point etc. */
-                tmp_length = xsum (tmp_length, 12);
-
-                if (tmp_length < width)
-                  tmp_length = width;
-
-                tmp_length = xsum (tmp_length, 1); /* account for trailing NUL */
-
-                if (tmp_length <= sizeof (tmpbuf) / sizeof (DCHAR_T))
-                  tmp = tmpbuf;
-                else
-                  {
-                    size_t tmp_memsize = xtimes (tmp_length, sizeof (DCHAR_T));
-
-                    if (size_overflow_p (tmp_memsize))
-                      /* Overflow, would lead to out of memory.  */
-                      goto out_of_memory;
-                    tmp = (DCHAR_T *) malloc (tmp_memsize);
-                    if (tmp == NULL)
-                      /* Out of memory.  */
-                      goto out_of_memory;
-                  }
-
-                pad_ptr = NULL;
-                p = tmp;
-
-# if NEED_PRINTF_LONG_DOUBLE || NEED_PRINTF_INFINITE_LONG_DOUBLE
-#  if NEED_PRINTF_DOUBLE || NEED_PRINTF_INFINITE_DOUBLE
-                if (type == TYPE_LONGDOUBLE)
-#  endif
-                  {
-                    long double arg = a.arg[dp->arg_index].a.a_longdouble;
-
-                    if (isnanl (arg))
-                      {
-                        if (dp->conversion >= 'A' && dp->conversion <= 'Z')
-                          {
-                            *p++ = 'N'; *p++ = 'A'; *p++ = 'N';
-                          }
-                        else
-                          {
-                            *p++ = 'n'; *p++ = 'a'; *p++ = 'n';
-                          }
-                      }
-                    else
-                      {
-                        int sign = 0;
-                        DECL_LONG_DOUBLE_ROUNDING
-
-                        BEGIN_LONG_DOUBLE_ROUNDING ();
-
-                        if (signbit (arg)) /* arg < 0.0L or negative zero */
-                          {
-                            sign = -1;
-                            arg = -arg;
-                          }
-
-                        if (sign < 0)
-                          *p++ = '-';
-                        else if (flags & FLAG_SHOWSIGN)
-                          *p++ = '+';
-                        else if (flags & FLAG_SPACE)
-                          *p++ = ' ';
-
-                        if (arg > 0.0L && arg + arg == arg)
-                          {
-                            if (dp->conversion >= 'A' && dp->conversion <= 'Z')
-                              {
-                                *p++ = 'I'; *p++ = 'N'; *p++ = 'F';
-                              }
-                            else
-                              {
-                                *p++ = 'i'; *p++ = 'n'; *p++ = 'f';
-                              }
-                          }
-                        else
-                          {
-#  if NEED_PRINTF_LONG_DOUBLE
-                            pad_ptr = p;
-
-                            if (dp->conversion == 'f' || dp->conversion == 'F')
-                              {
-                                char *digits;
-                                size_t ndigits;
-
-                                digits =
-                                  scale10_round_decimal_long_double (arg, precision);
-                                if (digits == NULL)
-                                  {
-                                    END_LONG_DOUBLE_ROUNDING ();
-                                    goto out_of_memory;
-                                  }
-                                ndigits = strlen (digits);
-
-                                if (ndigits > precision)
-                                  do
-                                    {
-                                      --ndigits;
-                                      *p++ = digits[ndigits];
-                                    }
-                                  while (ndigits > precision);
-                                else
-                                  *p++ = '0';
-                                /* Here ndigits <= precision.  */
-                                if ((flags & FLAG_ALT) || precision > 0)
-                                  {
-                                    *p++ = decimal_point_char ();
-                                    for (; precision > ndigits; precision--)
-                                      *p++ = '0';
-                                    while (ndigits > 0)
-                                      {
-                                        --ndigits;
-                                        *p++ = digits[ndigits];
-                                      }
-                                  }
-
-                                free (digits);
-                              }
-                            else if (dp->conversion == 'e' || dp->conversion == 'E')
-                              {
-                                int exponent;
-
-                                if (arg == 0.0L)
-                                  {
-                                    exponent = 0;
-                                    *p++ = '0';
-                                    if ((flags & FLAG_ALT) || precision > 0)
-                                      {
-                                        *p++ = decimal_point_char ();
-                                        for (; precision > 0; precision--)
-                                          *p++ = '0';
-                                      }
-                                  }
-                                else
-                                  {
-                                    /* arg > 0.0L.  */
-                                    int adjusted;
-                                    char *digits;
-                                    size_t ndigits;
-
-                                    exponent = floorlog10l (arg);
-                                    adjusted = 0;
-                                    for (;;)
-                                      {
-                                        digits =
-                                          scale10_round_decimal_long_double (arg,
-                                                                             (int)precision - exponent);
-                                        if (digits == NULL)
-                                          {
-                                            END_LONG_DOUBLE_ROUNDING ();
-                                            goto out_of_memory;
-                                          }
-                                        ndigits = strlen (digits);
-
-                                        if (ndigits == precision + 1)
-                                          break;
-                                        if (ndigits < precision
-                                            || ndigits > precision + 2)
-                                          /* The exponent was not guessed
-                                             precisely enough.  */
-                                          abort ();
-                                        if (adjusted)
-                                          /* None of two values of exponent is
-                                             the right one.  Prevent an endless
-                                             loop.  */
-                                          abort ();
-                                        free (digits);
-                                        if (ndigits == precision)
-                                          exponent -= 1;
-                                        else
-                                          exponent += 1;
-                                        adjusted = 1;
-                                      }
-                                    /* Here ndigits = precision+1.  */
-                                    if (is_borderline (digits, precision))
-                                      {
-                                        /* Maybe the exponent guess was too high
-                                           and a smaller exponent can be reached
-                                           by turning a 10...0 into 9...9x.  */
-                                        char *digits2 =
-                                          scale10_round_decimal_long_double (arg,
-                                                                             (int)precision - exponent + 1);
-                                        if (digits2 == NULL)
-                                          {
-                                            free (digits);
-                                            END_LONG_DOUBLE_ROUNDING ();
-                                            goto out_of_memory;
-                                          }
-                                        if (strlen (digits2) == precision + 1)
-                                          {
-                                            free (digits);
-                                            digits = digits2;
-                                            exponent -= 1;
-                                          }
-                                        else
-                                          free (digits2);
-                                      }
-                                    /* Here ndigits = precision+1.  */
-
-                                    *p++ = digits[--ndigits];
-                                    if ((flags & FLAG_ALT) || precision > 0)
-                                      {
-                                        *p++ = decimal_point_char ();
-                                        while (ndigits > 0)
-                                          {
-                                            --ndigits;
-                                            *p++ = digits[ndigits];
-                                          }
-                                      }
-
-                                    free (digits);
-                                  }
-
-                                *p++ = dp->conversion; /* 'e' or 'E' */
-#   if WIDE_CHAR_VERSION
-                                {
-                                  static const wchar_t decimal_format[] =
-                                    { '%', '+', '.', '2', 'd', '\0' };
-                                  SNPRINTF (p, 6 + 1, decimal_format, exponent);
-                                }
-                                while (*p != '\0')
-                                  p++;
-#   else
-                                if (sizeof (DCHAR_T) == 1)
-                                  {
-                                    sprintf ((char *) p, "%+.2d", exponent);
-                                    while (*p != '\0')
-                                      p++;
-                                  }
-                                else
-                                  {
-                                    char expbuf[6 + 1];
-                                    const char *ep;
-                                    sprintf (expbuf, "%+.2d", exponent);
-                                    for (ep = expbuf; (*p = *ep) != '\0'; ep++)
-                                      p++;
-                                  }
-#   endif
-                              }
-                            else if (dp->conversion == 'g' || dp->conversion == 'G')
-                              {
-                                if (precision == 0)
-                                  precision = 1;
-                                /* precision >= 1.  */
-
-                                if (arg == 0.0L)
-                                  /* The exponent is 0, >= -4, < precision.
