comparison src/regex.c @ 28473:975fe3d8922e

* regex.c (PTR_TO_OFFSET) [!emacs]: Remove. (RE_MULTIBYTE_P, RE_STRING_CHAR_AND_LENGTH): New macros. (GET_CHAR_BEFORE_2): Moved from charset.h plus fixed minor bug when we are between str1 and str2. (MAX_MULTIBYTE_LENGTH, CHAR_STRING) [!emacs]: Provide trivial default. (PATFETCH): Use `TRANSLATE'. (PATFETCH_RAW): Fetch multibyte char if applicable. (PATUNFETCH): Remove. (regex_compile): Rely on PATFETCH to do most of the multibyte magic. When writing a char, write it directly into the pattern buffer rather than going needlessly through a temp char-array. (re_match_2_internal): Similarly, rely on RE_STRING_CHAR to do the multibyte magic and remove the useless `#ifdef emacs'. (bcmp_translate): Don't compare as multibyte chars when in a unibyte buffer. * regex.h (struct re_pattern_buffer): Make field `multibyte' conditional on `emacs'. * charset.h (GET_CHAR_BEFORE_2): Moved to regex.c.

28472:bae9218986ac

   along with this program; if not, write to the Free Software
   Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307,
   USA.	 */

/* TODO:
   - clean up multibyte issues
   - structure the opcode space into opcode+flag.
   - merge with glibc's regex.[ch]
 */

/* AIX requires this to be the first thing in the file. */

#ifdef emacs
/* Converts the pointer to the char to BEG-based offset from the start.	 */
#define PTR_TO_OFFSET(d) POS_AS_IN_BUFFER (POINTER_TO_OFFSET (d))
#define POS_AS_IN_BUFFER(p) ((p) + (NILP (re_match_object) || BUFFERP (re_match_object)))
#else
#define PTR_TO_OFFSET(d) 0
#endif

#ifdef HAVE_CONFIG_H
#include <config.h>
#endif

#define malloc xmalloc
#define realloc xrealloc
#define free xfree

#define RE_STRING_CHAR(p, s) \
  (multibyte ? (STRING_CHAR (p, s)) : (*(p)))

#else  /* not emacs */

/* If we are not linking with Emacs proper,
   we can't use the relocating allocator

/* Dummy macros for non-Emacs environments.  */
#define BASE_LEADING_CODE_P(c) (0)
#define CHAR_CHARSET(c) 0
#define CHARSET_LEADING_CODE_BASE(c) 0
#define WORD_BOUNDARY_P(c1, c2) (0)
#define CHAR_HEAD_P(p) (1)
#define SINGLE_BYTE_CHAR_P(c) (1)
#define SAME_CHARSET_P(c1, c2) (1)
#define MULTIBYTE_FORM_LENGTH(p, s) (1)
#define STRING_CHAR(p, s) (*(p))
#define RE_STRING_CHAR STRING_CHAR
#define STRING_CHAR_AND_LENGTH(p, s, actual_len) ((actual_len) = 1, *(p))
#define GET_CHAR_BEFORE_2(c, p, str1, end1, str2, end2) \
  (c = ((p) == (str2) ? *((end1) - 1) : *((p) - 1)))
#endif /* not emacs */

#ifndef RE_TRANSLATE

   string passed to us by the user to an unsigned char that we can use
   as an array index (in, e.g., `translate').  */
#define PATFETCH(c)							\
  do {									\
    PATFETCH_RAW (c);							\
    if (RE_TRANSLATE_P (translate)) c = RE_TRANSLATE (translate, c);	\
  } while (0)

/* Fetch the next character in the uncompiled pattern, with no
   translation.	 */
#define PATFETCH_RAW(c)							\
  do {if (p == pend) return REG_EEND;					\
    c = *p++;								\
  } while (0)

/* Go backwards one character in the pattern.  */
#define PATUNFETCH p--


/* If `translate' is non-null, return translate[D], else just D.  We
   cast the subscript to translate because some data is declared as
   `char *', to avoid warnings when a string constant is passed.  But

  regnum_t regnum = 0;

  /* Work area for range table of charset.  */
  struct range_table_work_area range_table_work;

#ifdef DEBUG
  debug++;
  DEBUG_PRINT1 ("\nCompiling pattern: ");
  if (debug > 0)
    {

     at the end.  */
  bufp->used = 0;

  /* Always count groups, whether or not bufp->no_sub is set.  */
  bufp->re_nsub = 0;

