vfscanf.c

绝对正真的stdio.h的实现

	    }
	  break;

	case L_('c'):	/* Match characters.  */
	  if ((flags & LONG) == 0)
	    {
	      if (!(flags & SUPPRESS))
		{
		  str = ARG (char *);
		  if (str == NULL)
		    conv_error ();
		}

	      c = inchar ();
	      if (c == EOF)
		input_error ();

	      if (width == -1)
		width = 1;

#ifdef COMPILE_WSCANF
	      /* We have to convert the wide character(s) into multibyte
		 characters and store the result.  */
	      memset (&state, '\0', sizeof (state));

	      do
		{
		  size_t n;

		  n = __wcrtomb (!(flags & SUPPRESS) ? str : NULL, c, &state);
		  if (n == (size_t) -1)
		    /* No valid wide character.  */
		    input_error ();

		  /* Increment the output pointer.  Even if we don't
		     write anything.  */
		  str += n;
		}
	      while (--width > 0 && inchar () != EOF);
#else
	      if (!(flags & SUPPRESS))
		{
		  do
		    *str++ = c;
		  while (--width > 0 && inchar () != EOF);
		}
	      else
		while (--width > 0 && inchar () != EOF);
#endif

	      if (!(flags & SUPPRESS))
		++done;

	      break;
	    }
	  /* FALLTHROUGH */
	case L_('C'):
	  if (!(flags & SUPPRESS))
	    {
	      wstr = ARG (wchar_t *);
	      if (wstr == NULL)
		conv_error ();
	    }

	  c = inchar ();
	  if (c == EOF)
	    input_error ();

#ifdef COMPILE_WSCANF
	  /* Just store the incoming wide characters.  */
	  if (!(flags & SUPPRESS))
	    {
	      do
		*wstr++ = c;
	      while (--width > 0 && inchar () != EOF);
	    }
	  else
	    while (--width > 0 && inchar () != EOF);
#else
	  {
	    /* We have to convert the multibyte input sequence to wide
	       characters.  */
	    char buf[1];
	    mbstate_t cstate;

	    memset (&cstate, '\0', sizeof (cstate));

	    do
	      {
		/* This is what we present the mbrtowc function first.  */
		buf[0] = c;

		while (1)
		  {
		    size_t n;

		    n = __mbrtowc (!(flags & SUPPRESS) ? wstr : NULL,
				   buf, 1, &cstate);

		    if (n == (size_t) -2)
		      {
			/* Possibly correct character, just not enough
			   input.  */
			if (inchar () == EOF)
			  encode_error ();

			buf[0] = c;
			continue;
		      }

		    if (n != 1)
		      encode_error ();

		    /* We have a match.  */
		    break;
		  }

		/* Advance the result pointer.  */
		++wstr;
	      }
	    while (--width > 0 && inchar () != EOF);
	  }
#endif

	  if (!(flags & SUPPRESS))
	    ++done;

	  break;

	case L_('s'):		/* Read a string.  */
	  if (!(flags & LONG))
	    {
#define STRING_ARG(Str, Type)						      \
	      do if (!(flags & SUPPRESS))				      \
		{							      \
		  if (flags & MALLOC)					      \
		    {							      \
		      /* The string is to be stored in a malloc'd buffer.  */ \
		      strptr = ARG (char **);				      \
		      if (strptr == NULL)				      \
			conv_error ();					      \
		      /* Allocate an initial buffer.  */		      \
		      strsize = 100;					      \
		      *strptr = (char *) malloc (strsize * sizeof (Type));    \
		      Str = (Type *) *strptr;				      \
		    }							      \
		  else							      \
		    Str = ARG (Type *);					      \
		  if (Str == NULL)					      \
		    conv_error ();					      \
		} while (0)
	      STRING_ARG (str, char);

	      c = inchar ();
	      if (c == EOF)
		input_error ();

#ifdef COMPILE_WSCANF
	      memset (&state, '\0', sizeof (state));
#endif

	      do
		{
		  if (ISSPACE (c))
		    {
		      ungetc_not_eof (c, s);
		      break;
		    }

