vfscanf.c

绝对正真的stdio.h的实现

		      /*  And get the mbdigits + extra_digit string.  */
		      *(char *) __mempcpy (__mempcpy (mb_extended, mbdigits[n],
						      mbdigits_len),
					   extra_mbdigit, mblen) = '\0';
		      mbdigits_extended[n] = mb_extended;
#endif
		    }
		}

	      /* Read the number into workspace.  */
	      while (c != EOF && width != 0)
		{
		  /* In this round we get the pointer to the digit strings
		     and also perform the first round of comparisons.  */
		  for (n = 0; n < 10; ++n)
		    {
		      /* Get the string for the digits with value N.  */
#ifdef COMPILE_WSCANF
		      if (__builtin_expect (map != NULL, 0))
			wcdigits[n] = wcdigits_extended[n];
		      else
			wcdigits[n] = (const wchar_t *)
			  _NL_CURRENT (LC_CTYPE, _NL_CTYPE_INDIGITS0_WC + n);
		      wcdigits[n] += from_level;

		      if (c == (wint_t) *wcdigits[n])
			{
			  to_level = from_level;
			  break;
			}

		      /* Advance the pointer to the next string.  */
		      ++wcdigits[n];
#else
		      const char *cmpp;
		      int avail = width > 0 ? width : INT_MAX;

		      if (__builtin_expect (map != NULL, 0))
			mbdigits[n] = mbdigits_extended[n];
		      else
			mbdigits[n]
			  = curctype->values[_NL_CTYPE_INDIGITS0_MB + n].string;

		      for (level = 0; level < from_level; level++)
			mbdigits[n] = strchr (mbdigits[n], '\0') + 1;

		      cmpp = mbdigits[n];
		      while ((unsigned char) *cmpp == c && avail > 0)
			{
			  if (*++cmpp == '\0')
			    break;
			  else
			    {
			      if ((c = inchar ()) == EOF)
				break;
			      --avail;
			    }
			}

		      if (*cmpp == '\0')
			{
			  if (width > 0)
			    width = avail;
			  to_level = from_level;
			  break;
			}

		      /* We are pushing all read characters back.  */
		      if (cmpp > mbdigits[n])
			{
			  ungetc (c, s);
			  while (--cmpp > mbdigits[n])
			    ungetc_not_eof ((unsigned char) *cmpp, s);
			  c = (unsigned char) *cmpp;
			}

		      /* Advance the pointer to the next string.  */
		      mbdigits[n] = strchr (mbdigits[n], '\0') + 1;
#endif
		    }

		  if (n == 10)
		    {
		      /* Have not yet found the digit.  */
		      for (level = from_level + 1; level <= to_level; ++level)
			{
			  /* Search all ten digits of this level.  */
			  for (n = 0; n < 10; ++n)
			    {
#ifdef COMPILE_WSCANF
			      if (c == (wint_t) *wcdigits[n])
				break;

			      /* Advance the pointer to the next string.  */
			      ++wcdigits[n];
#else
			      const char *cmpp;
			      int avail = width > 0 ? width : INT_MAX;

			      cmpp = mbdigits[n];
			      while ((unsigned char) *cmpp == c && avail > 0)
				{
				  if (*++cmpp == '\0')
				    break;
				  else
				    {
				      if ((c = inchar ()) == EOF)
					break;
				      --avail;
				    }
				}

			      if (*cmpp == '\0')
				{
				  if (width > 0)
				    width = avail;
				  break;
				}

			      /* We are pushing all read characters back.  */
			      if (cmpp > mbdigits[n])
				{
				  ungetc (c, s);
				  while (--cmpp > mbdigits[n])
				    ungetc_not_eof ((unsigned char) *cmpp, s);
				  c = (unsigned char) *cmpp;
				}