-                                     Use fixed-point notation.  */
-                                  {
-                                    size_t ndigits = precision;
-                                    /* Number of trailing zeroes that have to be
-                                       dropped.  */
-                                    size_t nzeroes =
-                                      (flags & FLAG_ALT ? 0 : precision - 1);
-
-                                    --ndigits;
-                                    *p++ = '0';
-                                    if ((flags & FLAG_ALT) || ndigits > nzeroes)
-                                      {
-                                        *p++ = decimal_point_char ();
-                                        while (ndigits > nzeroes)
-                                          {
-                                            --ndigits;
-                                            *p++ = '0';
-                                          }
-                                      }
-                                  }
-                                else
-                                  {
-                                    /* arg > 0.0L.  */
-                                    int exponent;
-                                    int adjusted;
-                                    char *digits;
-                                    size_t ndigits;
-                                    size_t nzeroes;
-
-                                    exponent = floorlog10l (arg);
-                                    adjusted = 0;
-                                    for (;;)
-                                      {
-                                        digits =
-                                          scale10_round_decimal_long_double (arg,
-                                                                             (int)(precision - 1) - exponent);
-                                        if (digits == NULL)
-                                          {
-                                            END_LONG_DOUBLE_ROUNDING ();
-                                            goto out_of_memory;
-                                          }
-                                        ndigits = strlen (digits);
-
-                                        if (ndigits == precision)
-                                          break;
-                                        if (ndigits < precision - 1
-                                            || ndigits > precision + 1)
-                                          /* The exponent was not guessed
-                                             precisely enough.  */
-                                          abort ();
-                                        if (adjusted)
-                                          /* None of two values of exponent is
-                                             the right one.  Prevent an endless
-                                             loop.  */
-                                          abort ();
-                                        free (digits);
-                                        if (ndigits < precision)
-                                          exponent -= 1;
-                                        else
-                                          exponent += 1;
-                                        adjusted = 1;
-                                      }
-                                    /* Here ndigits = precision.  */
-                                    if (is_borderline (digits, precision - 1))
-                                      {
-                                        /* Maybe the exponent guess was too high
-                                           and a smaller exponent can be reached
-                                           by turning a 10...0 into 9...9x.  */
-                                        char *digits2 =
-                                          scale10_round_decimal_long_double (arg,
-                                                                             (int)(precision - 1) - exponent + 1);
-                                        if (digits2 == NULL)
-                                          {
-                                            free (digits);
-                                            END_LONG_DOUBLE_ROUNDING ();
-                                            goto out_of_memory;
-                                          }
-                                        if (strlen (digits2) == precision)
-                                          {
-                                            free (digits);
-                                            digits = digits2;
-                                            exponent -= 1;
-                                          }
-                                        else
-                                          free (digits2);
-                                      }
-                                    /* Here ndigits = precision.  */
-
-                                    /* Determine the number of trailing zeroes
-                                       that have to be dropped.  */
-                                    nzeroes = 0;
-                                    if ((flags & FLAG_ALT) == 0)
-                                      while (nzeroes < ndigits
-                                             && digits[nzeroes] == '0')
-                                        nzeroes++;
-
-                                    /* The exponent is now determined.  */
-                                    if (exponent >= -4
-                                        && exponent < (long)precision)
-                                      {
-                                        /* Fixed-point notation:
-                                           max(exponent,0)+1 digits, then the
-                                           decimal point, then the remaining
-                                           digits without trailing zeroes.  */
-                                        if (exponent >= 0)
-                                          {
-                                            size_t count = exponent + 1;
-                                            /* Note: count <= precision = ndigits.  */
-                                            for (; count > 0; count--)
-                                              *p++ = digits[--ndigits];
-                                            if ((flags & FLAG_ALT) || ndigits > nzeroes)
-                                              {
-                                                *p++ = decimal_point_char ();
-                                                while (ndigits > nzeroes)
-                                                  {
-                                                    --ndigits;
-                                                    *p++ = digits[ndigits];
-                                                  }
-                                              }
-                                          }
-                                        else
-                                          {
-                                            size_t count = -exponent - 1;
-                                            *p++ = '0';
-                                            *p++ = decimal_point_char ();
-                                            for (; count > 0; count--)
-                                              *p++ = '0';
-                                            while (ndigits > nzeroes)
-                                              {
-                                                --ndigits;
-                                                *p++ = digits[ndigits];
-                                              }
-                                          }
-                                      }
-                                    else
-                                      {
-                                        /* Exponential notation.  */
-                                        *p++ = digits[--ndigits];
-                                        if ((flags & FLAG_ALT) || ndigits > nzeroes)
-                                          {
-                                            *p++ = decimal_point_char ();
-                                            while (ndigits > nzeroes)
-                                              {
-                                                --ndigits;
-                                                *p++ = digits[ndigits];
-                                              }
-                                          }
-                                        *p++ = dp->conversion - 'G' + 'E'; /* 'e' or 'E' */
-#   if WIDE_CHAR_VERSION
-                                        {
-                                          static const wchar_t decimal_format[] =
-                                            { '%', '+', '.', '2', 'd', '\0' };
-                                          SNPRINTF (p, 6 + 1, decimal_format, exponent);
-                                        }
-                                        while (*p != '\0')
-                                          p++;
-#   else
-                                        if (sizeof (DCHAR_T) == 1)
-                                          {
-                                            sprintf ((char *) p, "%+.2d", exponent);
-                                            while (*p != '\0')
-                                              p++;
-                                          }
-                                        else
-                                          {
-                                            char expbuf[6 + 1];
-                                            const char *ep;
-                                            sprintf (expbuf, "%+.2d", exponent);
-                                            for (ep = expbuf; (*p = *ep) != '\0'; ep++)
-                                              p++;
-                                          }
-#   endif
-                                      }
-
-                                    free (digits);
-                                  }
-                              }
-                            else
-                              abort ();
-#  else
-                            /* arg is finite.  */
-                            if (!(arg == 0.0L))
-                              abort ();
-
-                            pad_ptr = p;
-
-                            if (dp->conversion == 'f' || dp->conversion == 'F')
-                              {
-                                *p++ = '0';
-                                if ((flags & FLAG_ALT) || precision > 0)
-                                  {
-                                    *p++ = decimal_point_char ();
-                                    for (; precision > 0; precision--)
-                                      *p++ = '0';
-                                  }
-                              }
-                            else if (dp->conversion == 'e' || dp->conversion == 'E')
-                              {
-                                *p++ = '0';
-                                if ((flags & FLAG_ALT) || precision > 0)
-                                  {
-                                    *p++ = decimal_point_char ();
-                                    for (; precision > 0; precision--)
-                                      *p++ = '0';
-                                  }
-                                *p++ = dp->conversion; /* 'e' or 'E' */
-                                *p++ = '+';
-                                *p++ = '0';
-                                *p++ = '0';
-                              }
-                            else if (dp->conversion == 'g' || dp->conversion == 'G')
-                              {
-                                *p++ = '0';
-                                if (flags & FLAG_ALT)
-                                  {
-                                    size_t ndigits =
-                                      (precision > 0 ? precision - 1 : 0);
-                                    *p++ = decimal_point_char ();
-                                    for (; ndigits > 0; --ndigits)
-                                      *p++ = '0';
-                                  }
-                              }
-                            else if (dp->conversion == 'a' || dp->conversion == 'A')
-                              {
-                                *p++ = '0';
-                                *p++ = dp->conversion - 'A' + 'X';
-                                pad_ptr = p;
-                                *p++ = '0';
-                                if ((flags & FLAG_ALT) || precision > 0)