#ifdef emacs
  /* bufp->multibyte is set before regex_compile is called, so don't alter
     it. */
#else  /* not emacs */
  /* Nothing is recognized as a multibyte character.  */
  bufp->multibyte = 0;
#endif

#if !defined (emacs) && !defined (SYNTAX_TABLE)
  /* Initialize the syntax table.  */
   init_syntax_once ();
#endif

	      SET_LIST_BIT ('\n');

	    /* Read in characters and ranges, setting map bits.	 */
	    for (;;)
	      {
		int len;
		boolean escaped_char = false;

		if (p == pend) FREE_STACK_RETURN (REG_EBRACK);

		PATFETCH (c);

		else
		  {
		    /* Could be the end of the bracket expression.	If it's
		       not (i.e., when the bracket expression is `[]' so
		       far), the ']' character bit gets set way below.  */
		    if (c == ']' && p != p1 + 1)
		      break;
		  }

		/* If C indicates start of multibyte char, get the
		   actual character code in C, and set the pattern
		   pointer P to the next character boundary.  */
		if (bufp->multibyte && BASE_LEADING_CODE_P (c))
		  {
		    PATUNFETCH;
		    c = STRING_CHAR_AND_LENGTH (p, pend - p, len);
		    p += len;
		  }
		/* What should we do for the character which is
		   greater than 0x7F, but not BASE_LEADING_CODE_P?
		   XXX */

		/* See if we're at the beginning of a possible character
		   class.  */

		else if (!escaped_char &&
			 syntax & RE_CHAR_CLASSES && c == '[' && *p == ':')
		  {
		    /* Leave room for the null.	 */
		    char str[CHAR_CLASS_MAX_LENGTH + 1];
		    const unsigned char *class_beg;

			   just set a flag.  Exceptions are is_blank,
			   is_digit, is_cntrl, and is_xdigit, since
			   they can only match ASCII characters.  We
			   don't need to handle them for multibyte.  */

			if (bufp->multibyte)
			  {
			    int bit = 0;

			    if (is_alnum) bit = BIT_ALNUM;
			    if (is_alpha) bit = BIT_ALPHA;

		    /* Discard the `-'. */
		    PATFETCH (c1);

		    /* Fetch the character which ends the range. */
		    PATFETCH (c1);
		    if (bufp->multibyte && BASE_LEADING_CODE_P (c1))
		      {
			PATUNFETCH;
			c1 = STRING_CHAR_AND_LENGTH (p, pend - p, len);
			p += len;
		      }

		    if (SINGLE_BYTE_CHAR_P (c)
			&& ! SINGLE_BYTE_CHAR_P (c1))
		      {
			/* Handle a range such as \177-\377 in multibyte mode.

	default:
	/* Expects the character in `c'.  */
	normal_char:
	  p1 = p - 1;		/* P1 points the head of C.  */
#ifdef emacs
	  if (bufp->multibyte)
	    {
	      c = STRING_CHAR (p1, pend - p1);
	      c = TRANSLATE (c);
	      /* Set P to the next character boundary.  */
	      p += MULTIBYTE_FORM_LENGTH (p1, pend - p1) - 1;
	    }
#endif
	      /* If no exactn currently being built.  */
	  if (!pending_exact

	      /* If last exactn not at current position.  */
	      || pending_exact + *pending_exact + 1 != b

	      /* We have only one byte following the exactn for the count.  */
	      || *pending_exact >= (1 << BYTEWIDTH) - (p - p1)

	      /* If followed by a repetition operator.	*/
	      || (p != pend && (*p == '*' || *p == '^'))
	      || ((syntax & RE_BK_PLUS_QM)
		  ? p + 1 < pend && *p == '\\' && (p[1] == '+' || p[1] == '?')

	      BUF_PUSH_2 (exactn, 0);
	      pending_exact = b - 1;
	    }

#ifdef emacs
	  if (! SINGLE_BYTE_CHAR_P (c))
	    {
	      unsigned char str[MAX_MULTIBYTE_LENGTH];
	      int i = CHAR_STRING (c, str);
	      int j;
	      for (j = 0; j < i; j++)
		{
		  BUF_PUSH (str[j]);
		  (*pending_exact)++;
		}
	    }
	  else
#endif
	    {
	      BUF_PUSH (c);
	      (*pending_exact)++;
	    }
	  break;
	} /* switch (c) */
    } /* while p != pend */

  bzero (fastmap, 1 << BYTEWIDTH);  /* Assume nothing's valid.	*/
  bufp->fastmap_accurate = 1;	    /* It will be when we're done.  */

  analysis = analyse_first (bufp->buffer, bufp->buffer + bufp->used,
			    fastmap, bufp->multibyte);
  if (analysis < -1)
    return analysis;
  bufp->can_be_null = (analysis != 0);
  return 0;
} /* re_compile_fastmap */

  int total_size = size1 + size2;
  int endpos = startpos + range;
  int anchored_start = 0;