#ifdef COMPILE_WSCANF
		  /* This is quite complicated.  We have to convert the
		     wide characters into multibyte characters and then
		     store them.  */
		  {
		    size_t n;

		    if (!(flags & SUPPRESS) && (flags & MALLOC)
			&& str + MB_CUR_MAX >= *strptr + strsize)
		      {
			/* We have to enlarge the buffer if the `a' flag
			   was given.  */
			size_t strleng = str - *strptr;
			char *newstr;

			newstr = (char *) realloc (*strptr, strsize * 2);
			if (newstr == NULL)
			  {
			    /* Can't allocate that much.  Last-ditch
			       effort.  */
			    newstr = (char *) realloc (*strptr,
						       strleng + MB_CUR_MAX);
			    if (newstr == NULL)
			      {
				/* We lose.  Oh well.  Terminate the
				   string and stop converting,
				   so at least we don't skip any input.  */
				((char *) (*strptr))[strleng] = '\0';
				++done;
				conv_error ();
			      }
			    else
			      {
				*strptr = newstr;
				str = newstr + strleng;
				strsize = strleng + MB_CUR_MAX;
			      }
			  }
			else
			  {
			    *strptr = newstr;
			    str = newstr + strleng;
			    strsize *= 2;
			  }
		      }

		    n = __wcrtomb (!(flags & SUPPRESS) ? str : NULL, c,
				   &state);
		    if (n == (size_t) -1)
		      encode_error ();

		    assert (n <= MB_CUR_MAX);
		    str += n;
		  }
#else
		  /* This is easy.  */
		  if (!(flags & SUPPRESS))
		    {
		      *str++ = c;
		      if ((flags & MALLOC)
			  && (char *) str == *strptr + strsize)
			{
			  /* Enlarge the buffer.  */
			  str = (char *) realloc (*strptr, 2 * strsize);
			  if (str == NULL)
			    {
			      /* Can't allocate that much.  Last-ditch
				 effort.  */
			      str = (char *) realloc (*strptr, strsize + 1);
			      if (str == NULL)
				{
				  /* We lose.  Oh well.  Terminate the
				     string and stop converting,
				     so at least we don't skip any input.  */
				  ((char *) (*strptr))[strsize - 1] = '\0';
				  ++done;
				  conv_error ();
				}
			      else
				{
				  *strptr = (char *) str;
				  str += strsize;
				  ++strsize;
				}
			    }
			  else
			    {
			      *strptr = (char *) str;
			      str += strsize;
			      strsize *= 2;
			    }
			}
		    }
#endif
		}
	      while ((width <= 0 || --width > 0) && inchar () != EOF);

	      if (!(flags & SUPPRESS))
		{
#ifdef COMPILE_WSCANF
		  /* We have to emit the code to get into the initial
		     state.  */
		  char buf[MB_LEN_MAX];
		  size_t n = __wcrtomb (buf, L'\0', &state);
		  if (n > 0 && (flags & MALLOC)
		      && str + n >= *strptr + strsize)
		    {
		      /* Enlarge the buffer.  */
		      size_t strleng = str - *strptr;
		      char *newstr;

		      newstr = (char *) realloc (*strptr, strleng + n + 1);
		      if (newstr == NULL)
			{
			  /* We lose.  Oh well.  Terminate the string
			     and stop converting, so at least we don't
			     skip any input.  */
			  ((char *) (*strptr))[strleng] = '\0';
			  ++done;
			  conv_error ();
			}
		      else
			{
			  *strptr = newstr;
			  str = newstr + strleng;
			  strsize = strleng + n + 1;
			}
		    }

		  str = __mempcpy (str, buf, n);
#endif
		  *str++ = '\0';

		  if ((flags & MALLOC) && str - *strptr != strsize)
		    {
		      char *cp = (char *) realloc (*strptr, str - *strptr);
		      if (cp != NULL)
			*strptr = cp;
		    }

		  ++done;
		}
	      break;
	    }
	  /* FALLTHROUGH */

	case L_('S'):
	  {
#ifndef COMPILE_WSCANF
	    mbstate_t cstate;
#endif

	    /* Wide character string.  */
	    STRING_ARG (wstr, wchar_t);

	    c = inchar ();
	    if (c == EOF)
	      input_error ();

