/[pcre]/code/trunk/pcre_compile.c

Diff of /code/trunk/pcre_compile.c

Parent Directory | Revision Log | View Patch Patch

-revision 773 by ph10,
Wed Nov 30 18:10:27 2011 UTC
+revision 1305 by ph10,
Mon Apr  1 14:50:45 2013 UTC
 Line 6
  and semantics are as close as possible to those of the Perl 5 language.
                         Written by Philip Hazel
-            Copyright (c) 1997-2011 University of Cambridge
+            Copyright (c) 1997-2012 University of Cambridge
  -----------------------------------------------------------------------------
  Redistribution and use in source and binary forms, with or without
 Line 53 
 supporting internal functions that are n
  #include "pcre_internal.h"
- /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
+ /* When PCRE_DEBUG is defined, we need the pcre(16|32)_printint() function, which
- also used by pcretest. PCRE_DEBUG is not defined when building a production
+ is also used by pcretest. PCRE_DEBUG is not defined when building a production
- library. */
+ library. We do not need to select pcre16_printint.c specially, because the
+ COMPILE_PCREx macro will already be appropriately set. */
  #ifdef PCRE_DEBUG
- #include "pcre_printint.src"
+ /* pcre_printint.c should not include any headers */
+ #define PCRE_INCLUDED
+ #include "pcre_printint.c"
+ #undef PCRE_INCLUDED
  #endif
  /* Macro for setting individual bits in class bitmaps. */
- #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
+ #define SETBIT(a,b) a[(b)/8] |= (1 << ((b)&7))
  /* Maximum length value to check against when making sure that the integer that
  holds the compiled pattern length does not overflow. We make it a bit less than
-Line 73 
 to check them every time. */
+Line 77 
 to check them every time. */
  #define OFLOW_MAX (INT_MAX - 20)
+ /* Definitions to allow mutual recursion */
+ static int
+   add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
+     const pcre_uint32 *, unsigned int);
+ static BOOL
+   compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
+     pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
+     compile_data *, int *);
  /*************************************************
  *      Code parameters and static tables         *
-Line 88 
 so this number is very generous.
+Line 104 
 so this number is very generous.
  The same workspace is used during the second, actual compile phase for
  remembering forward references to groups so that they can be filled in at the
  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
  is 4 there is plenty of room for most patterns. However, the memory can get
  filled up by repetitions of forward references, for example patterns like
  /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
  that the workspace is expanded using malloc() in this situation. The value
  below is therefore a minimum, and we put a maximum on it for safety. The
  minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
  kicks in at the same number of forward references in all cases. */
  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
-Line 104 
 overrun before it actually does run off
+Line 120 
 overrun before it actually does run off
  #define WORK_SIZE_SAFETY_MARGIN (100)
+ /* Private flags added to firstchar and reqchar. */
+ #define REQ_CASELESS    (1 << 0)        /* Indicates caselessness */
+ #define REQ_VARY        (1 << 1)        /* Reqchar followed non-literal item */
+ /* Negative values for the firstchar and reqchar flags */
+ #define REQ_UNSET       (-2)
+ #define REQ_NONE        (-1)
+ /* Repeated character flags. */
+ #define UTF_LENGTH     0x10000000l      /* The char contains its length. */
  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
  are simple data values; negative values are for special things like \d and so
-Line 238 
 static const char posix_names[] =
+Line 265 
 static const char posix_names[] =
    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
    STRING_word0  STRING_xdigit;
- static const uschar posix_name_lengths[] = {
+ static const pcre_uint8 posix_name_lengths[] = {
 , 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
  /* Table of class bit maps for each POSIX class. Each class is formed from a
-Line 273 
 substitutes must be in the order of the
+Line 300 
 substitutes must be in the order of the
  both positive and negative cases. NULL means no substitute. */
  #ifdef SUPPORT_UCP
- static const uschar *substitutes[] = {
+ static const pcre_uchar string_PNd[]  = {
-   (uschar *)"\\P{Nd}",    /* \D */
+   CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
-   (uschar *)"\\p{Nd}",    /* \d */
+   CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
-   (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */
+ static const pcre_uchar string_pNd[]  = {
-   (uschar *)"\\p{Xsp}",   /* \s */
+   CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
-   (uschar *)"\\P{Xwd}",   /* \W */
+   CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
-   (uschar *)"\\p{Xwd}"    /* \w */
+ static const pcre_uchar string_PXsp[] = {
+   CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
+   CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
+ static const pcre_uchar string_pXsp[] = {
+   CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
+   CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
+ static const pcre_uchar string_PXwd[] = {
+   CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
+   CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
+ static const pcre_uchar string_pXwd[] = {
+   CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
+   CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
+ static const pcre_uchar *substitutes[] = {
+   string_PNd,           /* \D */
+   string_pNd,           /* \d */
+   string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */
+   string_pXsp,          /* \s */
+   string_PXwd,          /* \W */
+   string_pXwd           /* \w */
  };
- static const uschar *posix_substitutes[] = {
+ static const pcre_uchar string_pL[] =   {
-   (uschar *)"\\p{L}",     /* alpha */
+   CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
-   (uschar *)"\\p{Ll}",    /* lower */
+   CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
-   (uschar *)"\\p{Lu}",    /* upper */
+ static const pcre_uchar string_pLl[] =  {
-   (uschar *)"\\p{Xan}",   /* alnum */
+   CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
-   NULL,                   /* ascii */
+   CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
-   (uschar *)"\\h",        /* blank */
+ static const pcre_uchar string_pLu[] =  {
-   NULL,                   /* cntrl */
+   CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
-   (uschar *)"\\p{Nd}",    /* digit */
+   CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
-   NULL,                   /* graph */
+ static const pcre_uchar string_pXan[] = {
-   NULL,                   /* print */
+   CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
-   NULL,                   /* punct */
+   CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
-   (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */
+ static const pcre_uchar string_h[] =    {
-   (uschar *)"\\p{Xwd}",   /* word */
+   CHAR_BACKSLASH, CHAR_h, '\0' };
-   NULL,                   /* xdigit */
+ static const pcre_uchar string_pXps[] = {
+   CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
+   CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
+ static const pcre_uchar string_PL[] =   {
+   CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
+   CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
+ static const pcre_uchar string_PLl[] =  {
+   CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
+   CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
+ static const pcre_uchar string_PLu[] =  {
+   CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
+   CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
+ static const pcre_uchar string_PXan[] = {
+   CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
+   CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
+ static const pcre_uchar string_H[] =    {
+   CHAR_BACKSLASH, CHAR_H, '\0' };
+ static const pcre_uchar string_PXps[] = {
+   CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
+   CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
+ static const pcre_uchar *posix_substitutes[] = {
+   string_pL,            /* alpha */
+   string_pLl,           /* lower */
+   string_pLu,           /* upper */
+   string_pXan,          /* alnum */
+   NULL,                 /* ascii */
+   string_h,             /* blank */
+   NULL,                 /* cntrl */
+   string_pNd,           /* digit */
+   NULL,                 /* graph */
+   NULL,                 /* print */
+   NULL,                 /* punct */
+   string_pXps,          /* space */    /* NOTE: Xps is POSIX space */
+   string_pXwd,          /* word */
+   NULL,                 /* xdigit */
    /* Negated cases */
-   (uschar *)"\\P{L}",     /* ^alpha */
+   string_PL,            /* ^alpha */
-   (uschar *)"\\P{Ll}",    /* ^lower */
+   string_PLl,           /* ^lower */
-   (uschar *)"\\P{Lu}",    /* ^upper */
+   string_PLu,           /* ^upper */
-   (uschar *)"\\P{Xan}",   /* ^alnum */
+   string_PXan,          /* ^alnum */
-   NULL,                   /* ^ascii */
+   NULL,                 /* ^ascii */
-   (uschar *)"\\H",        /* ^blank */
+   string_H,             /* ^blank */
-   NULL,                   /* ^cntrl */
+   NULL,                 /* ^cntrl */
-   (uschar *)"\\P{Nd}",    /* ^digit */
+   string_PNd,           /* ^digit */
-   NULL,                   /* ^graph */
+   NULL,                 /* ^graph */
-   NULL,                   /* ^print */
+   NULL,                 /* ^print */
-   NULL,                   /* ^punct */
+   NULL,                 /* ^punct */
-   (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */
+   string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */
-   (uschar *)"\\P{Xwd}",   /* ^word */
+   string_PXwd,          /* ^word */
-   NULL                    /* ^xdigit */
+   NULL                  /* ^xdigit */
  };
- #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))
+ #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
  #endif
  #define STRING(a)  # a
-Line 372 
 static const char error_texts[] =
+Line 453 
 static const char error_texts[] =
    /* 30 */
    "unknown POSIX class name\0"
    "POSIX collating elements are not supported\0"
-   "this version of PCRE is not compiled with PCRE_UTF8 support\0"
+   "this version of PCRE is compiled without UTF support\0"
    "spare error\0"  /** DEAD **/
    "character value in \\x{...} sequence is too large\0"
    /* 35 */
-Line 395 
 static const char error_texts[] =
+Line 476 
 static const char error_texts[] =
    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
    /* 50 */
    "repeated subpattern is too long\0"    /** DEAD **/
-   "octal value is greater than \\377 (not in UTF-8 mode)\0"
+   "octal value is greater than \\377 in 8-bit non-UTF-8 mode\0"
    "internal error: overran compiling workspace\0"
    "internal error: previously-checked referenced subpattern not found\0"
    "DEFINE group contains more than one branch\0"
-Line 414 
 static const char error_texts[] =
+Line 495 
 static const char error_texts[] =
    /* 65 */
    "different names for subpatterns of the same number are not allowed\0"
    "(*MARK) must have an argument\0"
-   "this version of PCRE is not compiled with PCRE_UCP support\0"
+   "this version of PCRE is not compiled with Unicode property support\0"
    "\\c must be followed by an ASCII character\0"
    "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
    /* 70 */
    "internal error: unknown opcode in find_fixedlength()\0"
    "\\N is not supported in a class\0"
    "too many forward references\0"
+   "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
+   "invalid UTF-16 string\0"
+   /* 75 */
+   "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
+   "character value in \\u.... sequence is too large\0"
+   "invalid UTF-32 string\0"
    ;
  /* Table to identify digits and hex digits. This is used when compiling
-Line 439 
 For convenience, we use the same bit def
+Line 526 
 For convenience, we use the same bit def
  Then we can use ctype_digit and ctype_xdigit in the code. */
+ /* Using a simple comparison for decimal numbers rather than a memory read
+ is much faster, and the resulting code is simpler (the compiler turns it
+ into a subtraction and unsigned comparison). */
+ #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
  #ifndef EBCDIC
  /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
  UTF-8 mode. */
- static const unsigned char digitab[] =
+ static const pcre_uint8 digitab[] =
    {
 x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
 x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */
-Line 483 
 static const unsigned char digitab[] =
+Line 576 
 static const unsigned char digitab[] =
  /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
- static const unsigned char digitab[] =
+ static const pcre_uint8 digitab[] =
    {
 x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
 x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */
-Line 518 
 static const unsigned char digitab[] =
+Line 611 
 static const unsigned char digitab[] =
 x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
 x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
- static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */
+ static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
 x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
 x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
 x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */
-Line 554 
 static const unsigned char ebcdic_charta
+Line 647 
 static const unsigned char ebcdic_charta
  #endif
- /* Definition to allow mutual recursion */
- static BOOL
-   compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,
-     int *, int *, branch_chain *, compile_data *, int *);
  /*************************************************
  *            Find an error text                  *
-Line 581 
 find_error_text(int n)
+Line 667 
 find_error_text(int n)
  const char *s = error_texts;
  for (; n > 0; n--)
    {
-   while (*s++ != 0) {};
+   while (*s++ != CHAR_NULL) {};
-   if (*s == 0) return "Error text not found (please report)";
+   if (*s == CHAR_NULL) return "Error text not found (please report)";
    }
  return s;
  }
-Line 592 
 return s;
+Line 678 
 return s;
  *           Expand the workspace                 *
  *************************************************/
  /* This function is called during the second compiling phase, if the number of
  forward references fills the existing workspace, which is originally a block on
  the stack. A larger block is obtained from malloc() unless the ultimate limit
  has been reached or the increase will be rather small.