  /* Nonzero if we have to concern multibyte character.	 */
  const boolean multibyte = bufp->multibyte;

  /* Check for out-of-range STARTPOS.  */
  if (startpos < 0 || startpos > total_size)
    return -1;

	      startpos += irange - range;
	    }
	  else				/* Searching backwards.	 */
	    {
	      buf_ch = STRING_CHAR (d, (startpos >= size1
					? size2 + size1 - startpos
					: size1 - startpos));
	      if (RE_TRANSLATE_P (translate))
		buf_ch = RE_TRANSLATE (translate, buf_ch);

	      if (! (buf_ch >= 0400
		     || fastmap[buf_ch]))
		goto advance;
	    }

  return -1;
} /* re_search_2 */

/* Declarations and macros for re_match_2.  */

static int bcmp_translate ();

/* This converts PTR, a pointer into one of the search strings `string1'
   and `string2' into an offset from the beginning of that string.  */
#define POINTER_TO_OFFSET(ptr)			\
  (FIRST_STRING_P (ptr)				\

mutually_exclusive_p (bufp, p1, p2)
     struct re_pattern_buffer *bufp;
     unsigned char *p1, *p2;
{
  re_opcode_t op2;
  const boolean multibyte = bufp->multibyte;
  unsigned char *pend = bufp->buffer + bufp->used;

  assert (p1 >= bufp->buffer && p1 < pend
	  && p2 >= bufp->buffer && p2 <= pend);

  /* We use this to map every character in the string.	*/
  RE_TRANSLATE_TYPE translate = bufp->translate;

  /* Nonzero if we have to concern multibyte character.	 */
  const boolean multibyte = bufp->multibyte;

  /* Failure point stack.  Each place that can handle a failure further
     down the line pushes a failure point on this stack.  It consists of
     regstart, and regend for all registers corresponding to
     the subexpressions we're currently inside, plus the number of such

	  /* This is written out as an if-else so we don't waste time
	     testing `translate' inside the loop.  */
	  if (RE_TRANSLATE_P (translate))
	    {
#ifdef emacs
	      if (multibyte)
		do
		  {
		    int pat_charlen, buf_charlen;
		    unsigned int pat_ch, buf_ch;

		    d += buf_charlen;
		    mcnt -= pat_charlen;
		  }
		while (mcnt > 0);
	      else
#endif /* not emacs */
		do
		  {
		    PREFETCH ();
		    if (RE_TRANSLATE (translate, *d) != *p++)
		      {

	    unsigned int buf_ch;

	    DEBUG_PRINT1 ("EXECUTING anychar.\n");

	    PREFETCH ();

#ifdef emacs
	    if (multibyte)
	      buf_ch = STRING_CHAR_AND_LENGTH (d, dend - d, buf_charlen);
	    else
#endif /* not emacs */
	      {
		buf_ch = *d;
		buf_charlen = 1;
	      }

	    buf_ch = TRANSLATE (buf_ch);

	    if ((!(bufp->syntax & RE_DOT_NEWLINE)
		 && buf_ch == '\n')
		|| ((bufp->syntax & RE_DOT_NOT_NULL)

	       in the initial byte-length of the command.  */
	    int count = 0;

	    DEBUG_PRINT2 ("EXECUTING charset%s.\n", not ? "_not" : "");

	    PREFETCH ();
	    c = *d;

	    range_table_exists = CHARSET_RANGE_TABLE_EXISTS_P (&p[-1]);

#ifdef emacs
	    if (range_table_exists)
	      {
		range_table = CHARSET_RANGE_TABLE (&p[-1]); /* Past the bitmap.  */
		EXTRACT_NUMBER_AND_INCR (count, range_table);
	      }

	    if (multibyte && BASE_LEADING_CODE_P (c))
	      c = STRING_CHAR_AND_LENGTH (d, dend - d, len);
#endif /* emacs */

	    if (SINGLE_BYTE_CHAR_P (c))
	      {			/* Lookup bitmap.  */
		c = TRANSLATE (c); /* The character to match.  */
		len = 1;