			      /* Advance the pointer to the next string.  */
			      mbdigits[n] = strchr (mbdigits[n], '\0') + 1;
#endif
			    }

			  if (n < 10)
			    {
			      /* Found it.  */
			      from_level = level;
			      to_level = level;
			      break;
			    }
			}
		    }

		  if (n < 10)
		    c = L_('0') + n;
		  else if ((flags & GROUP)
#ifdef COMPILE_WSCANF
			   && thousands != L'\0'
#else
			   && thousands != NULL
#endif
			   )
		    {
		      /* Try matching against the thousands separator.  */
#ifdef COMPILE_WSCANF
		      if (c != thousands)
			  break;
#else
		      const char *cmpp = thousands;
		      int avail = width > 0 ? width : INT_MAX;

		      while ((unsigned char) *cmpp == c && avail > 0)
			{
			  ADDW (c);
			  if (*++cmpp == '\0')
			    break;
			  else
			    {
			      if ((c = inchar ()) == EOF)
				break;
			      --avail;
			    }
			}

		      if (*cmpp != '\0')
			{
			  /* We are pushing all read characters back.  */
			  if (cmpp > thousands)
			    {
			      wpsize -= cmpp - thousands;
			      ungetc (c, s);
			      while (--cmpp > thousands)
				ungetc_not_eof ((unsigned char) *cmpp, s);
			      c = (unsigned char) *cmpp;
			    }
			  break;
			}

		      if (width > 0)
			width = avail;

		      /* The last thousands character will be added back by
			 the ADDW below.  */
			--wpsize;
#endif
		    }
		  else
		    break;

		  ADDW (c);
		  if (width > 0)
		    --width;

		  c = inchar ();
		}
	    }
	  else
	    /* Read the number into workspace.  */
	    while (c != EOF && width != 0)
	      {
		if (base == 16)
		  {
		    if (!ISXDIGIT (c))
		      break;
		  }
		else if (!ISDIGIT (c) || (int) (c - L_('0')) >= base)
		  {
		    if (base == 10 && (flags & GROUP)
#ifdef COMPILE_WSCANF
			&& thousands != L'\0'
#else
			&& thousands != NULL
#endif
			)
		      {
			/* Try matching against the thousands separator.  */
#ifdef COMPILE_WSCANF
			if (c != thousands)
			  break;
#else
			const char *cmpp = thousands;
			int avail = width > 0 ? width : INT_MAX;

			while ((unsigned char) *cmpp == c && avail > 0)
			  {
			    ADDW (c);
			    if (*++cmpp == '\0')
			      break;
			    else
			      {
				if ((c = inchar ()) == EOF)
				  break;
				--avail;
			      }
			  }

			if (*cmpp != '\0')
			  {
			    /* We are pushing all read characters back.  */
			    if (cmpp > thousands)
			      {
				wpsize -= cmpp - thousands;
				ungetc (c, s);
				while (--cmpp > thousands)
				  ungetc_not_eof ((unsigned char) *cmpp, s);
				c = (unsigned char) *cmpp;
			      }
			    break;
			  }

			if (width > 0)
			  width = avail;

			/* The last thousands character will be added back by
			   the ADDW below.  */
			--wpsize;
#endif
		      }
		    else
		      break;
		  }
		ADDW (c);
		if (width > 0)
		  --width;

		c = inchar ();
	      }

	  if (wpsize == 0
	      || (wpsize == 1 && (wp[0] == L_('+') || wp[0] == L_('-'))))
	    {
	      /* There was no number.  If we are supposed to read a pointer
		 we must recognize "(nil)" as well.  */
	      if (wpsize == 0 && read_pointer && (width < 0 || width >= 0)
		  && c == '('
		  && TOLOWER (inchar ()) == L_('n')
		  && TOLOWER (inchar ()) == L_('i')
		  && TOLOWER (inchar ()) == L_('l')
		  && inchar () == L_(')'))
		/* We must produce the value of a NULL pointer.  A single
		   '0' digit is enough.  */
		ADDW (L_('0'));
	      else
		{
		  /* The last read character is not part of the number
		     anymore.  */
		  ungetc (c, s);

		  conv_error ();
		}
	    }
	  else
	    /* The just read character is not part of the number anymore.  */
	    ungetc (c, s);