-                                  {
-                                    *p++ = decimal_point_char ();
-                                    for (; precision > 0; precision--)
-                                      *p++ = '0';
-                                  }
-                                *p++ = dp->conversion - 'A' + 'P';
-                                *p++ = '+';
-                                *p++ = '0';
-                              }
-                            else
-                              abort ();
-#  endif
-                          }
-
-                        END_LONG_DOUBLE_ROUNDING ();
-                      }
-                  }
-#  if NEED_PRINTF_DOUBLE || NEED_PRINTF_INFINITE_DOUBLE
-                else
-#  endif
-# endif
-# if NEED_PRINTF_DOUBLE || NEED_PRINTF_INFINITE_DOUBLE
-                  {
-                    double arg = a.arg[dp->arg_index].a.a_double;
-
-                    if (isnand (arg))
-                      {
-                        if (dp->conversion >= 'A' && dp->conversion <= 'Z')
-                          {
-                            *p++ = 'N'; *p++ = 'A'; *p++ = 'N';
-                          }
-                        else
-                          {
-                            *p++ = 'n'; *p++ = 'a'; *p++ = 'n';
-                          }
-                      }
-                    else
-                      {
-                        int sign = 0;
-
-                        if (signbit (arg)) /* arg < 0.0 or negative zero */
-                          {
-                            sign = -1;
-                            arg = -arg;
-                          }
-
-                        if (sign < 0)
-                          *p++ = '-';
-                        else if (flags & FLAG_SHOWSIGN)
-                          *p++ = '+';
-                        else if (flags & FLAG_SPACE)
-                          *p++ = ' ';
-
-                        if (arg > 0.0 && arg + arg == arg)
-                          {
-                            if (dp->conversion >= 'A' && dp->conversion <= 'Z')
-                              {
-                                *p++ = 'I'; *p++ = 'N'; *p++ = 'F';
-                              }
-                            else
-                              {
-                                *p++ = 'i'; *p++ = 'n'; *p++ = 'f';
-                              }
-                          }
-                        else
-                          {
-#  if NEED_PRINTF_DOUBLE
-                            pad_ptr = p;
-
-                            if (dp->conversion == 'f' || dp->conversion == 'F')
-                              {
-                                char *digits;
-                                size_t ndigits;
-
-                                digits =
-                                  scale10_round_decimal_double (arg, precision);
-                                if (digits == NULL)
-                                  goto out_of_memory;
-                                ndigits = strlen (digits);
-
-                                if (ndigits > precision)
-                                  do
-                                    {
-                                      --ndigits;
-                                      *p++ = digits[ndigits];
-                                    }
-                                  while (ndigits > precision);
-                                else
-                                  *p++ = '0';
-                                /* Here ndigits <= precision.  */
-                                if ((flags & FLAG_ALT) || precision > 0)
-                                  {
-                                    *p++ = decimal_point_char ();
-                                    for (; precision > ndigits; precision--)
-                                      *p++ = '0';
-                                    while (ndigits > 0)
-                                      {
-                                        --ndigits;
-                                        *p++ = digits[ndigits];
-                                      }
-                                  }
-
-                                free (digits);
-                              }
-                            else if (dp->conversion == 'e' || dp->conversion == 'E')
-                              {
-                                int exponent;
-
-                                if (arg == 0.0)
-                                  {
-                                    exponent = 0;
-                                    *p++ = '0';
-                                    if ((flags & FLAG_ALT) || precision > 0)
-                                      {
-                                        *p++ = decimal_point_char ();
-                                        for (; precision > 0; precision--)
-                                          *p++ = '0';
-                                      }
-                                  }
-                                else
-                                  {
-                                    /* arg > 0.0.  */
-                                    int adjusted;
-                                    char *digits;
-                                    size_t ndigits;
-
-                                    exponent = floorlog10 (arg);
-                                    adjusted = 0;
-                                    for (;;)
-                                      {
-                                        digits =
-                                          scale10_round_decimal_double (arg,
-                                                                        (int)precision - exponent);
-                                        if (digits == NULL)
-                                          goto out_of_memory;
-                                        ndigits = strlen (digits);
-
-                                        if (ndigits == precision + 1)
-                                          break;
-                                        if (ndigits < precision
-                                            || ndigits > precision + 2)
-                                          /* The exponent was not guessed
-                                             precisely enough.  */
-                                          abort ();
-                                        if (adjusted)
-                                          /* None of two values of exponent is
-                                             the right one.  Prevent an endless
-                                             loop.  */
-                                          abort ();
-                                        free (digits);
-                                        if (ndigits == precision)
-                                          exponent -= 1;
-                                        else
-                                          exponent += 1;
-                                        adjusted = 1;
-                                      }
-                                    /* Here ndigits = precision+1.  */
-                                    if (is_borderline (digits, precision))
-                                      {
-                                        /* Maybe the exponent guess was too high
-                                           and a smaller exponent can be reached
-                                           by turning a 10...0 into 9...9x.  */
-                                        char *digits2 =
-                                          scale10_round_decimal_double (arg,
-                                                                        (int)precision - exponent + 1);
-                                        if (digits2 == NULL)
-                                          {
-                                            free (digits);
-                                            goto out_of_memory;
-                                          }
-                                        if (strlen (digits2) == precision + 1)
-                                          {
-                                            free (digits);
-                                            digits = digits2;
-                                            exponent -= 1;
-                                          }
-                                        else
-                                          free (digits2);
-                                      }
-                                    /* Here ndigits = precision+1.  */
-
-                                    *p++ = digits[--ndigits];
-                                    if ((flags & FLAG_ALT) || precision > 0)
-                                      {
-                                        *p++ = decimal_point_char ();
-                                        while (ndigits > 0)
-                                          {
-                                            --ndigits;
-                                            *p++ = digits[ndigits];
-                                          }
-                                      }
-
-                                    free (digits);
-                                  }
-
-                                *p++ = dp->conversion; /* 'e' or 'E' */
-#   if WIDE_CHAR_VERSION
-                                {
-                                  static const wchar_t decimal_format[] =
-                                    /* Produce the same number of exponent digits
-                                       as the native printf implementation.  */
-#    if (defined _WIN32 || defined __WIN32__) && ! defined __CYGWIN__
-                                    { '%', '+', '.', '3', 'd', '\0' };
-#    else
-                                    { '%', '+', '.', '2', 'd', '\0' };
-#    endif
-                                  SNPRINTF (p, 6 + 1, decimal_format, exponent);
-                                }
-                                while (*p != '\0')
-                                  p++;
-#   else
-                                {
-                                  static const char decimal_format[] =
-                                    /* Produce the same number of exponent digits
-                                       as the native printf implementation.  */
-#    if (defined _WIN32 || defined __WIN32__) && ! defined __CYGWIN__
-                                    "%+.3d";
-#    else
-                                    "%+.2d";
-#    endif
-                                  if (sizeof (DCHAR_T) == 1)
-                                    {
-                                      sprintf ((char *) p, decimal_format, exponent);
-                                      while (*p != '\0')
-                                        p++;
-                                    }
-                                  else
-                                    {
-                                      char expbuf[6 + 1];
-                                      const char *ep;
-                                      sprintf (expbuf, decimal_format, exponent);
-                                      for (ep = expbuf; (*p = *ep) != '\0'; ep++)
-                                        p++;
-                                    }
-                                }
-#   endif
-                              }
-                            else if (dp->conversion == 'g' || dp->conversion == 'G')
-                              {
-                                if (precision == 0)
-                                  precision = 1;
-                                /* precision >= 1.  */
-
-                                if (arg == 0.0)
-                                  /* The exponent is 0, >= -4, < precision.
-                                     Use fixed-point notation.  */
-                                  {
-                                    size_t ndigits = precision;
-                                    /* Number of trailing zeroes that have to be
-                                       dropped.  */
-                                    size_t nzeroes =
-                                      (flags & FLAG_ALT ? 0 : precision - 1);
-
-                                    --ndigits;
-                                    *p++ = '0';
-                                    if ((flags & FLAG_ALT) || ndigits > nzeroes)
-                                      {
-                                        *p++ = decimal_point_char ();
-                                        while (ndigits > nzeroes)
-                                          {
-                                            --ndigits;
-                                            *p++ = '0';
-                                          }