		/* Cast to `unsigned' instead of `unsigned char' in
		   case the bit list is a full 32 bytes long.  */
		if (c < (unsigned) (CHARSET_BITMAP_SIZE (&p[-1]) * BYTEWIDTH)
		    && p[1 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH)))
		  not = !not;

		  mcnt = dend2 - d2;

		/* Compare that many; failure if mismatch, else move
		   past them.  */
		if (RE_TRANSLATE_P (translate)
		    ? bcmp_translate (d, d2, mcnt, translate)
		    : bcmp (d, d2, mcnt))
		  {
		    d = dfail;
		    goto fail;
		  }

	    {
	      /* C1 is the character before D, S1 is the syntax of C1, C2
		 is the character at D, and S2 is the syntax of C2.  */
	      int c1, c2, s1, s2;
#ifdef emacs
	      int charpos = SYNTAX_TABLE_BYTE_TO_CHAR (PTR_TO_OFFSET (d - 1));
	      UPDATE_SYNTAX_TABLE (charpos);
#endif
	      /* FIXME: This does a STRING_CHAR even for unibyte buffers.  */
	      GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2);
	      s1 = SYNTAX (c1);
#ifdef emacs
	      UPDATE_SYNTAX_TABLE_FORWARD (charpos + 1);
#endif
	      PREFETCH ();
	      /* FIXME: This does a STRING_CHAR even for unibyte buffers.  */
	      c2 = STRING_CHAR (d, dend - d);
	      s2 = SYNTAX (c2);

	      if (/* Case 2: Only one of S1 and S2 is Sword.  */
		  ((s1 == Sword) != (s2 == Sword))
		  /* Case 3: Both of S1 and S2 are Sword, and macro

	    {
	      /* C1 is the character before D, S1 is the syntax of C1, C2
		 is the character at D, and S2 is the syntax of C2.  */
	      int c1, c2, s1, s2;
#ifdef emacs
	      int charpos = SYNTAX_TABLE_BYTE_TO_CHAR (PTR_TO_OFFSET (d));
	      UPDATE_SYNTAX_TABLE (charpos);
#endif
	      PREFETCH ();
	      /* FIXME: This does a STRING_CHAR even for unibyte buffers.  */
	      c2 = STRING_CHAR (d, dend - d);
	      s2 = SYNTAX (c2);
	
	      /* Case 2: S2 is not Sword. */
	      if (s2 != Sword)
		goto fail;

	    {
	      /* C1 is the character before D, S1 is the syntax of C1, C2
		 is the character at D, and S2 is the syntax of C2.  */
	      int c1, c2, s1, s2;
#ifdef emacs
	      int charpos = SYNTAX_TABLE_BYTE_TO_CHAR (PTR_TO_OFFSET (d) - 1);
	      UPDATE_SYNTAX_TABLE (charpos);
#endif
	      GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2);
	      s1 = SYNTAX (c1);

	      /* Case 3: D is not at the end of string ... */
	      if (!AT_STRINGS_END (d))
		{
		  PREFETCH ();
		  /* FIXME: This does a STRING_CHAR even for unibyte buffers.  */
		  c2 = STRING_CHAR (d, dend - d);
#ifdef emacs
		  UPDATE_SYNTAX_TABLE_FORWARD (charpos);
#endif
		  s2 = SYNTAX (c2);

	  mcnt = *p++;
	  DEBUG_PRINT3 ("EXECUTING %ssyntaxspec %d.\n", not?"not":"", mcnt);
	  PREFETCH ();
#ifdef emacs
	  {
	    int pos1 = SYNTAX_TABLE_BYTE_TO_CHAR (PTR_TO_OFFSET (d));
	    UPDATE_SYNTAX_TABLE (pos1);
	  }
#endif
	  {
	    int c, len;

	    if (multibyte)
	      /* we must concern about multibyte form, ... */
	      c = STRING_CHAR_AND_LENGTH (d, dend - d, len);
	    else
	      /* everything should be handled as ASCII, even though it
		 looks like multibyte form.  */
	      c = *d, len = 1;

	    if ((SYNTAX (c) != (enum syntaxcode) mcnt) ^ not)
	      goto fail;
	    d += len;
	  }

	  mcnt = *p++;
	  DEBUG_PRINT3 ("EXECUTING %scategoryspec %d.\n", not?"not":"", mcnt);
	  PREFETCH ();
	  {
	    int c, len;

	    if (multibyte)
	      c = STRING_CHAR_AND_LENGTH (d, dend - d, len);
	    else
	      c = *d, len = 1;

	    if ((!CHAR_HAS_CATEGORY (c, mcnt)) ^ not)
	      goto fail;
	    d += len;
	  }

/* Return zero if TRANSLATE[S1] and TRANSLATE[S2] are identical for LEN
   bytes; nonzero otherwise.  */

static int
bcmp_translate (s1, s2, len, translate)
     unsigned char *s1, *s2;
     register int len;
     RE_TRANSLATE_TYPE translate;
{
  register unsigned char *p1 = s1, *p2 = s2;
  unsigned char *p1_end = s1 + len;
  unsigned char *p2_end = s2 + len;

  while (p1 != p1_end && p2 != p2_end)
    {
      int p1_charlen, p2_charlen;
      int p1_ch, p2_ch;