  Argument: pointer to the compile data block
-Line 604 
 Returns:  0 if all went well, else an er
+Line 690 
 Returns:  0 if all went well, else an er
  static int
  expand_workspace(compile_data *cd)
  {
- uschar *newspace;
+ pcre_uchar *newspace;
  int newsize = cd->workspace_size * 2;
  if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
-Line 612 
 if (cd->workspace_size >= COMPILE_WORK_S
+Line 698 
 if (cd->workspace_size >= COMPILE_WORK_S
      newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
   return ERR72;
- newspace = (pcre_malloc)(newsize);
+ newspace = (PUBL(malloc))(IN_UCHARS(newsize));
  if (newspace == NULL) return ERR21;
+ memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
- memcpy(newspace, cd->start_workspace, cd->workspace_size);
+ cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
- cd->hwm = (uschar *)newspace + (cd->hwm - cd->start_workspace);
+ if (cd->workspace_size > COMPILE_WORK_SIZE)
- if (cd->workspace_size > COMPILE_WORK_SIZE)
+   (PUBL(free))((void *)cd->start_workspace);
-   (pcre_free)((void *)cd->start_workspace);
  cd->start_workspace = newspace;
  cd->workspace_size = newsize;
  return 0;
-Line 642 
 Returns:    TRUE or FALSE
+Line 727 
 Returns:    TRUE or FALSE
  */
  static BOOL
- is_counted_repeat(const uschar *p)
+ is_counted_repeat(const pcre_uchar *p)
  {
- if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
+ if (!IS_DIGIT(*p)) return FALSE;
- while ((digitab[*p] & ctype_digit) != 0) p++;
+ p++;
+ while (IS_DIGIT(*p)) p++;
  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
  if (*p++ != CHAR_COMMA) return FALSE;
  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
- if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
+ if (!IS_DIGIT(*p)) return FALSE;
- while ((digitab[*p] & ctype_digit) != 0) p++;
+ p++;
+ while (IS_DIGIT(*p)) p++;
  return (*p == CHAR_RIGHT_CURLY_BRACKET);
  }
-Line 664 
 return (*p == CHAR_RIGHT_CURLY_BRACKET);
+Line 751 
 return (*p == CHAR_RIGHT_CURLY_BRACKET);
  *************************************************/
  /* This function is called when a \ has been encountered. It either returns a
- positive value for a simple escape such as \n, or a negative value which
+ positive value for a simple escape such as \n, or 0 for a data character
- encodes one of the more complicated things such as \d. A backreference to group
+ which will be placed in chptr. A backreference to group n is returned as
- n is returned as -(ESC_REF + n); ESC_REF is the highest ESC_xxx macro. When
+ negative n. When UTF-8 is enabled, a positive value greater than 255 may
- UTF-8 is enabled, a positive value greater than 255 may be returned. On entry,
+ be returned in chptr.
- ptr is pointing at the \. On exit, it is on the final character of the escape
+ On entry,ptr is pointing at the \. On exit, it is on the final character of the
- sequence.
+ escape sequence.
  Arguments:
    ptrptr         points to the pattern position pointer
+   chptr          points to the data character
    errorcodeptr   points to the errorcode variable
    bracount       number of previous extracting brackets
    options        the options bits
    isclass        TRUE if inside a character class
- Returns:         zero or positive => a data character
+ Returns:         zero => a data character
-                  negative => a special escape sequence
+                  positive => a special escape sequence
+                  negative => a back reference
                   on error, errorcodeptr is set
  */
  static int
- check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,
+ check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
-   int options, BOOL isclass)
+   int bracount, int options, BOOL isclass)
  {
- BOOL utf8 = (options & PCRE_UTF8) != 0;
+ /* PCRE_UTF16 has the same value as PCRE_UTF8. */
- const uschar *ptr = *ptrptr + 1;
+ BOOL utf = (options & PCRE_UTF8) != 0;
- int c, i;
+ const pcre_uchar *ptr = *ptrptr + 1;
+ pcre_uint32 c;
+ int escape = 0;
+ int i;
  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
  ptr--;                            /* Set pointer back to the last byte */
  /* If backslash is at the end of the pattern, it's an error. */
- if (c == 0) *errorcodeptr = ERR1;
+ if (c == CHAR_NULL) *errorcodeptr = ERR1;
  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
  in a table. A non-zero result is something that can be returned immediately.
  Otherwise further processing may be required. */
  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
- else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
+ /* Not alphanumeric */
- else if ((i = escapes[c - CHAR_0]) != 0) c = i;
+ else if (c < CHAR_0 || c > CHAR_z) {}
+ else if ((i = escapes[c - CHAR_0]) != 0)
+   { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
  #else           /* EBCDIC coding */
- else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
+ /* Not alphanumeric */
- else if ((i = escapes[c - 0x48]) != 0)  c = i;
+ else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
+ else if ((i = escapes[c - 0x48]) != 0)  { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
  #endif
  /* Escapes that need further processing, or are illegal. */
  else
    {
-   const uschar *oldptr;
+   const pcre_uchar *oldptr;
-   BOOL braced, negated;
+   BOOL braced, negated, overflow;
+   int s;
    switch (c)
      {
-Line 733 
 else
+Line 829 
 else
        {
        /* In JavaScript, \u must be followed by four hexadecimal numbers.
        Otherwise it is a lowercase u letter. */
-       if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0
+       if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
-            && (digitab[ptr[3]] & ctype_xdigit) != 0 && (digitab[ptr[4]] & ctype_xdigit) != 0)
+         && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
+         && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
+         && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
          {
          c = 0;
          for (i = 0; i < 4; ++i)
            {
-           register int cc = *(++ptr);
+           register pcre_uint32 cc = *(++ptr);
  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
-Line 748 
 else
+Line 846 
 else
            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
  #endif
            }
+ #if defined COMPILE_PCRE8
+         if (c > (utf ? 0x10ffffU : 0xffU))
+ #elif defined COMPILE_PCRE16
+         if (c > (utf ? 0x10ffffU : 0xffffU))
+ #elif defined COMPILE_PCRE32
+         if (utf && c > 0x10ffffU)
+ #endif
+           {
+           *errorcodeptr = ERR76;
+           }
+         else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
          }
        }
      else
-Line 774 
 else
+Line 884 
 else
      (3) For Oniguruma compatibility we also support \g followed by a name or a
      number either in angle brackets or in single quotes. However, these are
      (possibly recursive) subroutine calls, _not_ backreferences. Just return
-     the -ESC_g code (cf \k). */
+     the ESC_g code (cf \k). */
      case CHAR_g:
      if (isclass) break;
      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
        {
-       c = -ESC_g;
+       escape = ESC_g;
        break;
        }
-Line 788 
 else
+Line 898 
 else
      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
        {
-       const uschar *p;
+       const pcre_uchar *p;
-       for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
+       for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
-         if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
+         if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
-       if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
+       if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
          {
-         c = -ESC_k;
+         escape = ESC_k;
          break;
          }
        braced = TRUE;
-Line 808 
 else
+Line 918 
 else
        }
      else negated = FALSE;
-     c = 0;
+     /* The integer range is limited by the machine's int representation. */
-     while ((digitab[ptr[1]] & ctype_digit) != 0)
+     s = 0;
-       c = c * 10 + *(++ptr) - CHAR_0;
+     overflow = FALSE;
+     while (IS_DIGIT(ptr[1]))
-     if (c < 0)   /* Integer overflow */
        {
+       if (s > INT_MAX / 10 - 1) /* Integer overflow */
+         {
+         overflow = TRUE;
+         break;
+         }
+       s = s * 10 + (int)(*(++ptr) - CHAR_0);
+       }
+     if (overflow) /* Integer overflow */
+       {
+       while (IS_DIGIT(ptr[1]))
+         ptr++;
        *errorcodeptr = ERR61;
        break;
        }
-Line 824 
 else
+Line 944 
 else
        break;
        }
-     if (c == 0)
+     if (s == 0)
        {
        *errorcodeptr = ERR58;
        break;
-Line 832 
 else
+Line 952 
 else
      if (negated)
        {
-       if (c > bracount)
+       if (s > bracount)
          {
          *errorcodeptr = ERR15;
          break;
          }
-       c = bracount - (c - 1);
+       s = bracount - (s - 1);
        }
-     c = -(ESC_REF + c);
+     escape = -s;
      break;
      /* The handling of escape sequences consisting of a string of digits
-Line 861 
 else
+Line 981 
 else
      if (!isclass)
        {
        oldptr = ptr;
-       c -= CHAR_0;
+       /* The integer range is limited by the machine's int representation. */
-       while ((digitab[ptr[1]] & ctype_digit) != 0)
+       s = (int)(c -CHAR_0);
-         c = c * 10 + *(++ptr) - CHAR_0;
+       overflow = FALSE;
-       if (c < 0)    /* Integer overflow */
+       while (IS_DIGIT(ptr[1]))
          {
+         if (s > INT_MAX / 10 - 1) /* Integer overflow */
+           {
+           overflow = TRUE;
+           break;
+           }
+         s = s * 10 + (int)(*(++ptr) - CHAR_0);
+         }
+       if (overflow) /* Integer overflow */
+         {
+         while (IS_DIGIT(ptr[1]))
+           ptr++;
          *errorcodeptr = ERR61;
          break;
          }
-       if (c < 10 || c <= bracount)
+       if (s < 10 || s <= bracount)
          {
-         c = -(ESC_REF + c);
+         escape = -s;
          break;
          }
        ptr = oldptr;      /* Put the pointer back and fall through */
-Line 891 
 else
+Line 1022 
 else
      /* \0 always starts an octal number, but we may drop through to here with a
      larger first octal digit. The original code used just to take the least
      significant 8 bits of octal numbers (I think this is what early Perls used
-     to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
+     to do). Nowadays we allow for larger numbers in UTF-8 mode and 16-bit mode,
-     than 3 octal digits. */
+     but no more than 3 octal digits. */
      case CHAR_0:
      c -= CHAR_0;
      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
          c = c * 8 + *(++ptr) - CHAR_0;
-     if (!utf8 && c > 255) *errorcodeptr = ERR51;
+ #ifdef COMPILE_PCRE8
+     if (!utf && c > 0xff) *errorcodeptr = ERR51;
+ #endif
      break;
      /* \x is complicated. \x{ddd} is a character number which can be greater
-     than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is
+     than 0xff in utf or non-8bit mode, but only if the ddd are hex digits.
-     treated as a data character. */
+     If not, { is treated as a data character. */
      case CHAR_x:
      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
        {
        /* In JavaScript, \x must be followed by two hexadecimal numbers.
        Otherwise it is a lowercase x letter. */
-       if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0)
+       if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
+         && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
          {
          c = 0;
          for (i = 0; i < 2; ++i)
            {
-           register int cc = *(++ptr);
+           register pcre_uint32 cc = *(++ptr);
  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
-Line 930 
 else
+Line 1064 
 else
      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
        {
-       const uschar *pt = ptr + 2;
+       const pcre_uchar *pt = ptr + 2;
-       int count = 0;
        c = 0;
-       while ((digitab[*pt] & ctype_xdigit) != 0)
+       overflow = FALSE;
+       while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)
          {
-         register int cc = *pt++;
+         register pcre_uint32 cc = *pt++;
          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
-         count++;
+ #ifdef COMPILE_PCRE32
+         if (c >= 0x10000000l) { overflow = TRUE; break; }
+ #endif
  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
-Line 947 
 else
+Line 1084 
 else
          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
  #endif
+ #if defined COMPILE_PCRE8
+         if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
+ #elif defined COMPILE_PCRE16
+         if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
+ #elif defined COMPILE_PCRE32
+         if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
+ #endif
+         }
+       if (overflow)
+         {
+         while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;
+         *errorcodeptr = ERR34;
          }
        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
          {
-         if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
+         if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
          ptr = pt;
          break;
          }
-Line 963 
 else
+Line 1114 
 else
      /* Read just a single-byte hex-defined char */
      c = 0;
-     while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
+     while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
        {
-       int cc;                                  /* Some compilers don't like */
+       pcre_uint32 cc;                          /* Some compilers don't like */
        cc = *(++ptr);                           /* ++ in initializers */
  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
-Line 984 
 else
+Line 1135 
 else
      case CHAR_c:
      c = *(++ptr);
-     if (c == 0)
+     if (c == CHAR_NULL)
        {
        *errorcodeptr = ERR2;
        break;
-Line 1024 
 else
+Line 1175 
 else
  newline". PCRE does not support \N{name}. However, it does support
  quantification such as \N{2,3}. */
- if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
+ if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
       !is_counted_repeat(ptr+2))
    *errorcodeptr = ERR37;
  /* If PCRE_UCP is set, we change the values for \d etc. */
- if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
+ if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
-   c -= (ESC_DU - ESC_D);
+   escape += (ESC_DU - ESC_D);
  /* Set the pointer to the final character before returning. */
  *ptrptr = ptr;
- return c;
+ *chptr = c;
+ return escape;
  }
  #ifdef SUPPORT_UCP
  /*************************************************
  *               Handle \P and \p                 *
-Line 1054 
 escape sequence.
+Line 1204 
 escape sequence.