-                                      }
-                                  }
-                                else
-                                  {
-                                    /* arg > 0.0.  */
-                                    int exponent;
-                                    int adjusted;
-                                    char *digits;
-                                    size_t ndigits;
-                                    size_t nzeroes;
-
-                                    exponent = floorlog10 (arg);
-                                    adjusted = 0;
-                                    for (;;)
-                                      {
-                                        digits =
-                                          scale10_round_decimal_double (arg,
-                                                                        (int)(precision - 1) - exponent);
-                                        if (digits == NULL)
-                                          goto out_of_memory;
-                                        ndigits = strlen (digits);
-
-                                        if (ndigits == precision)
-                                          break;
-                                        if (ndigits < precision - 1
-                                            || ndigits > precision + 1)
-                                          /* The exponent was not guessed
-                                             precisely enough.  */
-                                          abort ();
-                                        if (adjusted)
-                                          /* None of two values of exponent is
-                                             the right one.  Prevent an endless
-                                             loop.  */
-                                          abort ();
-                                        free (digits);
-                                        if (ndigits < precision)
-                                          exponent -= 1;
-                                        else
-                                          exponent += 1;
-                                        adjusted = 1;
-                                      }
-                                    /* Here ndigits = precision.  */
-                                    if (is_borderline (digits, precision - 1))
-                                      {
-                                        /* Maybe the exponent guess was too high
-                                           and a smaller exponent can be reached
-                                           by turning a 10...0 into 9...9x.  */
-                                        char *digits2 =
-                                          scale10_round_decimal_double (arg,
-                                                                        (int)(precision - 1) - exponent + 1);
-                                        if (digits2 == NULL)
-                                          {
-                                            free (digits);
-                                            goto out_of_memory;
-                                          }
-                                        if (strlen (digits2) == precision)
-                                          {
-                                            free (digits);
-                                            digits = digits2;
-                                            exponent -= 1;
-                                          }
-                                        else
-                                          free (digits2);
-                                      }
-                                    /* Here ndigits = precision.  */
-
-                                    /* Determine the number of trailing zeroes
-                                       that have to be dropped.  */
-                                    nzeroes = 0;
-                                    if ((flags & FLAG_ALT) == 0)
-                                      while (nzeroes < ndigits
-                                             && digits[nzeroes] == '0')
-                                        nzeroes++;
-
-                                    /* The exponent is now determined.  */
-                                    if (exponent >= -4
-                                        && exponent < (long)precision)
-                                      {
-                                        /* Fixed-point notation:
-                                           max(exponent,0)+1 digits, then the
-                                           decimal point, then the remaining
-                                           digits without trailing zeroes.  */
-                                        if (exponent >= 0)
-                                          {
-                                            size_t count = exponent + 1;
-                                            /* Note: count <= precision = ndigits.  */
-                                            for (; count > 0; count--)
-                                              *p++ = digits[--ndigits];
-                                            if ((flags & FLAG_ALT) || ndigits > nzeroes)
-                                              {
-                                                *p++ = decimal_point_char ();
-                                                while (ndigits > nzeroes)
-                                                  {
-                                                    --ndigits;
-                                                    *p++ = digits[ndigits];
-                                                  }
-                                              }
-                                          }
-                                        else
-                                          {
-                                            size_t count = -exponent - 1;
-                                            *p++ = '0';
-                                            *p++ = decimal_point_char ();
-                                            for (; count > 0; count--)
-                                              *p++ = '0';
-                                            while (ndigits > nzeroes)
-                                              {
-                                                --ndigits;
-                                                *p++ = digits[ndigits];
-                                              }
-                                          }
-                                      }
-                                    else
-                                      {
-                                        /* Exponential notation.  */
-                                        *p++ = digits[--ndigits];
-                                        if ((flags & FLAG_ALT) || ndigits > nzeroes)
-                                          {
-                                            *p++ = decimal_point_char ();
-                                            while (ndigits > nzeroes)
-                                              {
-                                                --ndigits;
-                                                *p++ = digits[ndigits];
-                                              }
-                                          }
-                                        *p++ = dp->conversion - 'G' + 'E'; /* 'e' or 'E' */
-#   if WIDE_CHAR_VERSION
-                                        {
-                                          static const wchar_t decimal_format[] =
-                                            /* Produce the same number of exponent digits
-                                               as the native printf implementation.  */
-#    if (defined _WIN32 || defined __WIN32__) && ! defined __CYGWIN__
-                                            { '%', '+', '.', '3', 'd', '\0' };
-#    else
-                                            { '%', '+', '.', '2', 'd', '\0' };
-#    endif
-                                          SNPRINTF (p, 6 + 1, decimal_format, exponent);
-                                        }
-                                        while (*p != '\0')
-                                          p++;
-#   else
-                                        {
-                                          static const char decimal_format[] =
-                                            /* Produce the same number of exponent digits
-                                               as the native printf implementation.  */
-#    if (defined _WIN32 || defined __WIN32__) && ! defined __CYGWIN__
-                                            "%+.3d";
-#    else
-                                            "%+.2d";
-#    endif
-                                          if (sizeof (DCHAR_T) == 1)
-                                            {
-                                              sprintf ((char *) p, decimal_format, exponent);
-                                              while (*p != '\0')
-                                                p++;
-                                            }
-                                          else
-                                            {
-                                              char expbuf[6 + 1];
-                                              const char *ep;
-                                              sprintf (expbuf, decimal_format, exponent);
-                                              for (ep = expbuf; (*p = *ep) != '\0'; ep++)
-                                                p++;
-                                            }
-                                        }
-#   endif
-                                      }
-
-                                    free (digits);
-                                  }
-                              }
-                            else
-                              abort ();
-#  else
-                            /* arg is finite.  */
-                            if (!(arg == 0.0))
-                              abort ();
-
-                            pad_ptr = p;
-
-                            if (dp->conversion == 'f' || dp->conversion == 'F')
-                              {
-                                *p++ = '0';
-                                if ((flags & FLAG_ALT) || precision > 0)
-                                  {
-                                    *p++ = decimal_point_char ();
-                                    for (; precision > 0; precision--)
-                                      *p++ = '0';
-                                  }
-                              }
-                            else if (dp->conversion == 'e' || dp->conversion == 'E')
-                              {
-                                *p++ = '0';
-                                if ((flags & FLAG_ALT) || precision > 0)
-                                  {
-                                    *p++ = decimal_point_char ();
-                                    for (; precision > 0; precision--)
-                                      *p++ = '0';
-                                  }
-                                *p++ = dp->conversion; /* 'e' or 'E' */
-                                *p++ = '+';
-                                /* Produce the same number of exponent digits as
-                                   the native printf implementation.  */
-#   if (defined _WIN32 || defined __WIN32__) && ! defined __CYGWIN__
-                                *p++ = '0';
-#   endif
-                                *p++ = '0';
-                                *p++ = '0';
-                              }
-                            else if (dp->conversion == 'g' || dp->conversion == 'G')
-                              {
-                                *p++ = '0';
-                                if (flags & FLAG_ALT)
-                                  {
-                                    size_t ndigits =
-                                      (precision > 0 ? precision - 1 : 0);
-                                    *p++ = decimal_point_char ();
-                                    for (; ndigits > 0; --ndigits)
-                                      *p++ = '0';
-                                  }
-                              }
-                            else
-                              abort ();
-#  endif
-                          }
-                      }
-                  }
-# endif
-
-                /* The generated string now extends from tmp to p, with the
-                   zero padding insertion point being at pad_ptr.  */
-                if (has_width && p - tmp < width)
-                  {