      /* FIXME: This assumes `multibyte = true'.  */
      p1_ch = STRING_CHAR_AND_LENGTH (p1, p1_end - p1, p1_charlen);
      p2_ch = STRING_CHAR_AND_LENGTH (p2, p2_end - p2, p2_charlen);

      if (RE_TRANSLATE (translate, p1_ch)
	  != RE_TRANSLATE (translate, p2_ch))
	return 1;

28473:975fe3d8922e

   along with this program; if not, write to the Free Software
   Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307,
   USA.	 */

/* TODO:
   - structure the opcode space into opcode+flag.
   - merge with glibc's regex.[ch]
 */

/* AIX requires this to be the first thing in the file. */

#ifdef emacs
/* Converts the pointer to the char to BEG-based offset from the start.	 */
#define PTR_TO_OFFSET(d) POS_AS_IN_BUFFER (POINTER_TO_OFFSET (d))
#define POS_AS_IN_BUFFER(p) ((p) + (NILP (re_match_object) || BUFFERP (re_match_object)))
#endif

#ifdef HAVE_CONFIG_H
#include <config.h>
#endif

#define malloc xmalloc
#define realloc xrealloc
#define free xfree

#define RE_MULTIBYTE_P(bufp) ((bufp)->multibyte)
#define RE_STRING_CHAR(p, s) \
  (multibyte ? (STRING_CHAR (p, s)) : (*(p)))
#define RE_STRING_CHAR_AND_LENGTH(p, s, len) \
  (multibyte ? (STRING_CHAR_AND_LENGTH (p, s, len)) : ((len) = 1, *(p)))

/* Set C a (possibly multibyte) character before P.  P points into a
   string which is the virtual concatenation of STR1 (which ends at
   END1) or STR2 (which ends at END2).  */
#define GET_CHAR_BEFORE_2(c, p, str1, end1, str2, end2)			\
  do {									\
    if (multibyte)							\
       {						    		\
         re_char *dtemp = (p) == (str2) ? (end1) : (p);		    	\
         re_char *dlimit = ((p) > (str2) && (p) <= (end2)) ? (str2) : (str1); \
         while (dtemp-- > dlimit && !CHAR_HEAD_P (*dtemp));		\
         c = STRING_CHAR (dtemp, (p) - dtemp);				\
       }						    		\
     else						    		\
       (c = ((p) == (str2) ? (end1) : (p))[-1]);			\
  } while (0)


#else  /* not emacs */

/* If we are not linking with Emacs proper,
   we can't use the relocating allocator

/* Dummy macros for non-Emacs environments.  */
#define BASE_LEADING_CODE_P(c) (0)
#define CHAR_CHARSET(c) 0
#define CHARSET_LEADING_CODE_BASE(c) 0
#define MAX_MULTIBYTE_LENGTH 1
#define RE_MULTIBYTE_P(x) 0
#define WORD_BOUNDARY_P(c1, c2) (0)
#define CHAR_HEAD_P(p) (1)
#define SINGLE_BYTE_CHAR_P(c) (1)
#define SAME_CHARSET_P(c1, c2) (1)
#define MULTIBYTE_FORM_LENGTH(p, s) (1)
#define STRING_CHAR(p, s) (*(p))
#define RE_STRING_CHAR STRING_CHAR
#define CHAR_STRING(c, s) (*(s) = (c), 1)
#define STRING_CHAR_AND_LENGTH(p, s, actual_len) ((actual_len) = 1, *(p))
#define RE_STRING_CHAR_AND_LENGTH STRING_CHAR_AND_LENGTH
#define GET_CHAR_BEFORE_2(c, p, str1, end1, str2, end2) \
  (c = ((p) == (str2) ? *((end1) - 1) : *((p) - 1)))
#endif /* not emacs */

#ifndef RE_TRANSLATE

   string passed to us by the user to an unsigned char that we can use
   as an array index (in, e.g., `translate').  */
#define PATFETCH(c)							\
  do {									\
    PATFETCH_RAW (c);							\
    c = TRANSLATE (c);							\
  } while (0)