  Argument:
    ptrptr         points to the pattern position pointer
    negptr         points to a boolean that is set TRUE for negation else FALSE
-   dptr           points to an int that is set to the detailed property value
+   ptypeptr       points to an unsigned int that is set to the type value
+   pdataptr       points to an unsigned int that is set to the detailed property value
    errorcodeptr   points to the error code variable
- Returns:         type value from ucp_type_table, or -1 for an invalid type
+ Returns:         TRUE if the type value was found, or FALSE for an invalid type
  */
- static int
+ static BOOL
- get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
+ get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
+   unsigned int *pdataptr, int *errorcodeptr)
  {
- int c, i, bot, top;
+ pcre_uchar c;
- const uschar *ptr = *ptrptr;
+ int i, bot, top;
- char name[32];
+ const pcre_uchar *ptr = *ptrptr;
+ pcre_uchar name[32];
  c = *(++ptr);
- if (c == 0) goto ERROR_RETURN;
+ if (c == CHAR_NULL) goto ERROR_RETURN;
  *negptr = FALSE;
-Line 1082 
 if (c == CHAR_LEFT_CURLY_BRACKET)
+Line 1235 
 if (c == CHAR_LEFT_CURLY_BRACKET)
      *negptr = TRUE;
      ptr++;
      }
-   for (i = 0; i < (int)sizeof(name) - 1; i++)
+   for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
      {
      c = *(++ptr);
-     if (c == 0) goto ERROR_RETURN;
+     if (c == CHAR_NULL) goto ERROR_RETURN;
      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
      name[i] = c;
      }
-Line 1106 
 else
+Line 1259 
 else
  /* Search for a recognized property name using binary chop */
  bot = 0;
- top = _pcre_utt_size;
+ top = PRIV(utt_size);
  while (bot < top)
    {
+   int r;
    i = (bot + top) >> 1;
-   c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
+   r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
-   if (c == 0)
+   if (r == 0)
      {
-     *dptr = _pcre_utt[i].value;
+     *ptypeptr = PRIV(utt)[i].type;
-     return _pcre_utt[i].type;
+     *pdataptr = PRIV(utt)[i].value;
+     return TRUE;
      }
-   if (c > 0) bot = i + 1; else top = i;
+   if (r > 0) bot = i + 1; else top = i;
    }
  *errorcodeptr = ERR47;
  *ptrptr = ptr;
- return -1;
+ return FALSE;
  ERROR_RETURN:
  *errorcodeptr = ERR46;
  *ptrptr = ptr;
- return -1;
+ return FALSE;
  }
  #endif
-Line 1153 
 Returns:         pointer to '}' on succe
+Line 1308 
 Returns:         pointer to '}' on succe
                   current ptr on error, with errorcodeptr set non-zero
  */
- static const uschar *
+ static const pcre_uchar *
- read_repeat_counts(const uschar *p, int *minp, int *maxp, int *errorcodeptr)
+ read_repeat_counts(const pcre_uchar *p, int *minp, int *maxp, int *errorcodeptr)
  {
  int min = 0;
  int max = -1;
-Line 1162 
 int max = -1;
+Line 1317 
 int max = -1;
  /* Read the minimum value and do a paranoid check: a negative value indicates
  an integer overflow. */
- while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
+ while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);
  if (min < 0 || min > 65535)
    {
    *errorcodeptr = ERR5;
-Line 1177 
 if (*p == CHAR_RIGHT_CURLY_BRACKET) max
+Line 1332 
 if (*p == CHAR_RIGHT_CURLY_BRACKET) max
    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
      {
      max = 0;
-     while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
+     while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);
      if (max < 0 || max > 65535)
        {
        *errorcodeptr = ERR5;
-Line 1232 
 Arguments:
+Line 1387 
 Arguments:
    name         name to seek, or NULL if seeking a numbered subpattern
    lorn         name length, or subpattern number if name is NULL
    xmode        TRUE if we are in /x mode
-   utf8         TRUE if we are in UTF-8 mode
+   utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode
    count        pointer to the current capturing subpattern number (updated)
  Returns:       the number of the named subpattern, or -1 if not found
  */
  static int
- find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
+ find_parens_sub(pcre_uchar **ptrptr, compile_data *cd, const pcre_uchar *name, int lorn,
-   BOOL xmode, BOOL utf8, int *count)
+   BOOL xmode, BOOL utf, int *count)
  {
- uschar *ptr = *ptrptr;
+ pcre_uchar *ptr = *ptrptr;
  int start_count = *count;
  int hwm_count = start_count;
  BOOL dup_parens = FALSE;
-Line 1254 
 if (ptr[0] == CHAR_LEFT_PARENTHESIS)
+Line 1409 
 if (ptr[0] == CHAR_LEFT_PARENTHESIS)
    {
    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
-   if (ptr[1] == CHAR_ASTERISK) ptr += 2;
+   if (ptr[1] == CHAR_ASTERISK)
+     {
+     ptr += 2;
+     while (ptr < cd->end_pattern && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
+     }
    /* Handle a normal, unnamed capturing parenthesis. */
-Line 1278 
 if (ptr[0] == CHAR_LEFT_PARENTHESIS)
+Line 1437 
 if (ptr[0] == CHAR_LEFT_PARENTHESIS)
    else if (ptr[2] == CHAR_NUMBER_SIGN)
      {
-     for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
+     for (ptr += 3; *ptr != CHAR_NULL; ptr++)
+       if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
      goto FAIL_EXIT;
      }
-Line 1291 
 if (ptr[0] == CHAR_LEFT_PARENTHESIS)
+Line 1451 
 if (ptr[0] == CHAR_LEFT_PARENTHESIS)
      ptr += 2;
      if (ptr[1] != CHAR_QUESTION_MARK)
        {
-       while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
+       while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
-       if (*ptr != 0) ptr++;
+       if (*ptr != CHAR_NULL) ptr++;
        }
      }
-Line 1308 
 if (ptr[0] == CHAR_LEFT_PARENTHESIS)
+Line 1468 
 if (ptr[0] == CHAR_LEFT_PARENTHESIS)
      if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
          ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
        {
-       int term;
+       pcre_uchar term;
-       const uschar *thisname;
+       const pcre_uchar *thisname;
        *count += 1;
        if (name == NULL && *count == lorn) return *count;
        term = *ptr++;
        if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
        thisname = ptr;
        while (*ptr != term) ptr++;
-       if (name != NULL && lorn == ptr - thisname &&
+       if (name != NULL && lorn == (int)(ptr - thisname) &&
-           strncmp((const char *)name, (const char *)thisname, lorn) == 0)
+           STRNCMP_UC_UC(name, thisname, (unsigned int)lorn) == 0)
          return *count;
        term++;
        }
-Line 1335 
 for (; ptr < cd->end_pattern; ptr++)
+Line 1495 
 for (; ptr < cd->end_pattern; ptr++)
    if (*ptr == CHAR_BACKSLASH)
      {
-     if (*(++ptr) == 0) goto FAIL_EXIT;
+     if (*(++ptr) == CHAR_NULL) goto FAIL_EXIT;
      if (*ptr == CHAR_Q) for (;;)
        {
-       while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
+       while (*(++ptr) != CHAR_NULL && *ptr != CHAR_BACKSLASH) {};
-       if (*ptr == 0) goto FAIL_EXIT;
+       if (*ptr == CHAR_NULL) goto FAIL_EXIT;
        if (*(++ptr) == CHAR_E) break;
        }
      continue;
-Line 1360 
 for (; ptr < cd->end_pattern; ptr++)
+Line 1520 
 for (; ptr < cd->end_pattern; ptr++)
          {
          if (ptr[2] == CHAR_E)
            ptr+= 2;
-         else if (strncmp((const char *)ptr+2,
+         else if (STRNCMP_UC_C8(ptr + 2,
                   STR_Q STR_BACKSLASH STR_E, 3) == 0)
            ptr += 4;
          else
-Line 1383 
 for (; ptr < cd->end_pattern; ptr++)
+Line 1543 
 for (; ptr < cd->end_pattern; ptr++)
      while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
        {
-       if (*ptr == 0) return -1;
+       if (*ptr == CHAR_NULL) return -1;
        if (*ptr == CHAR_BACKSLASH)
          {
-         if (*(++ptr) == 0) goto FAIL_EXIT;
+         if (*(++ptr) == CHAR_NULL) goto FAIL_EXIT;
          if (*ptr == CHAR_Q) for (;;)
            {
-           while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
+           while (*(++ptr) != CHAR_NULL && *ptr != CHAR_BACKSLASH) {};
-           if (*ptr == 0) goto FAIL_EXIT;
+           if (*ptr == CHAR_NULL) goto FAIL_EXIT;
            if (*(++ptr) == CHAR_E) break;
            }
          continue;
-Line 1404 
 for (; ptr < cd->end_pattern; ptr++)
+Line 1564 
 for (; ptr < cd->end_pattern; ptr++)
    if (xmode && *ptr == CHAR_NUMBER_SIGN)
      {
      ptr++;
-     while (*ptr != 0)
+     while (*ptr != CHAR_NULL)
        {
        if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
        ptr++;
- #ifdef SUPPORT_UTF8
+ #ifdef SUPPORT_UTF
-       if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
+       if (utf) FORWARDCHAR(ptr);
  #endif
        }
-     if (*ptr == 0) goto FAIL_EXIT;
+     if (*ptr == CHAR_NULL) goto FAIL_EXIT;
      continue;
      }
-Line 1420 
 for (; ptr < cd->end_pattern; ptr++)
+Line 1580 
 for (; ptr < cd->end_pattern; ptr++)
    if (*ptr == CHAR_LEFT_PARENTHESIS)
      {
-     int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
+     int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);
      if (rc > 0) return rc;
-     if (*ptr == 0) goto FAIL_EXIT;
+     if (*ptr == CHAR_NULL) goto FAIL_EXIT;
      }
    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
-Line 1466 
 Arguments:
+Line 1626 
 Arguments:
    name         name to seek, or NULL if seeking a numbered subpattern
    lorn         name length, or subpattern number if name is NULL
    xmode        TRUE if we are in /x mode
-   utf8         TRUE if we are in UTF-8 mode
+   utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode
  Returns:       the number of the found subpattern, or -1 if not found
  */
  static int
- find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
+ find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,
-   BOOL utf8)
+   BOOL utf)
  {
- uschar *ptr = (uschar *)cd->start_pattern;
+ pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;
  int count = 0;
  int rc;
-Line 1486 
 matching closing parens. That is why we
+Line 1646 
 matching closing parens. That is why we
  for (;;)
    {
-   rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
+   rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);
-   if (rc > 0 || *ptr++ == 0) break;
+   if (rc > 0 || *ptr++ == CHAR_NULL) break;
    }
  return rc;
-Line 1513 
 Arguments:
+Line 1673 
 Arguments:
  Returns:       pointer to the first significant opcode
  */
- static const uschar*
+ static const pcre_uchar*
- first_significant_code(const uschar *code, BOOL skipassert)
+ first_significant_code(const pcre_uchar *code, BOOL skipassert)
  {
  for (;;)
    {
-Line 1525 
 for (;;)
+Line 1685 
 for (;;)
      case OP_ASSERTBACK_NOT:
      if (!skipassert) return code;
      do code += GET(code, 1); while (*code == OP_ALT);
-     code += _pcre_OP_lengths[*code];
+     code += PRIV(OP_lengths)[*code];
      break;
      case OP_WORD_BOUNDARY:
-Line 1539 
 for (;;)
+Line 1699 
 for (;;)
      case OP_RREF:
      case OP_NRREF:
      case OP_DEF:
-     code += _pcre_OP_lengths[*code];
+     code += PRIV(OP_lengths)[*code];
      break;
      default:
-Line 1569 
 and doing the check at the end; a flag s
+Line 1729 
 and doing the check at the end; a flag s
  Arguments:
    code     points to the start of the pattern (the bracket)
-   utf8     TRUE in UTF-8 mode
+   utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
    atend    TRUE if called when the pattern is complete
    cd       the "compile data" structure
-Line 1581 
 Returns:   the fixed length,
+Line 1741 
 Returns:   the fixed length,
  */
  static int
- find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
+ find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
  {
  int length = -1;
  register int branchlength = 0;
- register uschar *cc = code + 1 + LINK_SIZE;
+ register pcre_uchar *cc = code + 1 + LINK_SIZE;
  /* Scan along the opcodes for this branch. If we get to the end of the
  branch, check the length against that of the other branches. */
-Line 1594 
 branch, check the length against that of
+Line 1754 
 branch, check the length against that of
  for (;;)
    {
    int d;
-   uschar *ce, *cs;
+   pcre_uchar *ce, *cs;
-   register int op = *cc;
+   register pcre_uchar op = *cc;
    switch (op)
      {
      /* We only need to continue for OP_CBRA (normal capturing bracket) and
-Line 1608 
 for (;;)
+Line 1769 
 for (;;)
      case OP_ONCE:
      case OP_ONCE_NC:
      case OP_COND:
-     d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
+     d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
      if (d < 0) return d;
      branchlength += d;
      do cc += GET(cc, 1); while (*cc == OP_ALT);
-Line 1639 
 for (;;)
+Line 1800 
 for (;;)
      case OP_RECURSE:
      if (!atend) return -3;
-     cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
+     cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
-     do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
+     do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
-     if (cc > cs && cc < ce) return -1;                /* Recursion */
+     if (cc > cs && cc < ce) return -1;                    /* Recursion */
-     d = find_fixedlength(cs + 2, utf8, atend, cd);
+     d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
      if (d < 0) return d;
      branchlength += d;
      cc += 1 + LINK_SIZE;
-Line 1655 
 for (;;)
+Line 1816 
 for (;;)
      case OP_ASSERTBACK:
      case OP_ASSERTBACK_NOT:
      do cc += GET(cc, 1); while (*cc == OP_ALT);
-     /* Fall through */
+     cc += PRIV(OP_lengths)[*cc];
+     break;
      /* Skip over things that don't match chars */
-Line 1663 
 for (;;)
+Line 1825 
 for (;;)
      case OP_PRUNE_ARG:
      case OP_SKIP_ARG:
      case OP_THEN_ARG:
-     cc += cc[1] + _pcre_OP_lengths[*cc];
+     cc += cc[1] + PRIV(OP_lengths)[*cc];
      break;
      case OP_CALLOUT:
-Line 1690 
 for (;;)
+Line 1852 
 for (;;)
      case OP_SOM:
      case OP_THEN:
      case OP_WORD_BOUNDARY:
-     cc += _pcre_OP_lengths[*cc];
+     cc += PRIV(OP_lengths)[*cc];
      break;
      /* Handle literal characters */
-Line 1701 
 for (;;)
+Line 1863 
 for (;;)
      case OP_NOTI:
      branchlength++;
      cc += 2;
- #ifdef SUPPORT_UTF8
+ #ifdef SUPPORT_UTF
-     if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
+     if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
  #endif
      break;
-Line 1713 
 for (;;)
+Line 1875 
 for (;;)
      case OP_EXACTI:
      case OP_NOTEXACT:
      case OP_NOTEXACTI:
-     branchlength += GET2(cc,1);
+     branchlength += (int)GET2(cc,1);
-     cc += 4;
+     cc += 2 + IMM2_SIZE;
- #ifdef SUPPORT_UTF8
+ #ifdef SUPPORT_UTF
-     if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
+     if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
  #endif
      break;
      case OP_TYPEEXACT:
      branchlength += GET2(cc,1);
-     if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
+     if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
-     cc += 4;
+       cc += 2;
+     cc += 1 + IMM2_SIZE + 1;
      break;
      /* Handle single-char matchers */
-Line 1749 
 for (;;)
+Line 1912 
 for (;;)
      cc++;
      break;
      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
      otherwise \C is coded as OP_ALLANY. */
      case OP_ANYBYTE:
-Line 1757 
 for (;;)
+Line 1920 
 for (;;)
      /* Check a class for variable quantification */
- #ifdef SUPPORT_UTF8
-     case OP_XCLASS:
-     cc += GET(cc, 1) - 33;
-     /* Fall through */
- #endif
      case OP_CLASS:
      case OP_NCLASS:
-     cc += 33;
+ #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
+     case OP_XCLASS:
+     /* The original code caused an unsigned overflow in 64 bit systems,
+     so now we use a conditional statement. */
+     if (op == OP_XCLASS)
+       cc += GET(cc, 1);
+     else
+       cc += PRIV(OP_lengths)[OP_CLASS];
+ #else
+     cc += PRIV(OP_lengths)[OP_CLASS];
+ #endif
      switch (*cc)
        {
-Line 1779 
 for (;;)
+Line 1946 
 for (;;)
        case OP_CRRANGE:
        case OP_CRMINRANGE:
-       if (GET2(cc,1) != GET2(cc,3)) return -1;
+       if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
-       branchlength += GET2(cc,1);
+       branchlength += (int)GET2(cc,1);
-       cc += 5;
+       cc += 1 + 2 * IMM2_SIZE;
        break;
        default:
-Line 1896 
 length.