-                    size_t pad = width - (p - tmp);
-                    DCHAR_T *end = p + pad;
-
-                    if (flags & FLAG_LEFT)
-                      {
-                        /* Pad with spaces on the right.  */
-                        for (; pad > 0; pad--)
-                          *p++ = ' ';
-                      }
-                    else if ((flags & FLAG_ZERO) && pad_ptr != NULL)
-                      {
-                        /* Pad with zeroes.  */
-                        DCHAR_T *q = end;
-
-                        while (p > pad_ptr)
-                          *--q = *--p;
-                        for (; pad > 0; pad--)
-                          *p++ = '0';
-                      }
-                    else
-                      {
-                        /* Pad with spaces on the left.  */
-                        DCHAR_T *q = end;
-
-                        while (p > tmp)
-                          *--q = *--p;
-                        for (; pad > 0; pad--)
-                          *p++ = ' ';
-                      }
-
-                    p = end;
-                  }
-
-                {
-                  size_t count = p - tmp;
-
-                  if (count >= tmp_length)
-                    /* tmp_length was incorrectly calculated - fix the
-                       code above!  */
-                    abort ();
-
-                  /* Make room for the result.  */
-                  if (count >= allocated - length)
-                    {
-                      size_t n = xsum (length, count);
-
-                      ENSURE_ALLOCATION (n);
-                    }
-
-                  /* Append the result.  */
-                  memcpy (result + length, tmp, count * sizeof (DCHAR_T));
-                  if (tmp != tmpbuf)
-                    free (tmp);
-                  length += count;
-                }
-              }
-#endif
-            else
-              {
-                arg_type type = a.arg[dp->arg_index].type;
-                int flags = dp->flags;
-#if !USE_SNPRINTF || !HAVE_SNPRINTF_RETVAL_C99 || !DCHAR_IS_TCHAR || ENABLE_UNISTDIO || NEED_PRINTF_FLAG_LEFTADJUST || NEED_PRINTF_FLAG_ZERO || NEED_PRINTF_UNBOUNDED_PRECISION
-                int has_width;
-                size_t width;
-#endif
-#if !USE_SNPRINTF || !HAVE_SNPRINTF_RETVAL_C99 || NEED_PRINTF_UNBOUNDED_PRECISION
-                int has_precision;
-                size_t precision;
-#endif
-#if NEED_PRINTF_UNBOUNDED_PRECISION
-                int prec_ourselves;
-#else
-#               define prec_ourselves 0
-#endif
-#if NEED_PRINTF_FLAG_LEFTADJUST
-#               define pad_ourselves 1
-#elif !DCHAR_IS_TCHAR || ENABLE_UNISTDIO || NEED_PRINTF_FLAG_ZERO || NEED_PRINTF_UNBOUNDED_PRECISION
-                int pad_ourselves;
-#else
-#               define pad_ourselves 0
-#endif
-                TCHAR_T *fbp;
-                unsigned int prefix_count;
-                int prefixes[2] IF_LINT (= { 0 });
-#if !USE_SNPRINTF
-                size_t tmp_length;
-                TCHAR_T tmpbuf[700];
-                TCHAR_T *tmp;
-#endif
-
-#if !USE_SNPRINTF || !HAVE_SNPRINTF_RETVAL_C99 || !DCHAR_IS_TCHAR || ENABLE_UNISTDIO || NEED_PRINTF_FLAG_LEFTADJUST || NEED_PRINTF_FLAG_ZERO || NEED_PRINTF_UNBOUNDED_PRECISION
-                has_width = 0;
-                width = 0;
-                if (dp->width_start != dp->width_end)
-                  {
-                    if (dp->width_arg_index != ARG_NONE)
-                      {
-                        int arg;
-
-                        if (!(a.arg[dp->width_arg_index].type == TYPE_INT))
-                          abort ();
-                        arg = a.arg[dp->width_arg_index].a.a_int;
-                        if (arg < 0)
-                          {
-                            /* "A negative field width is taken as a '-' flag
-                                followed by a positive field width."  */
-                            flags |= FLAG_LEFT;
-                            width = (unsigned int) (-arg);
-                          }
-                        else
-                          width = arg;
-                      }
-                    else
-                      {
-                        const FCHAR_T *digitp = dp->width_start;
-
-                        do
-                          width = xsum (xtimes (width, 10), *digitp++ - '0');
-                        while (digitp != dp->width_end);
-                      }
-                    has_width = 1;
-                  }
-#endif
-
-#if !USE_SNPRINTF || !HAVE_SNPRINTF_RETVAL_C99 || NEED_PRINTF_UNBOUNDED_PRECISION
-                has_precision = 0;
-                precision = 6;
-                if (dp->precision_start != dp->precision_end)
-                  {
-                    if (dp->precision_arg_index != ARG_NONE)
-                      {
-                        int arg;
-
-                        if (!(a.arg[dp->precision_arg_index].type == TYPE_INT))
-                          abort ();
-                        arg = a.arg[dp->precision_arg_index].a.a_int;
-                        /* "A negative precision is taken as if the precision
-                            were omitted."  */
-                        if (arg >= 0)
-                          {
-                            precision = arg;
-                            has_precision = 1;
-                          }
-                      }
-                    else
-                      {
-                        const FCHAR_T *digitp = dp->precision_start + 1;
-
-                        precision = 0;
-                        while (digitp != dp->precision_end)
-                          precision = xsum (xtimes (precision, 10), *digitp++ - '0');
-                        has_precision = 1;
-                      }
-                  }
-#endif
-
-                /* Decide whether to handle the precision ourselves.  */
-#if NEED_PRINTF_UNBOUNDED_PRECISION
-                switch (dp->conversion)
-                  {
-                  case 'd': case 'i': case 'u':
-                  case 'o':
-                  case 'x': case 'X': case 'p':
-                    prec_ourselves = has_precision && (precision > 0);
-                    break;
-                  default:
-                    prec_ourselves = 0;
-                    break;
-                  }
-#endif
-
-                /* Decide whether to perform the padding ourselves.  */
-#if !NEED_PRINTF_FLAG_LEFTADJUST && (!DCHAR_IS_TCHAR || ENABLE_UNISTDIO || NEED_PRINTF_FLAG_ZERO || NEED_PRINTF_UNBOUNDED_PRECISION)
-                switch (dp->conversion)
-                  {
-# if !DCHAR_IS_TCHAR || ENABLE_UNISTDIO
-                  /* If we need conversion from TCHAR_T[] to DCHAR_T[], we need
-                     to perform the padding after this conversion.  Functions
-                     with unistdio extensions perform the padding based on
-                     character count rather than element count.  */
-                  case 'c': case 's':
-# endif
-# if NEED_PRINTF_FLAG_ZERO
-                  case 'f': case 'F': case 'e': case 'E': case 'g': case 'G':
-                  case 'a': case 'A':
-# endif
-                    pad_ourselves = 1;
-                    break;
-                  default:
-                    pad_ourselves = prec_ourselves;
-                    break;
-                  }
-#endif
-
-#if !USE_SNPRINTF
-                /* Allocate a temporary buffer of sufficient size for calling
-                   sprintf.  */
-                tmp_length =
-                  MAX_ROOM_NEEDED (&a, dp->arg_index, dp->conversion, type,
-                                   flags, width, has_precision, precision,
-                                   pad_ourselves);
-
-                if (tmp_length <= sizeof (tmpbuf) / sizeof (TCHAR_T))
-                  tmp = tmpbuf;
-                else
-                  {
-                    size_t tmp_memsize = xtimes (tmp_length, sizeof (TCHAR_T));
-
-                    if (size_overflow_p (tmp_memsize))
-                      /* Overflow, would lead to out of memory.  */
-                      goto out_of_memory;
-                    tmp = (TCHAR_T *) malloc (tmp_memsize);
-                    if (tmp == NULL)
-                      /* Out of memory.  */
-                      goto out_of_memory;
-                  }
-#endif
-
-                /* Construct the format string for calling snprintf or
-                   sprintf.  */
-                fbp = buf;
-                *fbp++ = '%';
-#if NEED_PRINTF_FLAG_GROUPING
-                /* The underlying implementation doesn't support the ' flag.
-                   Produce no grouping characters in this case; this is
-                   acceptable because the grouping is locale dependent.  */
-#else
-                if (flags & FLAG_GROUP)
-                  *fbp++ = '\'';
-#endif
-                if (flags & FLAG_LEFT)
-                  *fbp++ = '-';
-                if (flags & FLAG_SHOWSIGN)
-                  *fbp++ = '+';
-                if (flags & FLAG_SPACE)
-                  *fbp++ = ' ';
-                if (flags & FLAG_ALT)
-                  *fbp++ = '#';
-                if (!pad_ourselves)
-                  {
-                    if (flags & FLAG_ZERO)
-                      *fbp++ = '0';
-                    if (dp->width_start != dp->width_end)
-                      {
-                        size_t n = dp->width_end - dp->width_start;
-                        /* The width specification is known to consist only
-                           of standard ASCII characters.  */
-                        if (sizeof (FCHAR_T) == sizeof (TCHAR_T))
-                          {
-                            memcpy (fbp, dp->width_start, n * sizeof (TCHAR_T));
-                            fbp += n;
-                          }
-                        else
-                          {
-                            const FCHAR_T *mp = dp->width_start;
-                            do
-                              *fbp++ = (unsigned char) *mp++;
-                            while (--n > 0);
-                          }
-                      }
-                  }
-                if (!prec_ourselves)
-                  {
-                    if (dp->precision_start != dp->precision_end)
-                      {
-                        size_t n = dp->precision_end - dp->precision_start;
-                        /* The precision specification is known to consist only
-                           of standard ASCII characters.  */
-                        if (sizeof (FCHAR_T) == sizeof (TCHAR_T))
-                          {
-                            memcpy (fbp, dp->precision_start, n * sizeof (TCHAR_T));
-                            fbp += n;
-                          }
-                        else
-                          {
-                            const FCHAR_T *mp = dp->precision_start;
-                            do
-                              *fbp++ = (unsigned char) *mp++;
-                            while (--n > 0);
-                          }
-                      }
-                  }
-
-                switch (type)
-                  {
-#if HAVE_LONG_LONG_INT
-                  case TYPE_LONGLONGINT:
-                  case TYPE_ULONGLONGINT:
-# if (defined _WIN32 || defined __WIN32__) && ! defined __CYGWIN__
-                    *fbp++ = 'I';
-                    *fbp++ = '6';
-                    *fbp++ = '4';
-                    break;
-# else
-                    *fbp++ = 'l';
-                    /*FALLTHROUGH*/
-# endif
-#endif
-                  case TYPE_LONGINT:
-                  case TYPE_ULONGINT:
-#if HAVE_WINT_T
-                  case TYPE_WIDE_CHAR:
-#endif
-#if HAVE_WCHAR_T
-                  case TYPE_WIDE_STRING:
-#endif
-                    *fbp++ = 'l';
-                    break;
-                  case TYPE_LONGDOUBLE:
-                    *fbp++ = 'L';
-                    break;
-                  default:
-                    break;
-                  }
-#if NEED_PRINTF_DIRECTIVE_F
-                if (dp->conversion == 'F')
-                  *fbp = 'f';
-                else
-#endif
-                  *fbp = dp->conversion;
-#if USE_SNPRINTF
-# if !(__GLIBC__ > 2 || (__GLIBC__ == 2 && __GLIBC_MINOR__ >= 3) || ((defined _WIN32 || defined __WIN32__) && ! defined __CYGWIN__))
-                fbp[1] = '%';
-                fbp[2] = 'n';
-                fbp[3] = '\0';
-# else
-                /* On glibc2 systems from glibc >= 2.3 - probably also older
-                   ones - we know that snprintf's returns value conforms to
-                   ISO C 99: the gl_SNPRINTF_DIRECTIVE_N test passes.
-                   Therefore we can avoid using %n in this situation.