/* Fetch the next character in the uncompiled pattern, with no
   translation.	 */
#define PATFETCH_RAW(c)							\
  do {									\
    int len;								\
    if (p == pend) return REG_EEND;					\
    c = RE_STRING_CHAR_AND_LENGTH (p, pend - p, len);			\
    p += len;								\
  } while (0)


/* If `translate' is non-null, return translate[D], else just D.  We
   cast the subscript to translate because some data is declared as
   `char *', to avoid warnings when a string constant is passed.  But

  regnum_t regnum = 0;

  /* Work area for range table of charset.  */
  struct range_table_work_area range_table_work;

  /* If the object matched can contain multibyte characters.  */
  const boolean multibyte = RE_MULTIBYTE_P (bufp);

#ifdef DEBUG
  debug++;
  DEBUG_PRINT1 ("\nCompiling pattern: ");
  if (debug > 0)
    {

     at the end.  */
  bufp->used = 0;

  /* Always count groups, whether or not bufp->no_sub is set.  */
  bufp->re_nsub = 0;

#if !defined (emacs) && !defined (SYNTAX_TABLE)
  /* Initialize the syntax table.  */
   init_syntax_once ();
#endif

	      SET_LIST_BIT ('\n');

	    /* Read in characters and ranges, setting map bits.	 */
	    for (;;)
	      {
		boolean escaped_char = false;
		const unsigned char *p2 = p;

		if (p == pend) FREE_STACK_RETURN (REG_EBRACK);

		PATFETCH (c);

		else
		  {
		    /* Could be the end of the bracket expression.	If it's
		       not (i.e., when the bracket expression is `[]' so
		       far), the ']' character bit gets set way below.  */
		    if (c == ']' && p2 != p1)
		      break;
		  }

		/* What should we do for the character which is
		   greater than 0x7F, but not BASE_LEADING_CODE_P?
		   XXX */

		/* See if we're at the beginning of a possible character
		   class.  */

		if (!escaped_char &&
		    syntax & RE_CHAR_CLASSES && c == '[' && *p == ':')
		  {
		    /* Leave room for the null.	 */
		    char str[CHAR_CLASS_MAX_LENGTH + 1];
		    const unsigned char *class_beg;

			   just set a flag.  Exceptions are is_blank,
			   is_digit, is_cntrl, and is_xdigit, since
			   they can only match ASCII characters.  We
			   don't need to handle them for multibyte.  */

			if (multibyte)
			  {
			    int bit = 0;

			    if (is_alnum) bit = BIT_ALNUM;
			    if (is_alpha) bit = BIT_ALPHA;

		    /* Discard the `-'. */
		    PATFETCH (c1);

		    /* Fetch the character which ends the range. */
		    PATFETCH (c1);

		    if (SINGLE_BYTE_CHAR_P (c)
			&& ! SINGLE_BYTE_CHAR_P (c1))
		      {
			/* Handle a range such as \177-\377 in multibyte mode.

	default:
	/* Expects the character in `c'.  */
	normal_char:
	      /* If no exactn currently being built.  */
	  if (!pending_exact

	      /* If last exactn not at current position.  */
	      || pending_exact + *pending_exact + 1 != b

	      /* We have only one byte following the exactn for the count.  */
	      || *pending_exact >= (1 << BYTEWIDTH) - MAX_MULTIBYTE_LENGTH

	      /* If followed by a repetition operator.	*/
	      || (p != pend && (*p == '*' || *p == '^'))
	      || ((syntax & RE_BK_PLUS_QM)
		  ? p + 1 < pend && *p == '\\' && (p[1] == '+' || p[1] == '?')

	      BUF_PUSH_2 (exactn, 0);
	      pending_exact = b - 1;
	    }

	  GET_BUFFER_SPACE (MAX_MULTIBYTE_LENGTH);
	  {
	    int len = CHAR_STRING (c, b);
	    b += len;
	    (*pending_exact) += len;
	  }

	  break;
	} /* switch (c) */
    } /* while p != pend */

  bzero (fastmap, 1 << BYTEWIDTH);  /* Assume nothing's valid.	*/
  bufp->fastmap_accurate = 1;	    /* It will be when we're done.  */

  analysis = analyse_first (bufp->buffer, bufp->buffer + bufp->used,
			    fastmap, RE_MULTIBYTE_P (bufp));
  if (analysis < -1)
    return analysis;
  bufp->can_be_null = (analysis != 0);
  return 0;
} /* re_compile_fastmap */

  int total_size = size1 + size2;
  int endpos = startpos + range;
  int anchored_start = 0;