	  /* Convert the number.  */
	  ADDW (L_('\0'));
	  if (need_longlong && (flags & LONGDBL))
	    {
	      if (number_signed)
		num.q = __strtoll_internal (wp, &tw, base, flags & GROUP);
	      else
		num.uq = __strtoull_internal (wp, &tw, base, flags & GROUP);
	    }
	  else
	    {
	      if (number_signed)
		num.l = __strtol_internal (wp, &tw, base, flags & GROUP);
	      else
		num.ul = __strtoul_internal (wp, &tw, base, flags & GROUP);
	    }
	  if (wp == tw)
	    conv_error ();

	  if (!(flags & SUPPRESS))
	    {
	      if (! number_signed)
		{
		  if (need_longlong && (flags & LONGDBL))
		    *ARG (unsigned LONGLONG int *) = num.uq;
		  else if (need_long && (flags & LONG))
		    *ARG (unsigned long int *) = num.ul;
		  else if (flags & SHORT)
		    *ARG (unsigned short int *)
		      = (unsigned short int) num.ul;
		  else if (!(flags & CHAR))
		    *ARG (unsigned int *) = (unsigned int) num.ul;
		  else
		    *ARG (unsigned char *) = (unsigned char) num.ul;
		}
	      else
		{
		  if (need_longlong && (flags & LONGDBL))
		    *ARG (LONGLONG int *) = num.q;
		  else if (need_long && (flags & LONG))
		    *ARG (long int *) = num.l;
		  else if (flags & SHORT)
		    *ARG (short int *) = (short int) num.l;
		  else if (!(flags & CHAR))
		    *ARG (int *) = (int) num.l;
		  else
		    *ARG (signed char *) = (signed char) num.ul;
		}
	      ++done;
	    }
	  break;

	case L_('e'):	/* Floating-point numbers.  */
	case L_('E'):
	case L_('f'):
	case L_('F'):
	case L_('g'):
	case L_('G'):
	case L_('a'):
	case L_('A'):
	  c = inchar ();
	  if (c == EOF)
	    input_error ();

	  /* Check for a sign.  */
	  if (c == L_('-') || c == L_('+'))
	    {
	      negative = c == L_('-');
	      if (width == 0 || inchar () == EOF)
		/* EOF is only an input error before we read any chars.  */
		conv_error ();
	      if (! ISDIGIT (c) && TOLOWER (c) != L_('i')
		  && TOLOWER (c) != L_('n'))
		{
#ifdef COMPILE_WSCANF
		  if (c != decimal)
		    {
		      /* This is no valid number.  */
		      ungetc (c, s);
		      conv_error ();
		    }
#else
		  /* Match against the decimal point.  At this point
                     we are taking advantage of the fact that we can
                     push more than one character back.  This is
                     (almost) never necessary since the decimal point
                     string hopefully never contains more than one
                     byte.  */
		  const char *cmpp = decimal;
		  int avail = width > 0 ? width : INT_MAX;

		  while ((unsigned char) *cmpp == c && avail > 0)
		    if (*++cmpp == '\0')
		      break;
		    else
		      {
			if (inchar () == EOF)
			  break;
			--avail;
		      }

		  if (*cmpp != '\0')
		    {
		      /* This is no valid number.  */
		      while (1)
			{
			  ungetc (c, s);
			  if (cmpp == decimal)
			    break;
			  c = (unsigned char) *--cmpp;
			}

		      conv_error ();
		    }
		  if (width > 0)
		    width = avail;
#endif
		}
	      if (width > 0)
		--width;
	    }
	  else
	    negative = 0;