-                   On glibc2 systems from 2004-10-18 or newer, the use of %n
-                   in format strings in writable memory may crash the program
-                   (if compiled with _FORTIFY_SOURCE=2), so we should avoid it
-                   in this situation.  */
-                /* On native Win32 systems (such as mingw), we can avoid using
-                   %n because:
-                     - Although the gl_SNPRINTF_TRUNCATION_C99 test fails,
-                       snprintf does not write more than the specified number
-                       of bytes. (snprintf (buf, 3, "%d %d", 4567, 89) writes
-                       '4', '5', '6' into buf, not '4', '5', '\0'.)
-                     - Although the gl_SNPRINTF_RETVAL_C99 test fails, snprintf
-                       allows us to recognize the case of an insufficient
-                       buffer size: it returns -1 in this case.
-                   On native Win32 systems (such as mingw) where the OS is
-                   Windows Vista, the use of %n in format strings by default
-                   crashes the program. See
-                     <http://gcc.gnu.org/ml/gcc/2007-06/msg00122.html> and
-                     <http://msdn2.microsoft.com/en-us/library/ms175782(VS.80).aspx>
-                   So we should avoid %n in this situation.  */
-                fbp[1] = '\0';
-# endif
-#else
-                fbp[1] = '\0';
-#endif
-
-                /* Construct the arguments for calling snprintf or sprintf.  */
-                prefix_count = 0;
-                if (!pad_ourselves && dp->width_arg_index != ARG_NONE)
-                  {
-                    if (!(a.arg[dp->width_arg_index].type == TYPE_INT))
-                      abort ();
-                    prefixes[prefix_count++] = a.arg[dp->width_arg_index].a.a_int;
-                  }
-                if (!prec_ourselves && dp->precision_arg_index != ARG_NONE)
-                  {
-                    if (!(a.arg[dp->precision_arg_index].type == TYPE_INT))
-                      abort ();
-                    prefixes[prefix_count++] = a.arg[dp->precision_arg_index].a.a_int;
-                  }
-
-#if USE_SNPRINTF
-                /* The SNPRINTF result is appended after result[0..length].
-                   The latter is an array of DCHAR_T; SNPRINTF appends an
-                   array of TCHAR_T to it.  This is possible because
-                   sizeof (TCHAR_T) divides sizeof (DCHAR_T) and
-                   alignof (TCHAR_T) <= alignof (DCHAR_T).  */
-# define TCHARS_PER_DCHAR (sizeof (DCHAR_T) / sizeof (TCHAR_T))
-                /* Ensure that maxlen below will be >= 2.  Needed on BeOS,
-                   where an snprintf() with maxlen==1 acts like sprintf().  */
-                ENSURE_ALLOCATION (xsum (length,
-                                         (2 + TCHARS_PER_DCHAR - 1)
-                                         / TCHARS_PER_DCHAR));
-                /* Prepare checking whether snprintf returns the count
-                   via %n.  */
-                *(TCHAR_T *) (result + length) = '\0';
-#endif
-
-                for (;;)
-                  {
-                    int count = -1;
-
-#if USE_SNPRINTF
-                    int retcount = 0;
-                    size_t maxlen = allocated - length;
-                    /* SNPRINTF can fail if its second argument is
-                       > INT_MAX.  */
-                    if (maxlen > INT_MAX / TCHARS_PER_DCHAR)
-                      maxlen = INT_MAX / TCHARS_PER_DCHAR;
-                    maxlen = maxlen * TCHARS_PER_DCHAR;
-# define SNPRINTF_BUF(arg) \
-                    switch (prefix_count)                                   \
-                      {                                                     \
-                      case 0:                                               \
-                        retcount = SNPRINTF ((TCHAR_T *) (result + length), \
-                                             maxlen, buf,                   \
-                                             arg, &count);                  \
-                        break;                                              \
-                      case 1:                                               \
-                        retcount = SNPRINTF ((TCHAR_T *) (result + length), \
-                                             maxlen, buf,                   \
-                                             prefixes[0], arg, &count);     \
-                        break;                                              \
-                      case 2:                                               \
-                        retcount = SNPRINTF ((TCHAR_T *) (result + length), \
-                                             maxlen, buf,                   \
-                                             prefixes[0], prefixes[1], arg, \
-                                             &count);                       \
-                        break;                                              \
-                      default:                                              \
-                        abort ();                                           \
-                      }
-#else
-# define SNPRINTF_BUF(arg) \
-                    switch (prefix_count)                                   \
-                      {                                                     \
-                      case 0:                                               \
-                        count = sprintf (tmp, buf, arg);                    \
-                        break;                                              \
-                      case 1:                                               \
-                        count = sprintf (tmp, buf, prefixes[0], arg);       \
-                        break;                                              \
-                      case 2:                                               \
-                        count = sprintf (tmp, buf, prefixes[0], prefixes[1],\
-                                         arg);                              \
-                        break;                                              \
-                      default:                                              \
-                        abort ();                                           \
-                      }
-#endif
-
-                    errno = 0;
-                    switch (type)
-                      {
-                      case TYPE_SCHAR:
-                        {
-                          int arg = a.arg[dp->arg_index].a.a_schar;
-                          SNPRINTF_BUF (arg);
-                        }
-                        break;
-                      case TYPE_UCHAR:
-                        {
-                          unsigned int arg = a.arg[dp->arg_index].a.a_uchar;
-                          SNPRINTF_BUF (arg);
-                        }
-                        break;
-                      case TYPE_SHORT:
-                        {
-                          int arg = a.arg[dp->arg_index].a.a_short;
-                          SNPRINTF_BUF (arg);
-                        }
-                        break;
-                      case TYPE_USHORT:
-                        {
-                          unsigned int arg = a.arg[dp->arg_index].a.a_ushort;
-                          SNPRINTF_BUF (arg);
-                        }
-                        break;
-                      case TYPE_INT:
-                        {
-                          int arg = a.arg[dp->arg_index].a.a_int;
-                          SNPRINTF_BUF (arg);
-                        }
-                        break;
-                      case TYPE_UINT:
-                        {
-                          unsigned int arg = a.arg[dp->arg_index].a.a_uint;
-                          SNPRINTF_BUF (arg);
-                        }
-                        break;
-                      case TYPE_LONGINT:
-                        {
-                          long int arg = a.arg[dp->arg_index].a.a_longint;
-                          SNPRINTF_BUF (arg);
-                        }
-                        break;
-                      case TYPE_ULONGINT:
-                        {
-                          unsigned long int arg = a.arg[dp->arg_index].a.a_ulongint;
-                          SNPRINTF_BUF (arg);
-                        }
-                        break;
-#if HAVE_LONG_LONG_INT
-                      case TYPE_LONGLONGINT:
-                        {
-                          long long int arg = a.arg[dp->arg_index].a.a_longlongint;
-                          SNPRINTF_BUF (arg);
-                        }
-                        break;
-                      case TYPE_ULONGLONGINT:
-                        {
-                          unsigned long long int arg = a.arg[dp->arg_index].a.a_ulonglongint;
-                          SNPRINTF_BUF (arg);
-                        }
-                        break;
-#endif
-                      case TYPE_DOUBLE:
-                        {
-                          double arg = a.arg[dp->arg_index].a.a_double;
-                          SNPRINTF_BUF (arg);
-                        }
-                        break;
-                      case TYPE_LONGDOUBLE:
-                        {
-                          long double arg = a.arg[dp->arg_index].a.a_longdouble;
-                          SNPRINTF_BUF (arg);
-                        }
-                        break;
-                      case TYPE_CHAR:
-                        {
-                          int arg = a.arg[dp->arg_index].a.a_char;
-                          SNPRINTF_BUF (arg);
-                        }
-                        break;
-#if HAVE_WINT_T
-                      case TYPE_WIDE_CHAR:
-                        {
-                          wint_t arg = a.arg[dp->arg_index].a.a_wide_char;
-                          SNPRINTF_BUF (arg);
-                        }
-                        break;
-#endif
-                      case TYPE_STRING:
-                        {
-                          const char *arg = a.arg[dp->arg_index].a.a_string;
-                          SNPRINTF_BUF (arg);
-                        }
-                        break;
-#if HAVE_WCHAR_T
-                      case TYPE_WIDE_STRING:
-                        {
-                          const wchar_t *arg = a.arg[dp->arg_index].a.a_wide_string;
-                          SNPRINTF_BUF (arg);
-                        }
-                        break;
-#endif
-                      case TYPE_POINTER:
-                        {
-                          void *arg = a.arg[dp->arg_index].a.a_pointer;
-                          SNPRINTF_BUF (arg);
-                        }
-                        break;
-                      default:
-                        abort ();
-                      }
-
-#if USE_SNPRINTF
-                    /* Portability: Not all implementations of snprintf()
-                       are ISO C 99 compliant.  Determine the number of
-                       bytes that snprintf() has produced or would have
-                       produced.  */
-                    if (count >= 0)
-                      {
-                        /* Verify that snprintf() has NUL-terminated its
-                           result.  */
-                        if (count < maxlen
-                            && ((TCHAR_T *) (result + length)) [count] != '\0')
-                          abort ();
-                        /* Portability hack.  */
-                        if (retcount > count)
-                          count = retcount;
-                      }
-                    else
-                      {
-                        /* snprintf() doesn't understand the '%n'
-                           directive.  */
-                        if (fbp[1] != '\0')
-                          {
-                            /* Don't use the '%n' directive; instead, look
-                               at the snprintf() return value.  */
-                            fbp[1] = '\0';
-                            continue;
-                          }
-                        else
-                          {
-                            /* Look at the snprintf() return value.  */
-                            if (retcount < 0)
-                              {
-# if !HAVE_SNPRINTF_RETVAL_C99
-                                /* HP-UX 10.20 snprintf() is doubly deficient:
-                                   It doesn't understand the '%n' directive,
-                                   *and* it returns -1 (rather than the length
-                                   that would have been required) when the
-                                   buffer is too small.