  /* Nonzero if we have to concern multibyte character.	 */
  const boolean multibyte = RE_MULTIBYTE_P (bufp);

  /* Check for out-of-range STARTPOS.  */
  if (startpos < 0 || startpos > total_size)
    return -1;

	      startpos += irange - range;
	    }
	  else				/* Searching backwards.	 */
	    {
	      int room = (startpos >= size1
			  ? size2 + size1 - startpos
			  : size1 - startpos);
	      buf_ch = RE_STRING_CHAR (d, room);
	      buf_ch = TRANSLATE (buf_ch);

	      if (! (buf_ch >= 0400
		     || fastmap[buf_ch]))
		goto advance;
	    }

  return -1;
} /* re_search_2 */

/* Declarations and macros for re_match_2.  */

static int bcmp_translate _RE_ARGS((re_char *s1, re_char *s2,
				    register int len,
				    RE_TRANSLATE_TYPE translate,
				    const int multibyte));

/* This converts PTR, a pointer into one of the search strings `string1'
   and `string2' into an offset from the beginning of that string.  */
#define POINTER_TO_OFFSET(ptr)			\
  (FIRST_STRING_P (ptr)				\

mutually_exclusive_p (bufp, p1, p2)
     struct re_pattern_buffer *bufp;
     unsigned char *p1, *p2;
{
  re_opcode_t op2;
  const boolean multibyte = RE_MULTIBYTE_P (bufp);
  unsigned char *pend = bufp->buffer + bufp->used;

  assert (p1 >= bufp->buffer && p1 < pend
	  && p2 >= bufp->buffer && p2 <= pend);

  /* We use this to map every character in the string.	*/
  RE_TRANSLATE_TYPE translate = bufp->translate;

  /* Nonzero if we have to concern multibyte character.	 */
  const boolean multibyte = RE_MULTIBYTE_P (bufp);

  /* Failure point stack.  Each place that can handle a failure further
     down the line pushes a failure point on this stack.  It consists of
     regstart, and regend for all registers corresponding to
     the subexpressions we're currently inside, plus the number of such

	  /* This is written out as an if-else so we don't waste time
	     testing `translate' inside the loop.  */
	  if (RE_TRANSLATE_P (translate))
	    {
	      if (multibyte)
		do
		  {
		    int pat_charlen, buf_charlen;
		    unsigned int pat_ch, buf_ch;

		    d += buf_charlen;
		    mcnt -= pat_charlen;
		  }
		while (mcnt > 0);
	      else
		do
		  {
		    PREFETCH ();
		    if (RE_TRANSLATE (translate, *d) != *p++)
		      {

	    unsigned int buf_ch;

	    DEBUG_PRINT1 ("EXECUTING anychar.\n");

	    PREFETCH ();
	    buf_ch = RE_STRING_CHAR_AND_LENGTH (d, dend - d, buf_charlen);
	    buf_ch = TRANSLATE (buf_ch);

	    if ((!(bufp->syntax & RE_DOT_NEWLINE)
		 && buf_ch == '\n')
		|| ((bufp->syntax & RE_DOT_NOT_NULL)

	       in the initial byte-length of the command.  */
	    int count = 0;

	    DEBUG_PRINT2 ("EXECUTING charset%s.\n", not ? "_not" : "");

	    range_table_exists = CHARSET_RANGE_TABLE_EXISTS_P (&p[-1]);

	    if (range_table_exists)
	      {
		range_table = CHARSET_RANGE_TABLE (&p[-1]); /* Past the bitmap.  */
		EXTRACT_NUMBER_AND_INCR (count, range_table);
	      }

	    PREFETCH ();
	    c = RE_STRING_CHAR_AND_LENGTH (d, dend - d, len);
	    c = TRANSLATE (c); /* The character to match.  */

	    if (SINGLE_BYTE_CHAR_P (c))
	      {			/* Lookup bitmap.  */
		/* Cast to `unsigned' instead of `unsigned char' in
		   case the bit list is a full 32 bytes long.  */
		if (c < (unsigned) (CHARSET_BITMAP_SIZE (&p[-1]) * BYTEWIDTH)
		    && p[1 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH)))
		  not = !not;

		  mcnt = dend2 - d2;

		/* Compare that many; failure if mismatch, else move
		   past them.  */
		if (RE_TRANSLATE_P (translate)
		    ? bcmp_translate (d, d2, mcnt, translate, multibyte)
		    : bcmp (d, d2, mcnt))
		  {
		    d = dfail;
		    goto fail;
		  }