+Line 2063 
 length.
  Arguments:
    code        points to start of expression
-   utf8        TRUE in UTF-8 mode
+   utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
    number      the required bracket number or negative to find a lookbehind
  Returns:      pointer to the opcode for the bracket, or NULL if not found
  */
- const uschar *
+ const pcre_uchar *
- _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
+ PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
  {
  for (;;)
    {
-   register int c = *code;
+   register pcre_uchar c = *code;
    if (c == OP_END) return NULL;
-Line 1921 
 for (;;)
+Line 2088 
 for (;;)
    else if (c == OP_REVERSE)
      {
-     if (number < 0) return (uschar *)code;
+     if (number < 0) return (pcre_uchar *)code;
-     code += _pcre_OP_lengths[c];
+     code += PRIV(OP_lengths)[c];
      }
    /* Handle capturing bracket */
-Line 1930 
 for (;;)
+Line 2097 
 for (;;)
    else if (c == OP_CBRA || c == OP_SCBRA ||
             c == OP_CBRAPOS || c == OP_SCBRAPOS)
      {
-     int n = GET2(code, 1+LINK_SIZE);
+     int n = (int)GET2(code, 1+LINK_SIZE);
-     if (n == number) return (uschar *)code;
+     if (n == number) return (pcre_uchar *)code;
-     code += _pcre_OP_lengths[c];
+     code += PRIV(OP_lengths)[c];
      }
    /* Otherwise, we can get the item's length from the table, except that for
-Line 1960 
 for (;;)
+Line 2127 
 for (;;)
        case OP_TYPEMINUPTO:
        case OP_TYPEEXACT:
        case OP_TYPEPOSUPTO:
-       if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
+       if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
+         code += 2;
        break;
        case OP_MARK:
        case OP_PRUNE_ARG:
        case OP_SKIP_ARG:
-       code += code[1];
-       break;
        case OP_THEN_ARG:
        code += code[1];
        break;
-Line 1976 
 for (;;)
+Line 2141 
 for (;;)
      /* Add in the fixed length from the table */
-     code += _pcre_OP_lengths[c];
+     code += PRIV(OP_lengths)[c];
    /* In UTF-8 mode, opcodes that are followed by a character may be followed by
    a multi-byte character. The length in the table is a minimum, so we have to
    arrange to skip the extra bytes. */
- #ifdef SUPPORT_UTF8
+ #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
-     if (utf8) switch(c)
+     if (utf) switch(c)
        {
        case OP_CHAR:
        case OP_CHARI:
-Line 2013 
 for (;;)
+Line 2178 
 for (;;)
        case OP_MINQUERYI:
        case OP_POSQUERY:
        case OP_POSQUERYI:
-       if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
+       if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
        break;
        }
  #else
-     (void)(utf8);  /* Keep compiler happy by referencing function argument */
+     (void)(utf);  /* Keep compiler happy by referencing function argument */
  #endif
      }
    }
-Line 2034 
 instance of OP_RECURSE.
+Line 2199 
 instance of OP_RECURSE.
  Arguments:
    code        points to start of expression
-   utf8        TRUE in UTF-8 mode
+   utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
  Returns:      pointer to the opcode for OP_RECURSE, or NULL if not found
  */
- static const uschar *
+ static const pcre_uchar *
- find_recurse(const uschar *code, BOOL utf8)
+ find_recurse(const pcre_uchar *code, BOOL utf)
  {
  for (;;)
    {
-   register int c = *code;
+   register pcre_uchar c = *code;
    if (c == OP_END) return NULL;
    if (c == OP_RECURSE) return code;
-Line 2079 
 for (;;)
+Line 2244 
 for (;;)
        case OP_TYPEUPTO:
        case OP_TYPEMINUPTO:
        case OP_TYPEEXACT:
-       if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
+       if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
+         code += 2;
        break;
        case OP_MARK:
        case OP_PRUNE_ARG:
        case OP_SKIP_ARG:
-       code += code[1];
-       break;
        case OP_THEN_ARG:
        code += code[1];
        break;
-Line 2095 
 for (;;)
+Line 2258 
 for (;;)
      /* Add in the fixed length from the table */
-     code += _pcre_OP_lengths[c];
+     code += PRIV(OP_lengths)[c];
      /* In UTF-8 mode, opcodes that are followed by a character may be followed
      by a multi-byte character. The length in the table is a minimum, so we have
      to arrange to skip the extra bytes. */
- #ifdef SUPPORT_UTF8
+ #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
-     if (utf8) switch(c)
+     if (utf) switch(c)
        {
        case OP_CHAR:
        case OP_CHARI:
+       case OP_NOT:
+       case OP_NOTI:
        case OP_EXACT:
        case OP_EXACTI:
+       case OP_NOTEXACT:
+       case OP_NOTEXACTI:
        case OP_UPTO:
        case OP_UPTOI:
+       case OP_NOTUPTO:
+       case OP_NOTUPTOI:
        case OP_MINUPTO:
        case OP_MINUPTOI:
+       case OP_NOTMINUPTO:
+       case OP_NOTMINUPTOI:
        case OP_POSUPTO:
        case OP_POSUPTOI:
+       case OP_NOTPOSUPTO:
+       case OP_NOTPOSUPTOI:
        case OP_STAR:
        case OP_STARI:
+       case OP_NOTSTAR:
+       case OP_NOTSTARI:
        case OP_MINSTAR:
        case OP_MINSTARI:
+       case OP_NOTMINSTAR:
+       case OP_NOTMINSTARI:
        case OP_POSSTAR:
        case OP_POSSTARI:
+       case OP_NOTPOSSTAR:
+       case OP_NOTPOSSTARI:
        case OP_PLUS:
        case OP_PLUSI:
+       case OP_NOTPLUS:
+       case OP_NOTPLUSI:
        case OP_MINPLUS:
        case OP_MINPLUSI:
+       case OP_NOTMINPLUS:
+       case OP_NOTMINPLUSI:
        case OP_POSPLUS:
        case OP_POSPLUSI:
+       case OP_NOTPOSPLUS:
+       case OP_NOTPOSPLUSI:
        case OP_QUERY:
        case OP_QUERYI:
+       case OP_NOTQUERY:
+       case OP_NOTQUERYI:
        case OP_MINQUERY:
        case OP_MINQUERYI:
+       case OP_NOTMINQUERY:
+       case OP_NOTMINQUERYI:
        case OP_POSQUERY:
        case OP_POSQUERYI:
-       if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
+       case OP_NOTPOSQUERY:
+       case OP_NOTPOSQUERYI:
+       if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
        break;
        }
  #else
-     (void)(utf8);  /* Keep compiler happy by referencing function argument */
+     (void)(utf);  /* Keep compiler happy by referencing function argument */
  #endif
      }
    }
-Line 2159 
 bracket whose current branch will alread
+Line 2350 
 bracket whose current branch will alread
  Arguments:
    code        points to start of search
    endcode     points to where to stop
-   utf8        TRUE if in UTF8 mode
+   utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
    cd          contains pointers to tables etc.
  Returns:      TRUE if what is matched could be empty
  */
  static BOOL
- could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
+ could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
-   compile_data *cd)
+   BOOL utf, compile_data *cd)
  {
- register int c;
+ register pcre_uchar c;
- for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
+ for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
       code < endcode;
-      code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
+      code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
    {
-   const uschar *ccode;
+   const pcre_uchar *ccode;
    c = *code;
-Line 2197 
 for (code = first_significant_code(code
+Line 2388 
 for (code = first_significant_code(code
    if (c == OP_RECURSE)
      {
-     const uschar *scode;
+     const pcre_uchar *scode;
      BOOL empty_branch;
      /* Test for forward reference */
      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
-       if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
+       if ((int)GET(scode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
      /* Not a forward reference, test for completed backward reference */
-Line 2215 
 for (code = first_significant_code(code
+Line 2406 
 for (code = first_significant_code(code
      do
        {
-       if (could_be_empty_branch(scode, endcode, utf8, cd))
+       if (could_be_empty_branch(scode, endcode, utf, cd))
          {
          empty_branch = TRUE;
          break;
-Line 2233 
 for (code = first_significant_code(code
+Line 2424 
 for (code = first_significant_code(code
    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
        c == OP_BRAPOSZERO)
      {
-     code += _pcre_OP_lengths[c];
+     code += PRIV(OP_lengths)[c];
      do code += GET(code, 1); while (*code == OP_ALT);
      c = *code;
      continue;
-Line 2271 
 for (code = first_significant_code(code
+Line 2462 
 for (code = first_significant_code(code
        empty_branch = FALSE;
        do
          {
-         if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
+         if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))
            empty_branch = TRUE;
          code += GET(code, 1);
          }
-Line 2289 
 for (code = first_significant_code(code
+Line 2480 
 for (code = first_significant_code(code
      {
      /* Check for quantifiers after a class. XCLASS is used for classes that
      cannot be represented just by a bit map. This includes negated single
-     high-valued characters. The length in _pcre_OP_lengths[] is zero; the
+     high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
      actual length is stored in the compiled code, so we must update "code"
      here. */
- #ifdef SUPPORT_UTF8
+ #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
      case OP_XCLASS:
      ccode = code += GET(code, 1);
      goto CHECK_CLASS_REPEAT;
-Line 2301 
 for (code = first_significant_code(code
+Line 2492 
 for (code = first_significant_code(code
      case OP_CLASS:
      case OP_NCLASS:
-     ccode = code + 33;
+     ccode = code + PRIV(OP_lengths)[OP_CLASS];
- #ifdef SUPPORT_UTF8
+ #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
      CHECK_CLASS_REPEAT:
  #endif
-Line 2376 
 for (code = first_significant_code(code
+Line 2567 
 for (code = first_significant_code(code
      case OP_TYPEUPTO:
      case OP_TYPEMINUPTO:
      case OP_TYPEPOSUPTO:
-     if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
+     if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
+       code += 2;
      break;
      /* End of branch */
-Line 2391 
 for (code = first_significant_code(code
+Line 2583 
 for (code = first_significant_code(code
      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
      MINUPTO, and POSUPTO may be followed by a multibyte character */
- #ifdef SUPPORT_UTF8
+ #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
      case OP_STAR:
      case OP_STARI:
      case OP_MINSTAR:
-Line 2404 
 for (code = first_significant_code(code
+Line 2596 
 for (code = first_significant_code(code
      case OP_MINQUERYI:
      case OP_POSQUERY:
      case OP_POSQUERYI:
-     if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
+     if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
      break;
      case OP_UPTO:
-Line 2413 
 for (code = first_significant_code(code
+Line 2605 
 for (code = first_significant_code(code
      case OP_MINUPTOI:
      case OP_POSUPTO:
      case OP_POSUPTOI:
-     if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
+     if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
      break;
  #endif
-Line 2423 
 for (code = first_significant_code(code
+Line 2615 
 for (code = first_significant_code(code
      case OP_MARK:
      case OP_PRUNE_ARG:
      case OP_SKIP_ARG:
-     code += code[1];
-     break;
      case OP_THEN_ARG:
      code += code[1];
      break;
-Line 2457 
 Arguments:
+Line 2646 
 Arguments:
    code        points to start of the recursion
    endcode     points to where to stop (current RECURSE item)
    bcptr       points to the chain of current (unclosed) branch starts
-   utf8        TRUE if in UTF-8 mode
+   utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
    cd          pointers to tables etc
  Returns:      TRUE if what is matched could be empty
  */
  static BOOL
- could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
+ could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
-   BOOL utf8, compile_data *cd)
+   branch_chain *bcptr, BOOL utf, compile_data *cd)
  {
  while (bcptr != NULL && bcptr->current_branch >= code)
    {
-   if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
+   if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))
      return FALSE;
    bcptr = bcptr->outer;
    }
-Line 2521 
 Returns:   TRUE or FALSE
+Line 2710 
 Returns:   TRUE or FALSE
  */
  static BOOL
- check_posix_syntax(const uschar *ptr, const uschar **endptr)
+ check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
  {
- int terminator;          /* Don't combine these lines; the Solaris cc */
+ pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
- for (++ptr; *ptr != 0; ptr++)
+ for (++ptr; *ptr != CHAR_NULL; ptr++)
    {
    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
      ptr++;
-Line 2565 
 Returns:     a value representing the na
+Line 2754 
 Returns:     a value representing the na
  */
  static int
- check_posix_name(const uschar *ptr, int len)
+ check_posix_name(const pcre_uchar *ptr, int len)
  {
  const char *pn = posix_names;
  register int yield = 0;
  while (posix_name_lengths[yield] != 0)
    {
    if (len == posix_name_lengths[yield] &&
-     strncmp((const char *)ptr, pn, len) == 0) return yield;
+     STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
    pn += posix_name_lengths[yield] + 1;
    yield++;
    }
-Line 2604 
 value in the reference (which is a group
+Line 2793 
 value in the reference (which is a group
  Arguments:
    group      points to the start of the group
    adjust     the amount by which the group is to be moved
-   utf8       TRUE in UTF-8 mode
+   utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
    cd         contains pointers to tables etc.