-                                   But a failure at this point can also come
-                                   from other reasons than a too small buffer,
-                                   such as an invalid wide string argument to
-                                   the %ls directive, or possibly an invalid
-                                   floating-point argument.  */
-                                size_t tmp_length =
-                                  MAX_ROOM_NEEDED (&a, dp->arg_index,
-                                                   dp->conversion, type, flags,
-                                                   width, has_precision,
-                                                   precision, pad_ourselves);
-
-                                if (maxlen < tmp_length)
-                                  {
-                                    /* Make more room.  But try to do through
-                                       this reallocation only once.  */
-                                    size_t bigger_need =
-                                      xsum (length,
-                                            xsum (tmp_length,
-                                                  TCHARS_PER_DCHAR - 1)
-                                            / TCHARS_PER_DCHAR);
-                                    /* And always grow proportionally.
-                                       (There may be several arguments, each
-                                       needing a little more room than the
-                                       previous one.)  */
-                                    size_t bigger_need2 =
-                                      xsum (xtimes (allocated, 2), 12);
-                                    if (bigger_need < bigger_need2)
-                                      bigger_need = bigger_need2;
-                                    ENSURE_ALLOCATION (bigger_need);
-                                    continue;
-                                  }
-# endif
-                              }
-                            else
-                              count = retcount;
-                          }
-                      }
-#endif
-
-                    /* Attempt to handle failure.  */
-                    if (count < 0)
-                      {
-                        /* SNPRINTF or sprintf failed.  Save and use the errno
-                           that it has set, if any.  */
-                        int saved_errno = errno;
-
-                        if (!(result == resultbuf || result == NULL))
-                          free (result);
-                        if (buf_malloced != NULL)
-                          free (buf_malloced);
-                        CLEANUP ();
-                        errno =
-                          (saved_errno != 0
-                           ? saved_errno
-                           : (dp->conversion == 'c' || dp->conversion == 's'
-                              ? EILSEQ
-                              : EINVAL));
-                        return NULL;
-                      }
-
-#if USE_SNPRINTF
-                    /* Handle overflow of the allocated buffer.
-                       If such an overflow occurs, a C99 compliant snprintf()
-                       returns a count >= maxlen.  However, a non-compliant
-                       snprintf() function returns only count = maxlen - 1.  To
-                       cover both cases, test whether count >= maxlen - 1.  */
-                    if ((unsigned int) count + 1 >= maxlen)
-                      {
-                        /* If maxlen already has attained its allowed maximum,
-                           allocating more memory will not increase maxlen.
-                           Instead of looping, bail out.  */
-                        if (maxlen == INT_MAX / TCHARS_PER_DCHAR)
-                          goto overflow;
-                        else
-                          {
-                            /* Need at least (count + 1) * sizeof (TCHAR_T)
-                               bytes.  (The +1 is for the trailing NUL.)
-                               But ask for (count + 2) * sizeof (TCHAR_T)
-                               bytes, so that in the next round, we likely get
-                                 maxlen > (unsigned int) count + 1
-                               and so we don't get here again.
-                               And allocate proportionally, to avoid looping
-                               eternally if snprintf() reports a too small
-                               count.  */
-                            size_t n =
-                              xmax (xsum (length,
-                                          ((unsigned int) count + 2
-                                           + TCHARS_PER_DCHAR - 1)
-                                          / TCHARS_PER_DCHAR),
-                                    xtimes (allocated, 2));
-
-                            ENSURE_ALLOCATION (n);
-                            continue;
-                          }
-                      }
-#endif
-
-#if NEED_PRINTF_UNBOUNDED_PRECISION
-                    if (prec_ourselves)
-                      {
-                        /* Handle the precision.  */
-                        TCHAR_T *prec_ptr =
-# if USE_SNPRINTF
-                          (TCHAR_T *) (result + length);
-# else
-                          tmp;
-# endif
-                        size_t prefix_count;
-                        size_t move;
-
-                        prefix_count = 0;
-                        /* Put the additional zeroes after the sign.  */
-                        if (count >= 1
-                            && (*prec_ptr == '-' || *prec_ptr == '+'
-                                || *prec_ptr == ' '))
-                          prefix_count = 1;
-                        /* Put the additional zeroes after the 0x prefix if
-                           (flags & FLAG_ALT) || (dp->conversion == 'p').  */
-                        else if (count >= 2
-                                 && prec_ptr[0] == '0'
-                                 && (prec_ptr[1] == 'x' || prec_ptr[1] == 'X'))
-                          prefix_count = 2;
-
-                        move = count - prefix_count;
-                        if (precision > move)
-                          {
-                            /* Insert zeroes.  */
-                            size_t insert = precision - move;
-                            TCHAR_T *prec_end;
-
-# if USE_SNPRINTF
-                            size_t n =
-                              xsum (length,
-                                    (count + insert + TCHARS_PER_DCHAR - 1)
-                                    / TCHARS_PER_DCHAR);
-                            length += (count + TCHARS_PER_DCHAR - 1) / TCHARS_PER_DCHAR;
-                            ENSURE_ALLOCATION (n);
-                            length -= (count + TCHARS_PER_DCHAR - 1) / TCHARS_PER_DCHAR;
-                            prec_ptr = (TCHAR_T *) (result + length);
-# endif
-
-                            prec_end = prec_ptr + count;
-                            prec_ptr += prefix_count;
-
-                            while (prec_end > prec_ptr)
-                              {
-                                prec_end--;
-                                prec_end[insert] = prec_end[0];
-                              }
-
-                            prec_end += insert;
-                            do
-                              *--prec_end = '0';
-                            while (prec_end > prec_ptr);
-
-                            count += insert;
-                          }
-                      }
-#endif
-
-#if !USE_SNPRINTF
-                    if (count >= tmp_length)
-                      /* tmp_length was incorrectly calculated - fix the
-                         code above!  */
-                      abort ();
-#endif
-
-#if !DCHAR_IS_TCHAR
-                    /* Convert from TCHAR_T[] to DCHAR_T[].  */
-                    if (dp->conversion == 'c' || dp->conversion == 's')
-                      {
-                        /* type = TYPE_CHAR or TYPE_WIDE_CHAR or TYPE_STRING
-                           TYPE_WIDE_STRING.
-                           The result string is not certainly ASCII.  */
-                        const TCHAR_T *tmpsrc;
-                        DCHAR_T *tmpdst;
-                        size_t tmpdst_len;
-                        /* This code assumes that TCHAR_T is 'char'.  */
-                        typedef int TCHAR_T_verify
-                                    [2 * (sizeof (TCHAR_T) == 1) - 1];
-# if USE_SNPRINTF
-                        tmpsrc = (TCHAR_T *) (result + length);
-# else
-                        tmpsrc = tmp;
-# endif
-                        tmpdst =
-                          DCHAR_CONV_FROM_ENCODING (locale_charset (),
-                                                    iconveh_question_mark,
-                                                    tmpsrc, count,
-                                                    NULL,
-                                                    NULL, &tmpdst_len);
-                        if (tmpdst == NULL)
-                          {
-                            int saved_errno = errno;
-                            if (!(result == resultbuf || result == NULL))
-                              free (result);
-                            if (buf_malloced != NULL)
-                              free (buf_malloced);
-                            CLEANUP ();
-                            errno = saved_errno;
-                            return NULL;
-                          }
-                        ENSURE_ALLOCATION (xsum (length, tmpdst_len));
-                        DCHAR_CPY (result + length, tmpdst, tmpdst_len);
-                        free (tmpdst);
-                        count = tmpdst_len;
-                      }
-                    else
-                      {
-                        /* The result string is ASCII.