	    {
	      /* C1 is the character before D, S1 is the syntax of C1, C2
		 is the character at D, and S2 is the syntax of C2.  */
	      int c1, c2, s1, s2;
#ifdef emacs
	      int offset = PTR_TO_OFFSET (d - 1);
	      int charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset);
	      UPDATE_SYNTAX_TABLE (charpos);
#endif
	      GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2);
	      s1 = SYNTAX (c1);
#ifdef emacs
	      UPDATE_SYNTAX_TABLE_FORWARD (charpos + 1);
#endif
	      PREFETCH ();
	      c2 = RE_STRING_CHAR (d, dend - d);
	      s2 = SYNTAX (c2);

	      if (/* Case 2: Only one of S1 and S2 is Sword.  */
		  ((s1 == Sword) != (s2 == Sword))
		  /* Case 3: Both of S1 and S2 are Sword, and macro

	    {
	      /* C1 is the character before D, S1 is the syntax of C1, C2
		 is the character at D, and S2 is the syntax of C2.  */
	      int c1, c2, s1, s2;
#ifdef emacs
	      int offset = PTR_TO_OFFSET (d);
	      int charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset);
	      UPDATE_SYNTAX_TABLE (charpos);
#endif
	      PREFETCH ();
	      c2 = RE_STRING_CHAR (d, dend - d);
	      s2 = SYNTAX (c2);
	
	      /* Case 2: S2 is not Sword. */
	      if (s2 != Sword)
		goto fail;

	    {
	      /* C1 is the character before D, S1 is the syntax of C1, C2
		 is the character at D, and S2 is the syntax of C2.  */
	      int c1, c2, s1, s2;
#ifdef emacs
	      int offset = PTR_TO_OFFSET (d) - 1;
	      int charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset);
	      UPDATE_SYNTAX_TABLE (charpos);
#endif
	      GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2);
	      s1 = SYNTAX (c1);

	      /* Case 3: D is not at the end of string ... */
	      if (!AT_STRINGS_END (d))
		{
		  PREFETCH ();
		  c2 = RE_STRING_CHAR (d, dend - d);
#ifdef emacs
		  UPDATE_SYNTAX_TABLE_FORWARD (charpos);
#endif
		  s2 = SYNTAX (c2);

	  mcnt = *p++;
	  DEBUG_PRINT3 ("EXECUTING %ssyntaxspec %d.\n", not?"not":"", mcnt);
	  PREFETCH ();
#ifdef emacs
	  {
	    int offset = PTR_TO_OFFSET (d);
	    int pos1 = SYNTAX_TABLE_BYTE_TO_CHAR (offset);
	    UPDATE_SYNTAX_TABLE (pos1);
	  }
#endif
	  {
	    int c, len;

	    c = RE_STRING_CHAR_AND_LENGTH (d, dend - d, len);

	    if ((SYNTAX (c) != (enum syntaxcode) mcnt) ^ not)
	      goto fail;
	    d += len;
	  }

	  mcnt = *p++;
	  DEBUG_PRINT3 ("EXECUTING %scategoryspec %d.\n", not?"not":"", mcnt);
	  PREFETCH ();
	  {
	    int c, len;
	    c = RE_STRING_CHAR_AND_LENGTH (d, dend - d, len);

	    if ((!CHAR_HAS_CATEGORY (c, mcnt)) ^ not)
	      goto fail;
	    d += len;
	  }

/* Return zero if TRANSLATE[S1] and TRANSLATE[S2] are identical for LEN
   bytes; nonzero otherwise.  */

static int
bcmp_translate (s1, s2, len, translate, multibyte)
     re_char *s1, *s2;
     register int len;
     RE_TRANSLATE_TYPE translate;
     const int multibyte;
{
  register re_char *p1 = s1, *p2 = s2;
  re_char *p1_end = s1 + len;
  re_char *p2_end = s2 + len;

  while (p1 != p1_end && p2 != p2_end)
    {
      int p1_charlen, p2_charlen;
      int p1_ch, p2_ch;

      p1_ch = RE_STRING_CHAR_AND_LENGTH (p1, p1_end - p1, p1_charlen);
      p2_ch = RE_STRING_CHAR_AND_LENGTH (p2, p2_end - p2, p2_charlen);

      if (RE_TRANSLATE (translate, p1_ch)
	  != RE_TRANSLATE (translate, p2_ch))
	return 1;