#ifndef COMPILE_WSCANF
	    memset (&cstate, '\0', sizeof (cstate));
#endif

	    do
	      {
		if (ISSPACE (c))
		  {
		    ungetc_not_eof (c, s);
		    break;
		  }

#ifdef COMPILE_WSCANF
		/* This is easy.  */
		if (!(flags & SUPPRESS))
		  {
		    *wstr++ = c;
		    if ((flags & MALLOC)
			&& wstr == (wchar_t *) *strptr + strsize)
		      {
			/* Enlarge the buffer.  */
			wstr = (wchar_t *) realloc (*strptr,
						    (2 * strsize)
						    * sizeof (wchar_t));
			if (wstr == NULL)
			  {
			    /* Can't allocate that much.  Last-ditch
                               effort.  */
			    wstr = (wchar_t *) realloc (*strptr,
							(strsize + 1)
							* sizeof (wchar_t));
			    if (wstr == NULL)
			      {
				/* We lose.  Oh well.  Terminate the string
				   and stop converting, so at least we don't
				   skip any input.  */
				((wchar_t *) (*strptr))[strsize - 1] = L'\0';
				++done;
				conv_error ();
			      }
			    else
			      {
				*strptr = (char *) wstr;
				wstr += strsize;
				++strsize;
			      }
			  }
			else
			  {
			    *strptr = (char *) wstr;
			    wstr += strsize;
			    strsize *= 2;
			  }
		      }
		  }
#else
		{
		  char buf[1];

		  buf[0] = c;

		  while (1)
		    {
		      size_t n;

		      n = __mbrtowc (!(flags & SUPPRESS) ? wstr : NULL,
				     buf, 1, &cstate);

		      if (n == (size_t) -2)
			{
			  /* Possibly correct character, just not enough
			     input.  */
			  if (inchar () == EOF)
			    encode_error ();

			  buf[0] = c;
			  continue;
			}

		      if (n != 1)
			encode_error ();

		      /* We have a match.  */
		      ++wstr;
		      break;
		    }

		  if (!(flags & SUPPRESS) && (flags & MALLOC)
		      && wstr == (wchar_t *) *strptr + strsize)
		    {
		      /* Enlarge the buffer.  */
		      wstr = (wchar_t *) realloc (*strptr,
						  (2 * strsize
						   * sizeof (wchar_t)));
		      if (wstr == NULL)
			{
			  /* Can't allocate that much.  Last-ditch effort.  */
			  wstr = (wchar_t *) realloc (*strptr,
						      ((strsize + 1)
						       * sizeof (wchar_t)));
			  if (wstr == NULL)
			    {
			      /* We lose.  Oh well.  Terminate the
				 string and stop converting, so at
				 least we don't skip any input.  */
			      ((wchar_t *) (*strptr))[strsize - 1] = L'\0';
			      ++done;
			      conv_error ();
			    }
			  else
			    {
			      *strptr = (char *) wstr;
			      wstr += strsize;
			      ++strsize;
			    }
			}
		      else
			{
			  *strptr = (char *) wstr;
			  wstr += strsize;
			  strsize *= 2;
			}
		    }
		}
#endif
	      }
	    while ((width <= 0 || --width > 0) && inchar () != EOF);

	    if (!(flags & SUPPRESS))
	      {
		*wstr++ = L'\0';

		if ((flags & MALLOC) && wstr - (wchar_t *) *strptr != strsize)
		  {
		    wchar_t *cp = (wchar_t *) realloc (*strptr,
						       ((wstr
							 - (wchar_t *) *strptr)
							* sizeof(wchar_t)));
		    if (cp != NULL)
		      *strptr = (char *) cp;
		  }