-                           Simple 1:1 conversion.  */
-# if USE_SNPRINTF
-                        /* If sizeof (DCHAR_T) == sizeof (TCHAR_T), it's a
-                           no-op conversion, in-place on the array starting
-                           at (result + length).  */
-                        if (sizeof (DCHAR_T) != sizeof (TCHAR_T))
-# endif
-                          {
-                            const TCHAR_T *tmpsrc;
-                            DCHAR_T *tmpdst;
-                            size_t n;
-
-# if USE_SNPRINTF
-                            if (result == resultbuf)
-                              {
-                                tmpsrc = (TCHAR_T *) (result + length);
-                                /* ENSURE_ALLOCATION will not move tmpsrc
-                                   (because it's part of resultbuf).  */
-                                ENSURE_ALLOCATION (xsum (length, count));
-                              }
-                            else
-                              {
-                                /* ENSURE_ALLOCATION will move the array
-                                   (because it uses realloc().  */
-                                ENSURE_ALLOCATION (xsum (length, count));
-                                tmpsrc = (TCHAR_T *) (result + length);
-                              }
-# else
-                            tmpsrc = tmp;
-                            ENSURE_ALLOCATION (xsum (length, count));
-# endif
-                            tmpdst = result + length;
-                            /* Copy backwards, because of overlapping.  */
-                            tmpsrc += count;
-                            tmpdst += count;
-                            for (n = count; n > 0; n--)
-                              *--tmpdst = (unsigned char) *--tmpsrc;
-                          }
-                      }
-#endif
-
-#if DCHAR_IS_TCHAR && !USE_SNPRINTF
-                    /* Make room for the result.  */
-                    if (count > allocated - length)
-                      {
-                        /* Need at least count elements.  But allocate
-                           proportionally.  */
-                        size_t n =
-                          xmax (xsum (length, count), xtimes (allocated, 2));
-
-                        ENSURE_ALLOCATION (n);
-                      }
-#endif
-
-                    /* Here count <= allocated - length.  */
-
-                    /* Perform padding.  */
-#if !DCHAR_IS_TCHAR || ENABLE_UNISTDIO || NEED_PRINTF_FLAG_LEFTADJUST || NEED_PRINTF_FLAG_ZERO || NEED_PRINTF_UNBOUNDED_PRECISION
-                    if (pad_ourselves && has_width)
-                      {
-                        size_t w;
-# if ENABLE_UNISTDIO
-                        /* Outside POSIX, it's preferrable to compare the width
-                           against the number of _characters_ of the converted
-                           value.  */
-                        w = DCHAR_MBSNLEN (result + length, count);
-# else
-                        /* The width is compared against the number of _bytes_
-                           of the converted value, says POSIX.  */
-                        w = count;
-# endif
-                        if (w < width)
-                          {
-                            size_t pad = width - w;
-
-                            /* Make room for the result.  */
-                            if (xsum (count, pad) > allocated - length)
-                              {
-                                /* Need at least count + pad elements.  But
-                                   allocate proportionally.  */
-                                size_t n =
-                                  xmax (xsum3 (length, count, pad),
-                                        xtimes (allocated, 2));
-
-# if USE_SNPRINTF
-                                length += count;
-                                ENSURE_ALLOCATION (n);
-                                length -= count;
-# else
-                                ENSURE_ALLOCATION (n);
-# endif
-                              }
-                            /* Here count + pad <= allocated - length.  */
-
-                            {
-# if !DCHAR_IS_TCHAR || USE_SNPRINTF
-                              DCHAR_T * const rp = result + length;
-# else
-                              DCHAR_T * const rp = tmp;
-# endif
-                              DCHAR_T *p = rp + count;
-                              DCHAR_T *end = p + pad;
-                              DCHAR_T *pad_ptr;
-# if !DCHAR_IS_TCHAR || ENABLE_UNISTDIO
-                              if (dp->conversion == 'c'
-                                  || dp->conversion == 's')
-                                /* No zero-padding for string directives.  */
-                                pad_ptr = NULL;
-                              else
-# endif
-                                {
-                                  pad_ptr = (*rp == '-' ? rp + 1 : rp);
-                                  /* No zero-padding of "inf" and "nan".  */
-                                  if ((*pad_ptr >= 'A' && *pad_ptr <= 'Z')
-                                      || (*pad_ptr >= 'a' && *pad_ptr <= 'z'))
-                                    pad_ptr = NULL;
-                                }
-                              /* The generated string now extends from rp to p,
-                                 with the zero padding insertion point being at
-                                 pad_ptr.  */
-
-                              count = count + pad; /* = end - rp */
-
-                              if (flags & FLAG_LEFT)
-                                {
-                                  /* Pad with spaces on the right.  */
-                                  for (; pad > 0; pad--)
-                                    *p++ = ' ';
-                                }
-                              else if ((flags & FLAG_ZERO) && pad_ptr != NULL)
-                                {
-                                  /* Pad with zeroes.  */
-                                  DCHAR_T *q = end;
-
-                                  while (p > pad_ptr)
-                                    *--q = *--p;
-                                  for (; pad > 0; pad--)
-                                    *p++ = '0';
-                                }
-                              else
-                                {
-                                  /* Pad with spaces on the left.  */
-                                  DCHAR_T *q = end;
-
-                                  while (p > rp)
-                                    *--q = *--p;
-                                  for (; pad > 0; pad--)
-                                    *p++ = ' ';
-                                }
-                            }
-                          }
-                      }
-#endif
-
-                    /* Here still count <= allocated - length.  */
-
-#if !DCHAR_IS_TCHAR || USE_SNPRINTF
-                    /* The snprintf() result did fit.  */
-#else
-                    /* Append the sprintf() result.  */
-                    memcpy (result + length, tmp, count * sizeof (DCHAR_T));
-#endif
-#if !USE_SNPRINTF
-                    if (tmp != tmpbuf)
-                      free (tmp);
-#endif
-
-#if NEED_PRINTF_DIRECTIVE_F
-                    if (dp->conversion == 'F')
-                      {
-                        /* Convert the %f result to upper case for %F.  */
-                        DCHAR_T *rp = result + length;
-                        size_t rc;
-                        for (rc = count; rc > 0; rc--, rp++)
-                          if (*rp >= 'a' && *rp <= 'z')
-                            *rp = *rp - 'a' + 'A';
-                      }
-#endif
-
-                    length += count;
-                    break;
-                  }
-#undef pad_ourselves
-#undef prec_ourselves
-              }
-          }
-      }
-
-    /* Add the final NUL.  */
-    ENSURE_ALLOCATION (xsum (length, 1));
-    result[length] = '\0';
-
-    if (result != resultbuf && length + 1 < allocated)
-      {
-        /* Shrink the allocated memory if possible.  */
-        DCHAR_T *memory;
-
-        memory = (DCHAR_T *) realloc (result, (length + 1) * sizeof (DCHAR_T));
-        if (memory != NULL)
-          result = memory;
-      }
-
-    if (buf_malloced != NULL)
-      free (buf_malloced);
-    CLEANUP ();
-    *lengthp = length;
-    /* Note that we can produce a big string of a length > INT_MAX.  POSIX
-       says that snprintf() fails with errno = EOVERFLOW in this case, but
-       that's only because snprintf() returns an 'int'.  This function does
-       not have this limitation.  */
-    return result;
-
-#if USE_SNPRINTF
-  overflow:
-    if (!(result == resultbuf || result == NULL))
+  length_overflow:
+    /* We could produce such a big string, but its length doesn't fit into
+       an 'int'.  POSIX says that snprintf() fails with errno = EOVERFLOW in
+       this case.  */
+    if (result != resultbuf)
       free (result);
-    if (buf_malloced != NULL)
-      free (buf_malloced);
-    CLEANUP ();
     errno = EOVERFLOW;
     return NULL;
-#endif
 
   out_of_memory:
     if (!(result == resultbuf || result == NULL))
@@ -5552,17 +893,10 @@ VASNPRINTF (DCHAR_T *resultbuf, size_t *lengthp,
   }
 }
 
-#undef MAX_ROOM_NEEDED
-#undef TCHARS_PER_DCHAR
 #undef SNPRINTF
 #undef USE_SNPRINTF
-#undef DCHAR_SET
-#undef DCHAR_CPY
 #undef PRINTF_PARSE
 #undef DIRECTIVES
 #undef DIRECTIVE
-#undef DCHAR_IS_TCHAR
-#undef TCHAR_T
-#undef DCHAR_T
-#undef FCHAR_T
+#undef CHAR_T
 #undef VASNPRINTF