    save_hwm   the hwm forward reference pointer at the start of the group
-Line 2612 
 Returns:     nothing
+Line 2801 
 Returns:     nothing
  */
  static void
- adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
+ adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
-   uschar *save_hwm)
+   pcre_uchar *save_hwm)
  {
- uschar *ptr = group;
+ pcre_uchar *ptr = group;
- while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
+ while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
    {
    int offset;
-   uschar *hc;
+   pcre_uchar *hc;
    /* See if this recursion is on the forward reference list. If so, adjust the
    reference. */
    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
      {
-     offset = GET(hc, 0);
+     offset = (int)GET(hc, 0);
      if (cd->start_code + offset == ptr + 1)
        {
        PUT(hc, 0, offset + adjust);
-Line 2640 
 while ((ptr = (uschar *)find_recurse(ptr
+Line 2829 
 while ((ptr = (uschar *)find_recurse(ptr
    if (hc >= cd->hwm)
      {
-     offset = GET(ptr, 1);
+     offset = (int)GET(ptr, 1);
      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
      }
-Line 2665 
 Arguments:
+Line 2854 
 Arguments:
  Returns:         new code pointer
  */
- static uschar *
+ static pcre_uchar *
- auto_callout(uschar *code, const uschar *ptr, compile_data *cd)
+ auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
  {
  *code++ = OP_CALLOUT;
  *code++ = 255;
  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
  PUT(code, LINK_SIZE, 0);                       /* Default length */
- return code + 2*LINK_SIZE;
+ return code + 2 * LINK_SIZE;
  }
-Line 2694 
 Returns:             nothing
+Line 2883 
 Returns:             nothing
  */
  static void
- complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
+ complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
  {
  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
  PUT(previous_callout, 2 + LINK_SIZE, length);
-Line 2708 
 PUT(previous_callout, 2 + LINK_SIZE, len
+Line 2897 
 PUT(previous_callout, 2 + LINK_SIZE, len
  *************************************************/
  /* This function is passed the start and end of a class range, in UTF-8 mode
- with UCP support. It searches up the characters, looking for internal ranges of
+ with UCP support. It searches up the characters, looking for ranges of
  characters in the "other" case. Each call returns the next one, updating the
- start address.
+ start address. A character with multiple other cases is returned on its own
+ with a special return value.
  Arguments:
    cptr        points to starting character value; updated
-Line 2718 
 Arguments:
+Line 2908 
 Arguments:
    ocptr       where to put start of othercase range
    odptr       where to put end of othercase range
- Yield:        TRUE when range returned; FALSE when no more
+ Yield:        -1 when no more
+when a range is returned
+               >0 the CASESET offset for char with multiple other cases
+                 in this case, ocptr contains the original
  */
- static BOOL
+ static int
- get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
+ get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
-   unsigned int *odptr)
+   pcre_uint32 *odptr)
  {
- unsigned int c, othercase, next;
+ pcre_uint32 c, othercase, next;
+ unsigned int co;
+ /* Find the first character that has an other case. If it has multiple other
+ cases, return its case offset value. */
  for (c = *cptr; c <= d; c++)
-   { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
+   {
+   if ((co = UCD_CASESET(c)) != 0)
+     {
+     *ocptr = c++;   /* Character that has the set */
+     *cptr = c;      /* Rest of input range */
+     return (int)co;
+     }
+   if ((othercase = UCD_OTHERCASE(c)) != c) break;
+   }
- if (c > d) return FALSE;
+ if (c > d) return -1;  /* Reached end of range */
  *ocptr = othercase;
  next = othercase + 1;
-Line 2741 
 for (++c; c <= d; c++)
+Line 2946 
 for (++c; c <= d; c++)
    next++;
    }
- *odptr = next - 1;
+ *odptr = next - 1;     /* End of othercase range */
- *cptr = c;
+ *cptr = c;             /* Rest of input range */
+ return 0;
- return TRUE;
  }
-Line 2766 
 Returns:       TRUE if auto-possessifyin
+Line 2970 
 Returns:       TRUE if auto-possessifyin
  */
  static BOOL
- check_char_prop(int c, int ptype, int pdata, BOOL negated)
+ check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata, BOOL negated)
  {
+ #ifdef SUPPORT_UCP
+ const pcre_uint32 *p;
+ #endif
  const ucd_record *prop = GET_UCD(c);
  switch(ptype)
    {
    case PT_LAMP:
-Line 2777 
 switch(ptype)
+Line 2986 
 switch(ptype)
            prop->chartype == ucp_Lt) == negated;
    case PT_GC:
-   return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
+   return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
    case PT_PC:
    return (pdata == prop->chartype) == negated;
-Line 2788 
 switch(ptype)
+Line 2997 
 switch(ptype)
    /* These are specials */
    case PT_ALNUM:
-   return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
+   return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
-           _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
+           PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
    case PT_SPACE:    /* Perl space */
-   return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
+   return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
            == negated;
    case PT_PXSPACE:  /* POSIX space */
-   return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
+   return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
            c == CHAR_FF || c == CHAR_CR)
            == negated;
    case PT_WORD:
-   return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
+   return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
-           _pcre_ucp_gentype[prop->chartype] == ucp_N ||
+           PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
            c == CHAR_UNDERSCORE) == negated;
+ #ifdef SUPPORT_UCP
+   case PT_CLIST:
+   p = PRIV(ucd_caseless_sets) + prop->caseset;
+   for (;;)
+     {
+     if (c < *p) return !negated;
+     if (c == *p++) return negated;
+     }
+   break;  /* Control never reaches here */
+ #endif
    }
  return FALSE;
  }
  #endif  /* SUPPORT_UCP */
-Line 2823 
 sense to automatically possessify the re
+Line 3044 
 sense to automatically possessify the re
  Arguments:
    previous      pointer to the repeated opcode
-   utf8          TRUE in UTF-8 mode
+   utf           TRUE in UTF-8 / UTF-16 / UTF-32 mode
    ptr           next character in pattern
    options       options bits
    cd            contains pointers to tables etc.
-Line 2832 
 Returns:        TRUE if possessifying is
+Line 3053 
 Returns:        TRUE if possessifying is
  */
  static BOOL
- check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
+ check_auto_possessive(const pcre_uchar *previous, BOOL utf,
-   int options, compile_data *cd)
+   const pcre_uchar *ptr, int options, compile_data *cd)
  {
- int c, next;
+ pcre_uint32 c = NOTACHAR;
- int op_code = *previous++;
+ pcre_uint32 next;
+ int escape;
+ pcre_uchar op_code = *previous++;
  /* Skip whitespace and comments in extended mode */
-Line 2844 
 if ((options & PCRE_EXTENDED) != 0)
+Line 3067 
 if ((options & PCRE_EXTENDED) != 0)
    {
    for (;;)
      {
-     while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
+     while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
      if (*ptr == CHAR_NUMBER_SIGN)
        {
        ptr++;
-       while (*ptr != 0)
+       while (*ptr != CHAR_NULL)
          {
          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
          ptr++;
- #ifdef SUPPORT_UTF8
+ #ifdef SUPPORT_UTF
-         if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
+         if (utf) FORWARDCHAR(ptr);
  #endif
          }
        }
-Line 2867 
 value is a character, a negative value i
+Line 3090 
 value is a character, a negative value i
  if (*ptr == CHAR_BACKSLASH)
    {
    int temperrorcode = 0;
-   next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
+   escape = check_escape(&ptr, &next, &temperrorcode, cd->bracount, options,
+     FALSE);
    if (temperrorcode != 0) return FALSE;
    ptr++;    /* Point after the escape sequence */
    }
+ else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)
- else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
    {
- #ifdef SUPPORT_UTF8
+   escape = 0;
-   if (utf8) { GETCHARINC(next, ptr); } else
+ #ifdef SUPPORT_UTF
+   if (utf) { GETCHARINC(next, ptr); } else
  #endif
    next = *ptr++;
    }
  else return FALSE;
  /* Skip whitespace and comments in extended mode */
-Line 2888 
 if ((options & PCRE_EXTENDED) != 0)
+Line 3111 
 if ((options & PCRE_EXTENDED) != 0)
    {
    for (;;)
      {
-     while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
+     while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
      if (*ptr == CHAR_NUMBER_SIGN)
        {
        ptr++;
-       while (*ptr != 0)
+       while (*ptr != CHAR_NULL)
          {
          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
          ptr++;
- #ifdef SUPPORT_UTF8
+ #ifdef SUPPORT_UTF
-         if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
+         if (utf) FORWARDCHAR(ptr);
  #endif
          }
        }
-Line 2908 
 if ((options & PCRE_EXTENDED) != 0)
+Line 3131 
 if ((options & PCRE_EXTENDED) != 0)
  /* If the next thing is itself optional, we have to give up. */
  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
-   strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
+   STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
      return FALSE;
- /* Now compare the next item with the previous opcode. First, handle cases when
+ /* If the previous item is a character, get its value. */
- the next item is a character. */
- if (next >= 0) switch(op_code)
+ if (op_code == OP_CHAR || op_code == OP_CHARI ||
+     op_code == OP_NOT || op_code == OP_NOTI)
    {
-   case OP_CHAR:
+ #ifdef SUPPORT_UTF
- #ifdef SUPPORT_UTF8
    GETCHARTEST(c, previous);
  #else
    c = *previous;
  #endif
-   return c != next;
+   }
-   /* For CHARI (caseless character) we must check the other case. If we have
+ /* Now compare the next item with the previous opcode. First, handle cases when
-   Unicode property support, we can use it to test the other case of
+ the next item is a character. */
-   high-valued characters. */
-   case OP_CHARI:
+ if (escape == 0)
- #ifdef SUPPORT_UTF8
+   {
-   GETCHARTEST(c, previous);
+   /* For a caseless UTF match, the next character may have more than one other
- #else
+   case, which maps to the special PT_CLIST property. Check this first. */
-   c = *previous;
+ #ifdef SUPPORT_UCP
+   if (utf && c != NOTACHAR && (options & PCRE_CASELESS) != 0)
+     {
+     unsigned int ocs = UCD_CASESET(next);
+     if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);
+     }
  #endif
-   if (c == next) return FALSE;
- #ifdef SUPPORT_UTF8
+   switch(op_code)
-   if (utf8)
      {
-     unsigned int othercase;
+     case OP_CHAR:
-     if (next < 128) othercase = cd->fcc[next]; else
+     return c != next;
+     /* For CHARI (caseless character) we must check the other case. If we have
+     Unicode property support, we can use it to test the other case of
+     high-valued characters. We know that next can have only one other case,
+     because multi-other-case characters are dealt with above. */
+     case OP_CHARI:
+     if (c == next) return FALSE;
+ #ifdef SUPPORT_UTF
+     if (utf)
+       {
+       pcre_uint32 othercase;
+       if (next < 128) othercase = cd->fcc[next]; else
  #ifdef SUPPORT_UCP
-     othercase = UCD_OTHERCASE((unsigned int)next);
+       othercase = UCD_OTHERCASE(next);
  #else
-     othercase = NOTACHAR;
+       othercase = NOTACHAR;
  #endif
-     return (unsigned int)c != othercase;
+       return c != othercase;
-     }
+       }
-   else
+     else
- #endif  /* SUPPORT_UTF8 */
+ #endif  /* SUPPORT_UTF */
-   return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
+     return (c != TABLE_GET(next, cd->fcc, next));  /* Not UTF */
-   /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These
+     case OP_NOT:
-   opcodes are not used for multi-byte characters, because they are coded using
+     return c == next;
-   an XCLASS instead. */
+     case OP_NOTI:
-   case OP_NOT:
+     if (c == next) return TRUE;
-   return (c = *previous) == next;
+ #ifdef SUPPORT_UTF
+     if (utf)
-   case OP_NOTI:
+       {
-   if ((c = *previous) == next) return TRUE;
+       pcre_uint32 othercase;
- #ifdef SUPPORT_UTF8
+       if (next < 128) othercase = cd->fcc[next]; else
-   if (utf8)
-     {
-     unsigned int othercase;
-     if (next < 128) othercase = cd->fcc[next]; else
  #ifdef SUPPORT_UCP
-     othercase = UCD_OTHERCASE(next);
+       othercase = UCD_OTHERCASE(next);
  #else
-     othercase = NOTACHAR;
+       othercase = NOTACHAR;
  #endif
-     return (unsigned int)c == othercase;
+       return c == othercase;
-     }
+       }
-   else
+     else
- #endif  /* SUPPORT_UTF8 */
+ #endif  /* SUPPORT_UTF */
-   return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
+     return (c == TABLE_GET(next, cd->fcc, next));  /* Not UTF */
-   /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
+     /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
-   When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
+     When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
-   case OP_DIGIT:
+     case OP_DIGIT:
-   return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