		++done;
	      }
	  }
	  break;

	case L_('x'):	/* Hexadecimal integer.  */
	case L_('X'):	/* Ditto.  */
	  base = 16;
	  number_signed = 0;
	  goto number;

	case L_('o'):	/* Octal integer.  */
	  base = 8;
	  number_signed = 0;
	  goto number;

	case L_('u'):	/* Unsigned decimal integer.  */
	  base = 10;
	  number_signed = 0;
	  goto number;

	case L_('d'):	/* Signed decimal integer.  */
	  base = 10;
	  number_signed = 1;
	  goto number;

	case L_('i'):	/* Generic number.  */
	  base = 0;
	  number_signed = 1;

	number:
	  c = inchar ();
	  if (c == EOF)
	    input_error ();

	  /* Check for a sign.  */
	  if (c == L_('-') || c == L_('+'))
	    {
	      ADDW (c);
	      if (width > 0)
		--width;
	      c = inchar ();
	    }

	  /* Look for a leading indication of base.  */
	  if (width != 0 && c == L_('0'))
	    {
	      if (width > 0)
		--width;

	      ADDW (c);
	      c = inchar ();

	      if (width != 0 && TOLOWER (c) == L_('x'))
		{
		  if (base == 0)
		    base = 16;
		  if (base == 16)
		    {
		      if (width > 0)
			--width;
		      c = inchar ();
		    }
		}
	      else if (base == 0)
		base = 8;
	    }

	  if (base == 0)
	    base = 10;

	  if (base == 10 && (flags & I18N) != 0)
	    {
	      int from_level;
	      int to_level;
	      int level;
#ifdef COMPILE_WSCANF
	      const wchar_t *wcdigits[10];
	      const wchar_t *wcdigits_extended[10];
#else
	      const char *mbdigits[10];
	      const char *mbdigits_extended[10];
#endif
	      /*  "to_inpunct" is a map from ASCII digits to their
		  equivalent in locale. This is defined for locales
		  which use an extra digits set.  */
	      wctrans_t map = __wctrans ("to_inpunct");
	      int n;

	      from_level = 0;
#ifdef COMPILE_WSCANF
	      to_level = _NL_CURRENT_WORD (LC_CTYPE,
					   _NL_CTYPE_INDIGITS_WC_LEN) - 1;
#else
	      to_level = (uint32_t) curctype->values[_NL_ITEM_INDEX (_NL_CTYPE_INDIGITS_MB_LEN)].word - 1;
#endif

	      /* Get the alternative digit forms if there are any.  */
	      if (__builtin_expect (map != NULL, 0))
		{
		  /*  Adding new level for extra digits set in locale file.  */
		  ++to_level;

		  for (n = 0; n < 10; ++n)
		    {
#ifdef COMPILE_WSCANF
		      wcdigits[n] = (const wchar_t *)
                        _NL_CURRENT (LC_CTYPE, _NL_CTYPE_INDIGITS0_WC + n);

		      wchar_t *wc_extended = (wchar_t *)
			alloca ((to_level + 2) * sizeof (wchar_t));
		      __wmemcpy (wc_extended, wcdigits[n], to_level);
		      wc_extended[to_level] = __towctrans (L'0' + n, map);
		      wc_extended[to_level + 1] = '\0';
		      wcdigits_extended[n] = wc_extended;
#else
		      mbdigits[n]
                        = curctype->values[_NL_CTYPE_INDIGITS0_MB + n].string;

		      /*  Get the equivalent wide char in map.  */
		      wint_t extra_wcdigit = __towctrans (L'0' + n, map);

		      /*  Convert it to multibyte representation.  */
		      mbstate_t state;
		      memset (&state, '\0', sizeof (state));

		      char extra_mbdigit[MB_LEN_MAX];
		      size_t mblen
			= __wcrtomb (extra_mbdigit, extra_wcdigit, &state);

		      if (mblen == (size_t) -1)
			{
			  /*  Ignore this new level.  */
			  map = NULL;
			  break;
			}

		      /*  Calculate the length of mbdigits[n].  */
		      const char *last_char = mbdigits[n];
		      for (level = 0; level < to_level; ++level)
			last_char = strchr (last_char, '\0') + 1;

		      size_t mbdigits_len = last_char - mbdigits[n];

		      /*  Allocate memory for extended multibyte digit.  */
		      char *mb_extended;
		      mb_extended = (char *) alloca (mbdigits_len + mblen + 1);