	  /* Take care for the special arguments "nan" and "inf".  */
	  if (TOLOWER (c) == L_('n'))
	    {
	      /* Maybe "nan".  */
	      ADDW (c);
	      if (width == 0 || inchar () == EOF || TOLOWER (c) != L_('a'))
		conv_error ();
	      if (width > 0)
		--width;
	      ADDW (c);
	      if (width == 0 || inchar () == EOF || TOLOWER (c) != L_('n'))
		conv_error ();
	      if (width > 0)
		--width;
	      ADDW (c);
	      /* It is "nan".  */
	      goto scan_float;
	    }
	  else if (TOLOWER (c) == L_('i'))
	    {
	      /* Maybe "inf" or "infinity".  */
	      ADDW (c);
	      if (width == 0 || inchar () == EOF || TOLOWER (c) != L_('n'))
		conv_error ();
	      if (width > 0)
		--width;
	      ADDW (c);
	      if (width == 0 || inchar () == EOF || TOLOWER (c) != L_('f'))
		conv_error ();
	      if (width > 0)
		--width;
	      ADDW (c);
	      /* It is as least "inf".  */
	      if (width != 0 && inchar () != EOF)
		{
		  if (TOLOWER (c) == L_('i'))
		    {
		      if (width > 0)
			--width;
		      /* Now we have to read the rest as well.  */
		      ADDW (c);
		      if (width == 0 || inchar () == EOF
			  || TOLOWER (c) != L_('n'))
			conv_error ();
		      if (width > 0)
			--width;
		      ADDW (c);
		      if (width == 0 || inchar () == EOF
			  || TOLOWER (c) != L_('i'))
			conv_error ();
		      if (width > 0)
			--width;
		      ADDW (c);
		      if (width == 0 || inchar () == EOF
			  || TOLOWER (c) != L_('t'))
			conv_error ();
		      if (width > 0)
			--width;
		      ADDW (c);
		      if (width == 0 || inchar () == EOF
			  || TOLOWER (c) != L_('y'))
			conv_error ();
		      if (width > 0)
			--width;
		      ADDW (c);
		    }
		  else
		    /* Never mind.  */
		    ungetc (c, s);
		}
	      goto scan_float;
	    }

	  is_hexa = 0;
	  exp_char = L_('e');
	  if (width != 0 && c == L_('0'))
	    {
	      ADDW (c);
	      c = inchar ();
	      if (width > 0)
		--width;
	      if (width != 0 && TOLOWER (c) == L_('x'))
		{
		  /* It is a number in hexadecimal format.  */
		  ADDW (c);

		  is_hexa = 1;
		  exp_char = L_('p');

		  /* Grouping is not allowed.  */
		  flags &= ~GROUP;
		  c = inchar ();
		  if (width > 0)
		    --width;
		}
	    }

	  got_dot = got_e = 0;
	  do
	    {
	      if (ISDIGIT (c))
		ADDW (c);
	      else if (!got_e && is_hexa && ISXDIGIT (c))
		ADDW (c);
	      else if (got_e && wp[wpsize - 1] == exp_char
		       && (c == L_('-') || c == L_('+')))
		ADDW (c);
	      else if (wpsize > 0 && !got_e
		       && (CHAR_T) TOLOWER (c) == exp_char)
		{
		  ADDW (exp_char);
		  got_e = got_dot = 1;
		}
	      else
		{
#ifdef COMPILE_WSCANF
		  if (! got_dot && c == decimal)
		    {
		      ADDW (c);
		      got_dot = 1;
		    }
		  else if ((flags & GROUP) != 0 && thousands != L'\0'
			   && ! got_dot && c == thousands)
		    ADDW (c);
		  else
		    {
		      /* The last read character is not part of the number
			 anymore.  */
		      ungetc (c, s);
		      break;
		    }
#else
		  const char *cmpp = decimal;
		  int avail = width > 0 ? width : INT_MAX;

		  if (! got_dot)
		    {
		      while ((unsigned char) *cmpp == c && avail > 0)
			if (*++cmpp == '\0')
			  break;
			else
			  {
			    if (inchar () == EOF)
			      break;
			    --avail;
			  }
		    }

		  if (*cmpp == '\0')
		    {
		      /* Add all the characters.  */
		      for (cmpp = decimal; *cmpp != '\0'; ++cmpp)
			ADDW ((unsigned char) *cmpp);
		      if (width > 0)
			width = avail;
		      got_dot = 1;
		    }
		  else
		    {
		      /* Figure out whether it is a thousands separator.
			 There is one problem: we possibly read more than
			 one character.  We cannot push them back but since
			 we know that parts of the `decimal' string matched,
			 we can compare against it.  */
		      const char *cmp2p = thousands;

		      if ((flags & GROUP) != 0 && thousands != NULL
			  && ! got_dot)
			{
			  while (cmp2p - thousands < cmpp - decimal
				 && *cmp2p == decimal[cmp2p - thousands])
			    ++cmp2p;
			  if (cmp2p - thousands == cmpp - decimal)
			    {
			      while ((unsigned char) *cmp2p == c && avail > 0)
				if (*++cmp2p == '\0')
				  break;
				else
				  {
				    if (inchar () == EOF)
				      break;
				    --avail;
				  }
			    }
			}

		      if (cmp2p != NULL && *cmp2p == '\0')