+     return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;
-   case OP_NOT_DIGIT:
+     case OP_NOT_DIGIT:
-   return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
+     return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;
-   case OP_WHITESPACE:
+     case OP_WHITESPACE:
-   return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
+     return next > 255 || (cd->ctypes[next] & ctype_space) == 0;
-   case OP_NOT_WHITESPACE:
+     case OP_NOT_WHITESPACE:
-   return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
+     return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;
-   case OP_WORDCHAR:
+     case OP_WORDCHAR:
-   return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
+     return next > 255 || (cd->ctypes[next] & ctype_word) == 0;
-   case OP_NOT_WORDCHAR:
+     case OP_NOT_WORDCHAR:
-   return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
+     return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;
-   case OP_HSPACE:
+     case OP_HSPACE:
-   case OP_NOT_HSPACE:
+     case OP_NOT_HSPACE:
-   switch(next)
+     switch(next)
-     {
+       {
-     case 0x09:
+       HSPACE_CASES:
-     case 0x20:
+       return op_code == OP_NOT_HSPACE;
-     case 0xa0:
-     case 0x1680:
-     case 0x180e:
-     case 0x2000:
-     case 0x2001:
-     case 0x2002:
-     case 0x2003:
-     case 0x2004:
-     case 0x2005:
-     case 0x2006:
-     case 0x2007:
-     case 0x2008:
-     case 0x2009:
-     case 0x200A:
-     case 0x202f:
-     case 0x205f:
-     case 0x3000:
-     return op_code == OP_NOT_HSPACE;
-     default:
-     return op_code != OP_NOT_HSPACE;
-     }
-   case OP_ANYNL:
+       default:
-   case OP_VSPACE:
+       return op_code != OP_NOT_HSPACE;
-   case OP_NOT_VSPACE:
+       }
-   switch(next)
-     {
+     case OP_ANYNL:
-     case 0x0a:
+     case OP_VSPACE:
-     case 0x0b:
+     case OP_NOT_VSPACE:
-     case 0x0c:
+     switch(next)
-     case 0x0d:
+       {
-     case 0x85:
+       VSPACE_CASES:
-     case 0x2028:
+       return op_code == OP_NOT_VSPACE;
-     case 0x2029:
-     return op_code == OP_NOT_VSPACE;
+       default:
-     default:
+       return op_code != OP_NOT_VSPACE;
-     return op_code != OP_NOT_VSPACE;
+       }
-     }
  #ifdef SUPPORT_UCP
-   case OP_PROP:
+     case OP_PROP:
-   return check_char_prop(next, previous[0], previous[1], FALSE);
+     return check_char_prop(next, previous[0], previous[1], FALSE);
-   case OP_NOTPROP:
+     case OP_NOTPROP:
-   return check_char_prop(next, previous[0], previous[1], TRUE);
+     return check_char_prop(next, previous[0], previous[1], TRUE);
  #endif
-   default:
+     default:
-   return FALSE;
+     return FALSE;
+     }
    }
  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
  generated only when PCRE_UCP is *not* set, that is, when only ASCII
-Line 3065 
 switch(op_code)
+Line 3278 
 switch(op_code)
    {
    case OP_CHAR:
    case OP_CHARI:
- #ifdef SUPPORT_UTF8
+   switch(escape)
-   GETCHARTEST(c, previous);
- #else
-   c = *previous;
- #endif
-   switch(-next)
      {
      case ESC_d:
-     return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
+     return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;
      case ESC_D:
-     return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
+     return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;
      case ESC_s:
-     return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
+     return c > 255 || (cd->ctypes[c] & ctype_space) == 0;
      case ESC_S:
-     return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
+     return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;
      case ESC_w:
-     return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
+     return c > 255 || (cd->ctypes[c] & ctype_word) == 0;
      case ESC_W:
-     return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
+     return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;
      case ESC_h:
      case ESC_H:
      switch(c)
        {
-       case 0x09:
+       HSPACE_CASES:
-       case 0x20:
+       return escape != ESC_h;
-       case 0xa0:
-       case 0x1680:
-       case 0x180e:
-       case 0x2000:
-       case 0x2001:
-       case 0x2002:
-       case 0x2003:
-       case 0x2004:
-       case 0x2005:
-       case 0x2006:
-       case 0x2007:
-       case 0x2008:
-       case 0x2009:
-       case 0x200A:
-       case 0x202f:
-       case 0x205f:
-       case 0x3000:
-       return -next != ESC_h;
        default:
-       return -next == ESC_h;
+       return escape == ESC_h;
        }
      case ESC_v:
      case ESC_V:
      switch(c)
        {
-       case 0x0a:
+       VSPACE_CASES:
-       case 0x0b:
+       return escape != ESC_v;
-       case 0x0c:
-       case 0x0d:
-       case 0x85:
-       case 0x2028:
-       case 0x2029:
-       return -next != ESC_v;
        default:
-       return -next == ESC_v;
+       return escape == ESC_v;
        }
      /* When PCRE_UCP is set, these values get generated for \d etc. Find
      their substitutions and process them. The result will always be either
-     -ESC_p or -ESC_P. Then fall through to process those values. */
+     ESC_p or ESC_P. Then fall through to process those values. */
  #ifdef SUPPORT_UCP
      case ESC_du:
-Line 3147 
 switch(op_code)
+Line 3333 
 switch(op_code)
      case ESC_SU:
        {
        int temperrorcode = 0;
-       ptr = substitutes[-next - ESC_DU];
+       ptr = substitutes[escape - ESC_DU];
-       next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
+       escape = check_escape(&ptr, &next, &temperrorcode, 0, options, FALSE);
        if (temperrorcode != 0) return FALSE;
        ptr++;    /* For compatibility */
        }
-Line 3157 
 switch(op_code)
+Line 3343 
 switch(op_code)
      case ESC_p:
      case ESC_P:
        {
-       int ptype, pdata, errorcodeptr;
+       unsigned int ptype = 0, pdata = 0;
+       int errorcodeptr;
        BOOL negated;
        ptr--;      /* Make ptr point at the p or P */
-       ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
+       if (!get_ucp(&ptr, &negated, &ptype, &pdata, &errorcodeptr))
-       if (ptype < 0) return FALSE;
+         return FALSE;
        ptr++;      /* Point past the final curly ket */
        /* If the property item is optional, we have to give up. (When generated
-Line 3170 
 switch(op_code)
+Line 3357 
 switch(op_code)
        to the original \d etc. At this point, ptr will point to a zero byte. */
        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
-         strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
+         STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
            return FALSE;
        /* Do the property check. */
-       return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
+       return check_char_prop(c, ptype, pdata, (escape == ESC_P) != negated);
        }
  #endif
-Line 3190 
 switch(op_code)
+Line 3377 
 switch(op_code)
    these op-codes are never generated.) */
    case OP_DIGIT:
-   return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
+   return escape == ESC_D || escape == ESC_s || escape == ESC_W ||
-          next == -ESC_h || next == -ESC_v || next == -ESC_R;
+          escape == ESC_h || escape == ESC_v || escape == ESC_R;
    case OP_NOT_DIGIT:
-   return next == -ESC_d;
+   return escape == ESC_d;
    case OP_WHITESPACE:
-   return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
+   return escape == ESC_S || escape == ESC_d || escape == ESC_w;
    case OP_NOT_WHITESPACE:
-   return next == -ESC_s || next == -ESC_h || next == -ESC_v;
+   return escape == ESC_s || escape == ESC_h || escape == ESC_v || escape == ESC_R;
    case OP_HSPACE:
-   return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
+   return escape == ESC_S || escape == ESC_H || escape == ESC_d ||
-          next == -ESC_w || next == -ESC_v || next == -ESC_R;
+          escape == ESC_w || escape == ESC_v || escape == ESC_R;
    case OP_NOT_HSPACE:
-   return next == -ESC_h;
+   return escape == ESC_h;
    /* Can't have \S in here because VT matches \S (Perl anomaly) */
    case OP_ANYNL:
    case OP_VSPACE:
-   return next == -ESC_V || next == -ESC_d || next == -ESC_w;
+   return escape == ESC_V || escape == ESC_d || escape == ESC_w;
    case OP_NOT_VSPACE:
-   return next == -ESC_v || next == -ESC_R;
+   return escape == ESC_v || escape == ESC_R;
    case OP_WORDCHAR:
-   return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
+   return escape == ESC_W || escape == ESC_s || escape == ESC_h ||
-          next == -ESC_v || next == -ESC_R;
+          escape == ESC_v || escape == ESC_R;
    case OP_NOT_WORDCHAR:
-   return next == -ESC_w || next == -ESC_d;
+   return escape == ESC_w || escape == ESC_d;
    default:
    return FALSE;
-Line 3234 
 switch(op_code)
+Line 3421 
 switch(op_code)
  /*************************************************
+ *        Add a character or range to a class     *
+ *************************************************/
+ /* This function packages up the logic of adding a character or range of
+ characters to a class. The character values in the arguments will be within the
+ valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
+ mutually recursive with the function immediately below.
+ Arguments:
+   classbits     the bit map for characters < 256
+   uchardptr     points to the pointer for extra data
+   options       the options word
+   cd            contains pointers to tables etc.
+   start         start of range character
+   end           end of range character
+ Returns:        the number of < 256 characters added
+                 the pointer to extra data is updated
+ */
+ static int
+ add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
+   compile_data *cd, pcre_uint32 start, pcre_uint32 end)
+ {
+ pcre_uint32 c;
+ int n8 = 0;
+ /* If caseless matching is required, scan the range and process alternate
+ cases. In Unicode, there are 8-bit characters that have alternate cases that
+ are greater than 255 and vice-versa. Sometimes we can just extend the original
+ range. */
+ if ((options & PCRE_CASELESS) != 0)
+   {
+ #ifdef SUPPORT_UCP
+   if ((options & PCRE_UTF8) != 0)
+     {
+     int rc;
+     pcre_uint32 oc, od;
+     options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
+     c = start;
+     while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
+       {
+       /* Handle a single character that has more than one other case. */
+       if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
+         PRIV(ucd_caseless_sets) + rc, oc);
+       /* Do nothing if the other case range is within the original range. */
+       else if (oc >= start && od <= end) continue;
+       /* Extend the original range if there is overlap, noting that if oc < c, we
+       can't have od > end because a subrange is always shorter than the basic
+       range. Otherwise, use a recursive call to add the additional range. */
+       else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
+       else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
+       else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
+       }
+     }
+   else
+ #endif  /* SUPPORT_UCP */
+   /* Not UTF-mode, or no UCP */
+   for (c = start; c <= end && c < 256; c++)
+     {
+     SETBIT(classbits, cd->fcc[c]);
+     n8++;
+     }
+   }
+ /* Now handle the original range. Adjust the final value according to the bit
+ length - this means that the same lists of (e.g.) horizontal spaces can be used
+ in all cases. */
+ #if defined COMPILE_PCRE8
+ #ifdef SUPPORT_UTF
+   if ((options & PCRE_UTF8) == 0)
+ #endif
+   if (end > 0xff) end = 0xff;
+ #elif defined COMPILE_PCRE16
+ #ifdef SUPPORT_UTF
+   if ((options & PCRE_UTF16) == 0)
+ #endif
+   if (end > 0xffff) end = 0xffff;
+ #endif /* COMPILE_PCRE[8|16] */
+ /* If all characters are less than 256, use the bit map. Otherwise use extra
+ data. */
+ if (end < 0x100)
+   {
+   for (c = start; c <= end; c++)
+     {
+     n8++;
+     SETBIT(classbits, c);
+     }
+   }
+ else
+   {
+   pcre_uchar *uchardata = *uchardptr;
+ #ifdef SUPPORT_UTF
+   if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
+     {
+     if (start < end)
+       {
+       *uchardata++ = XCL_RANGE;
+       uchardata += PRIV(ord2utf)(start, uchardata);
+       uchardata += PRIV(ord2utf)(end, uchardata);
+       }
+     else if (start == end)
+       {
+       *uchardata++ = XCL_SINGLE;
+       uchardata += PRIV(ord2utf)(start, uchardata);
+       }
+     }
+   else
+ #endif  /* SUPPORT_UTF */
+   /* Without UTF support, character values are constrained by the bit length,
+   and can only be > 256 for 16-bit and 32-bit libraries. */
+ #ifdef COMPILE_PCRE8
+     {}
+ #else
+   if (start < end)
+     {
+     *uchardata++ = XCL_RANGE;
+     *uchardata++ = start;
+     *uchardata++ = end;
+     }
+   else if (start == end)
+     {
+     *uchardata++ = XCL_SINGLE;
+     *uchardata++ = start;
+     }
+ #endif
+   *uchardptr = uchardata;   /* Updata extra data pointer */
+   }
+ return n8;    /* Number of 8-bit characters */
+ }
+ /*************************************************
+ *        Add a list of characters to a class     *
+ *************************************************/
+ /* This function is used for adding a list of case-equivalent characters to a
+ class, and also for adding a list of horizontal or vertical whitespace. If the
+ list is in order (which it should be), ranges of characters are detected and
+ handled appropriately. This function is mutually recursive with the function
+ above.
+ Arguments:
+   classbits     the bit map for characters < 256
+   uchardptr     points to the pointer for extra data
+   options       the options word
+   cd            contains pointers to tables etc.
+   p             points to row of 32-bit values, terminated by NOTACHAR
+   except        character to omit; this is used when adding lists of
+                   case-equivalent characters to avoid including the one we
+                   already know about
+ Returns:        the number of < 256 characters added
+                 the pointer to extra data is updated
+ */
+ static int
+ add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
+   compile_data *cd, const pcre_uint32 *p, unsigned int except)
+ {
+ int n8 = 0;
+ while (p[0] < NOTACHAR)
+   {
+   int n = 0;
+   if (p[0] != except)
+     {
+     while(p[n+1] == p[0] + n + 1) n++;
+     n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
+     }
+   p += n + 1;
+   }
+ return n8;
+ }
+ /*************************************************
+ *    Add characters not in a list to a class     *
+ *************************************************/
+ /* This function is used for adding the complement of a list of horizontal or
+ vertical whitespace to a class. The list must be in order.
+ Arguments:
+   classbits     the bit map for characters < 256
+   uchardptr     points to the pointer for extra data
+   options       the options word
+   cd            contains pointers to tables etc.
+   p             points to row of 32-bit values, terminated by NOTACHAR
+ Returns:        the number of < 256 characters added
+                 the pointer to extra data is updated
+ */
+ static int
+ add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
+   int options, compile_data *cd, const pcre_uint32 *p)
+ {
+ BOOL utf = (options & PCRE_UTF8) != 0;
+ int n8 = 0;
+ if (p[0] > 0)
+   n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
+ while (p[0] < NOTACHAR)
+   {
+   while (p[1] == p[0] + 1) p++;
+   n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
+     (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
+   p++;
+   }
+ return n8;
+ }
+ /*************************************************
  *           Compile one branch                   *
  *************************************************/
-Line 3248 
 Arguments:
+Line 3673 
 Arguments:
    codeptr        points to the pointer to the current code point
    ptrptr         points to the current pattern pointer
    errorcodeptr   points to error code variable
-   firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
+   firstcharptr    place to put the first required character
-   reqbyteptr     set to the last literal character required, else < 0
+   firstcharflagsptr place to put the first character flags, or a negative number
+   reqcharptr     place to put the last required character
+   reqcharflagsptr place to put the last required character flags, or a negative number
    bcptr          points to current branch chain
    cond_depth     conditional nesting depth
    cd             contains pointers to tables etc.
-Line 3261 
 Returns:         TRUE on success
+Line 3688 
 Returns:         TRUE on success
  */
  static BOOL
- compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
+ compile_branch(int *optionsptr, pcre_uchar **codeptr,
-   int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
+   const pcre_uchar **ptrptr, int *errorcodeptr,
-   int cond_depth, compile_data *cd, int *lengthptr)
+   pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
+   pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
+   branch_chain *bcptr, int cond_depth,
+   compile_data *cd, int *lengthptr)
  {
  int repeat_type, op_type;
  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
  int bravalue = 0;
  int greedy_default, greedy_non_default;
- int firstbyte, reqbyte;
+ pcre_uint32 firstchar, reqchar;
- int zeroreqbyte, zerofirstbyte;
+ pcre_int32 firstcharflags, reqcharflags;
- int req_caseopt, reqvary, tempreqvary;
+ pcre_uint32 zeroreqchar, zerofirstchar;
+ pcre_int32 zeroreqcharflags, zerofirstcharflags;
+ pcre_int32 req_caseopt, reqvary, tempreqvary;
  int options = *optionsptr;               /* May change dynamically */
  int after_manual_callout = 0;
  int length_prevgroup = 0;
- register int c;
+ register pcre_uint32 c;
- register uschar *code = *codeptr;
+ int escape;
- uschar *last_code = code;
+ register pcre_uchar *code = *codeptr;
- uschar *orig_code = code;
+ pcre_uchar *last_code = code;
- uschar *tempcode;
+ pcre_uchar *orig_code = code;
+ pcre_uchar *tempcode;
  BOOL inescq = FALSE;
- BOOL groupsetfirstbyte = FALSE;
+ BOOL groupsetfirstchar = FALSE;
- const uschar *ptr = *ptrptr;
+ const pcre_uchar *ptr = *ptrptr;
- const uschar *tempptr;
+ const pcre_uchar *tempptr;
- const uschar *nestptr = NULL;
+ const pcre_uchar *nestptr = NULL;
- uschar *previous = NULL;
+ pcre_uchar *previous = NULL;
- uschar *previous_callout = NULL;
+ pcre_uchar *previous_callout = NULL;
- uschar *save_hwm = NULL;
+ pcre_uchar *save_hwm = NULL;
- uschar classbits[32];
+ pcre_uint8 classbits[32];
  /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
  must not do this for other options (e.g. PCRE_EXTENDED) because they may change
  dynamically as we process the pattern. */
- #ifdef SUPPORT_UTF8
+ #ifdef SUPPORT_UTF
- BOOL class_utf8;
+ /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
- BOOL utf8 = (options & PCRE_UTF8) != 0;
+ BOOL utf = (options & PCRE_UTF8) != 0;
- uschar *class_utf8data;
+ #ifndef COMPILE_PCRE32
- uschar *class_utf8data_base;
+ pcre_uchar utf_chars[6];
- uschar utf8_char[6];
+ #endif
  #else
- BOOL utf8 = FALSE;
+ BOOL utf = FALSE;
+ #endif
+ /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
+ class_uchardata always so that it can be passed to add_to_class() always,
+ though it will not be used in non-UTF 8-bit cases. This avoids having to supply
+ alternative calls for the different cases. */
+ pcre_uchar *class_uchardata;
+ #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
+ BOOL xclass;
+ pcre_uchar *class_uchardata_base;
  #endif
  #ifdef PCRE_DEBUG
-Line 3315 
 greedy_non_default = greedy_default ^ 1;
+Line 3759 
 greedy_non_default = greedy_default ^ 1;
  /* Initialize no first byte, no required byte. REQ_UNSET means "no char
  matching encountered yet". It gets changed to REQ_NONE if we hit something that
- matches a non-fixed char first char; reqbyte just remains unset if we never
+ matches a non-fixed char first char; reqchar just remains unset if we never
  find one.
  When we hit a repeat whose minimum is zero, we may have to adjust these values
  to take the zero repeat into account. This is implemented by setting them to
- zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual
+ zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
  item types that can be repeated set these backoff variables appropriately. */
- firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;
+ firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
+ firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
- /* The variable req_caseopt contains either the REQ_CASELESS value or zero,
+ /* The variable req_caseopt contains either the REQ_CASELESS value
- according to the current setting of the caseless flag. REQ_CASELESS is a bit
+ or zero, according to the current setting of the caseless flag. The
- value > 255. It is added into the firstbyte or reqbyte variables to record the
+ REQ_CASELESS leaves the lower 28 bit empty. It is added into the
- case status of the value. This is used only for ASCII characters. */
+ firstchar or reqchar variables to record the case status of the
+ value. This is used only for ASCII characters. */
- req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
+ req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
  /* Switch on next character until the end of the branch */
-Line 3342 
 for (;; ptr++)
+Line 3788 
 for (;; ptr++)
    BOOL is_quantifier;
    BOOL is_recurse;
    BOOL reset_bracount;
-   int class_charcount;
+   int class_has_8bitchar;
-   int class_lastchar;
+   int class_one_char;
    int newoptions;
    int recno;
    int refsign;
    int skipbytes;
-   int subreqbyte;
+   pcre_uint32 subreqchar, subfirstchar;
-   int subfirstbyte;
+   pcre_int32 subreqcharflags, subfirstcharflags;
    int terminator;
-   int mclength;
+   unsigned int mclength;
-   int tempbracount;
+   unsigned int tempbracount;
-   uschar mcbuffer[8];
+   pcre_uint32 ec;
+   pcre_uchar mcbuffer[8];
-   /* Get next byte in the pattern */
+   /* Get next character in the pattern */
    c = *ptr;
    /* If we are at the end of a nested substitution, revert to the outer level
    string. Nesting only happens one level deep. */
-   if (c == 0 && nestptr != NULL)
+   if (c == CHAR_NULL && nestptr != NULL)
      {
      ptr = nestptr;
      nestptr = NULL;
-Line 3377 
 for (;; ptr++)
+Line 3824 
 for (;; ptr++)
  #ifdef PCRE_DEBUG
      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
  #endif
      if (code > cd->start_workspace + cd->workspace_size -
          WORK_SIZE_SAFETY_MARGIN)                       /* Check for overrun */
        {
        *errorcodeptr = ERR52;
-Line 3401 
 for (;; ptr++)
+Line 3848 
 for (;; ptr++)
        }
      *lengthptr += (int)(code - last_code);
-     DPRINTF(("length=%d added %d c=%c\n", *lengthptr, (int)(code - last_code),
+     DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
-       c));
+       (int)(code - last_code), c, c));
      /* If "previous" is set and it is not at the start of the work space, move
      it back to there, in order to avoid filling up the work space. Otherwise,
-Line 3412 
 for (;; ptr++)
+Line 3859 
 for (;; ptr++)
        {
        if (previous > orig_code)
          {
-         memmove(orig_code, previous, code - previous);
+         memmove(orig_code, previous, IN_UCHARS(code - previous));
          code -= previous - orig_code;
          previous = orig_code;
          }
-Line 3428 
 for (;; ptr++)
+Line 3875 
 for (;; ptr++)
    /* In the real compile phase, just check the workspace used by the forward
    reference list. */
    else if (cd->hwm > cd->start_workspace + cd->workspace_size -
             WORK_SIZE_SAFETY_MARGIN)
      {
      *errorcodeptr = ERR52;
-Line 3437 
 for (;; ptr++)
+Line 3884 
 for (;; ptr++)
    /* If in \Q...\E, check for the end; if not, we have a literal */
-   if (inescq && c != 0)
+   if (inescq && c != CHAR_NULL)
      {
      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
        {
-Line 3481 
 for (;; ptr++)
+Line 3928 
 for (;; ptr++)
    if ((options & PCRE_EXTENDED) != 0)
      {
-     if ((cd->ctypes[c] & ctype_space) != 0) continue;
+     if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;
      if (c == CHAR_NUMBER_SIGN)
        {
        ptr++;
-       while (*ptr != 0)
+       while (*ptr != CHAR_NULL)
          {
          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
          ptr++;
- #ifdef SUPPORT_UTF8
+ #ifdef SUPPORT_UTF
-         if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
+         if (utf) FORWARDCHAR(ptr);
  #endif
          }
-       if (*ptr != 0) continue;
+       if (*ptr != CHAR_NULL) continue;
        /* Else fall through to handle end of string */
        c = 0;
-Line 3514 
 for (;; ptr++)
+Line 3961 
 for (;; ptr++)
      case 0:                        /* The branch terminates at string end */
      case CHAR_VERTICAL_LINE:       /* or | or ) */
      case CHAR_RIGHT_PARENTHESIS:
-     *firstbyteptr = firstbyte;
+     *firstcharptr = firstchar;
-     *reqbyteptr = reqbyte;
+     *firstcharflagsptr = firstcharflags;
+     *reqcharptr = reqchar;
+     *reqcharflagsptr = reqcharflags;
      *codeptr = code;
      *ptrptr = ptr;
      if (lengthptr != NULL)
-Line 3539 
 for (;; ptr++)
+Line 3988 
 for (;; ptr++)
      previous = NULL;
      if ((options & PCRE_MULTILINE) != 0)
        {
-       if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
+       if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
        *code++ = OP_CIRCM;
        }
      else *code++ = OP_CIRC;
-Line 3551 
 for (;; ptr++)
+Line 4000 
 for (;; ptr++)
      break;
      /* There can never be a first char if '.' is first, whatever happens about
-     repeats. The value of reqbyte doesn't change either. */
+     repeats. The value of reqchar doesn't change either. */
      case CHAR_DOT:
-     if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
+     if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
-     zerofirstbyte = firstbyte;
+     zerofirstchar = firstchar;
-     zeroreqbyte = reqbyte;
+     zerofirstcharflags = firstcharflags;
+     zeroreqchar = reqchar;
+     zeroreqcharflags = reqcharflags;
      previous = code;
      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
      break;
-Line 3611 
 for (;; ptr++)
+Line 4062 
 for (;; ptr++)
          {
          if (ptr[1] == CHAR_E)
            ptr++;
-         else if (strncmp((const char *)ptr+1,
+         else if (STRNCMP_UC_C8(ptr + 1, STR_Q STR_BACKSLASH STR_E, 3) == 0)
-                           STR_Q STR_BACKSLASH STR_E, 3) == 0)
            ptr += 3;
          else
            break;
-Line 3631 
 for (;; ptr++)
+Line 4081 
 for (;; ptr++)
          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
        {
        *code++ = negate_class? OP_ALLANY : OP_FAIL;
-       if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
+       if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
-       zerofirstbyte = firstbyte;
+       zerofirstchar = firstchar;
+       zerofirstcharflags = firstcharflags;
        break;
        }
-Line 3642 
 for (;; ptr++)
+Line 4093 
 for (;; ptr++)
      should_flip_negation = FALSE;
-     /* Keep a count of chars with values < 256 so that we can optimize the case
+     /* For optimization purposes, we track some properties of the class:
-     of just a single character (as long as it's < 256). However, For higher
+     class_has_8bitchar will be non-zero if the class contains at least one <
-     valued UTF-8 characters, we don't yet do any optimization. */
+character; class_one_char will be 1 if the class contains just one
+     character. */
-     class_charcount = 0;
+     class_has_8bitchar = 0;
-     class_lastchar = -1;
+     class_one_char = 0;
      /* Initialize the 32-char bit map to all zeros. We build the map in a
-     temporary bit of memory, in case the class contains only 1 character (less
+     temporary bit of memory, in case the class contains fewer than two
-     than 256), because in that case the compiled code doesn't use the bit map.
+-bit characters because in that case the compiled code doesn't use the bit
-     */
+     map. */
-     memset(classbits, 0, 32 * sizeof(uschar));
+     memset(classbits, 0, 32 * sizeof(pcre_uint8));
- #ifdef SUPPORT_UTF8
+ #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
-     class_utf8 = FALSE;                       /* No chars >= 256 */
+     xclass = FALSE;
-     class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
+     class_uchardata = code + LINK_SIZE + 2;   /* For XCLASS items */
-     class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
+     class_uchardata_base = class_uchardata;   /* Save the start */
  #endif
      /* Process characters until ] is reached. By writing this as a "do" it
      means that an initial ] is taken as a data character. At the start of the
      loop, c contains the first byte of the character. */
-     if (c != 0) do
+     if (c != CHAR_NULL) do
        {
-       const uschar *oldptr;
+       const pcre_uchar *oldptr;
- #ifdef SUPPORT_UTF8
+ #ifdef SUPPORT_UTF
-       if (utf8 && c > 127)
+       if (utf && HAS_EXTRALEN(c))
          {                           /* Braces are required because the */
          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
          }
+ #endif
-       /* In the pre-compile phase, accumulate the length of any UTF-8 extra
+ #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
+       /* In the pre-compile phase, accumulate the length of any extra
        data and reset the pointer. This is so that very large classes that
-       contain a zillion UTF-8 characters no longer overwrite the work space
+       contain a zillion > 255 characters no longer overwrite the work space
-       (which is on the stack). */
+       (which is on the stack). We have to remember that there was XCLASS data,
+       however. */
-       if (lengthptr != NULL)
-         {
+       if (lengthptr != NULL && class_uchardata > class_uchardata_base)
-         *lengthptr += class_utf8data - class_utf8data_base;
+         {
-         class_utf8data = class_utf8data_base;
+         xclass = TRUE;
+         *lengthptr += class_uchardata - class_uchardata_base;
+         class_uchardata = class_uchardata_base;
          }
  #endif
        /* Inside \Q...\E everything is literal except \E */
-Line 3714 
 for (;; ptr++)
+Line 4169 
 for (;; ptr++)
          {
          BOOL local_negate = FALSE;
          int posix_class, taboffset, tabopt;
-         register const uschar *cbits = cd->cbits;
+         register const pcre_uint8 *cbits = cd->cbits;
-         uschar pbits[32];
+         pcre_uint8 pbits[32];
          if (ptr[1] != CHAR_COLON)
            {
-Line 3770 
 for (;; ptr++)
+Line 4225 
 for (;; ptr++)
          /* Copy in the first table (always present) */
          memcpy(pbits, cbits + posix_class_maps[posix_class],
-* sizeof(uschar));
+* sizeof(pcre_uint8));
          /* If there is a second table, add or remove it as required. */
-Line 3785 
 for (;; ptr++)
+Line 4240 
 for (;; ptr++)
              for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];
            }
-         /* Not see if we need to remove any special characters. An option
+         /* Now see if we need to remove any special characters. An option
          value of 1 removes vertical space and 2 removes underscore. */
          if (tabopt < 0) tabopt = -tabopt;
-Line 3801 
 for (;; ptr++)
+Line 4256 
 for (;; ptr++)
            for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
          ptr = tempptr + 1;
-         class_charcount = 10;  /* Set > 1; assumes more than 1 per class */
+         /* Every class contains at least one < 256 character. */
+         class_has_8bitchar = 1;
+         /* Every class contains at least two characters. */
+         class_one_char = 2;
          continue;    /* End of POSIX syntax handling */
          }
        /* Backslash may introduce a single character, or it may introduce one
        of the specials, which just set a flag. The sequence \b is a special
        case. Inside a class (and only there) it is treated as backspace. We
-       assume that other escapes have more than one character in them, so set
+       assume that other escapes have more than one character in them, so
-       class_charcount bigger than one. Unrecognized escapes fall through and
+       speculatively set both class_has_8bitchar and class_one_char bigger
-       are either treated as literal characters (by default), or are faulted if
+       than one. Unrecognized escapes fall through and are either treated
+       as literal characters (by default), or are faulted if
        PCRE_EXTRA is set. */
        if (c == CHAR_BACKSLASH)
          {
-         c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
+         escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options,
+           TRUE);
          if (*errorcodeptr != 0) goto FAILED;
+         if (escape == 0) c = ec;
-         if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */
+         else if (escape == ESC_b) c = CHAR_BS; /* \b is backspace in a class */
-         else if (-c == ESC_N)            /* \N is not supported in a class */
+         else if (escape == ESC_N)          /* \N is not supported in a class */
            {
            *errorcodeptr = ERR71;
            goto FAILED;
            }
-         else if (-c == ESC_Q)            /* Handle start of quoted string */
+         else if (escape == ESC_Q)            /* Handle start of quoted string */
            {
            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
              {
-Line 3833 
 for (;; ptr++)
+Line 4293 
 for (;; ptr++)
            else inescq = TRUE;
            continue;
            }
-         else if (-c == ESC_E) continue;  /* Ignore orphan \E */
+         else if (escape == ESC_E) continue;  /* Ignore orphan \E */
-         if (c < 0)
+         else
            {
-           register const uschar *cbits = cd->cbits;
+           register const pcre_uint8 *cbits = cd->cbits;
-           class_charcount += 2;     /* Greater than 1 is what matters */
+           /* Every class contains at least two < 256 characters. */
+           class_has_8bitchar++;
+           /* Every class contains at least two characters. */
+           class_one_char += 2;
-           switch (-c)
+           switch (escape)
              {
  #ifdef SUPPORT_UCP
              case ESC_du:     /* These are the values given for \d etc */
-Line 3850 
 for (;; ptr++)
+Line 4313 
 for (;; ptr++)
              case ESC_su:     /* of the default ASCII testing. */
              case ESC_SU:
              nestptr = ptr;
-             ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
+             ptr = substitutes[escape - ESC_DU] - 1;  /* Just before substitute */
-             class_charcount -= 2;                /* Undo! */
+             class_has_8bitchar--;                /* Undo! */
              continue;
  #endif
              case ESC_d:
-Line 3874 
 for (;; ptr++)
+Line 4337 
 for (;; ptr++)
              /* Perl 5.004 onwards omits VT from \s, but we must preserve it
              if it was previously set by something earlier in the character
-             class. */
+             class. Luckily, the value of CHAR_VT is 0x0b in both ASCII and
+             EBCDIC, so we lazily just adjust the appropriate bit. */
              case ESC_s:
              classbits[0] |= cbits[cbit_space];
-Line 3888 
 for (;; ptr++)
+Line 4352 
 for (;; ptr++)
              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
              continue;
+             /* The rest apply in both UCP and non-UCP cases. */
              case ESC_h:
-             SETBIT(classbits, 0x09); /* VT */
+             (void)add_list_to_class(classbits, &class_uchardata, options, cd,
-             SETBIT(classbits, 0x20); /* SPACE */
+               PRIV(hspace_list), NOTACHAR);
-             SETBIT(classbits, 0xa0); /* NSBP */
- #ifdef SUPPORT_UTF8
-             if (utf8)
-               {
-               class_utf8 = TRUE;
-               *class_utf8data++ = XCL_SINGLE;
-               class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
-               *class_utf8data++ = XCL_SINGLE;
-               class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
-               *class_utf8data++ = XCL_RANGE;
-               class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
-               class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
-               *class_utf8data++ = XCL_SINGLE;
-               class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
-               *class_utf8data++ = XCL_SINGLE;
-               class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
-               *class_utf8data++ = XCL_SINGLE;
-               class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
-               }
- #endif
              continue;
              case ESC_H:
-             for (c = 0; c < 32; c++)
+             (void)add_not_list_to_class(classbits, &class_uchardata, options,
-               {
+               cd, PRIV(hspace_list));
-               int x = 0xff;
-               switch (c)
-                 {
-                 case 0x09/8: x ^= 1 << (0x09%8); break;
-                 case 0x20/8: x ^= 1 << (0x20%8); break;
-                 case 0xa0/8: x ^= 1 << (0xa0%8); break;
-                 default: break;
-                 }
-               classbits[c] |= x;
-               }
- #ifdef SUPPORT_UTF8
-             if (utf8)
-               {
-               class_utf8 = TRUE;
-               *class_utf8data++ = XCL_RANGE;
-               class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
-               class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
-               *class_utf8data++ = XCL_RANGE;
-               class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
-               class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
-               *class_utf8data++ = XCL_RANGE;
-               class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
-               class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
-               *class_utf8data++ = XCL_RANGE;
-               class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
-               class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
-               *class_utf8data++ = XCL_RANGE;
-               class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
-               class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
-               *class_utf8data++ = XCL_RANGE;
-               class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
-               class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
-               *class_utf8data++ = XCL_RANGE;
-               class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
-               class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
-               }
- #endif
              continue;
              case ESC_v:
-             SETBIT(classbits, 0x0a); /* LF */
+             (void)add_list_to_class(classbits, &class_uchardata, options, cd,
-             SETBIT(classbits, 0x0b); /* VT */
+               PRIV(vspace_list), NOTACHAR);
-             SETBIT(classbits, 0x0c); /* FF */
-             SETBIT(classbits, 0x0d); /* CR */
-             SETBIT(classbits, 0x85); /* NEL */
- #ifdef SUPPORT_UTF8
-             if (utf8)
-               {
-               class_utf8 = TRUE;
-               *class_utf8data++ = XCL_RANGE;
-               class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
-               class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
-               }
- #endif
              continue;
              case ESC_V:
-             for (c = 0; c < 32; c++)
+             (void)add_not_list_to_class(classbits, &class_uchardata, options,
-               {
+               cd, PRIV(vspace_list));
-               int x = 0xff;
-               switch (c)
-                 {
-                 case 0x0a/8: x ^= 1 << (0x0a%8);
-                              x ^= 1 << (0x0b%8);
-                              x ^= 1 << (0x0c%8);
-                              x ^= 1 << (0x0d%8);
-                              break;
-                 case 0x85/8: x ^= 1 << (0x85%8); break;
-                 default: break;
-                 }
-               classbits[c] |= x;
-               }
- #ifdef SUPPORT_UTF8
-             if (utf8)
-               {
-               class_utf8 = TRUE;
-               *class_utf8data++ = XCL_RANGE;
-               class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
-               class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
-               *class_utf8data++ = XCL_RANGE;
-               class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
-               class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
-               }
- #endif
              continue;
  #ifdef SUPPORT_UCP
-Line 4009 
 for (;; ptr++)
+Line 4379 
 for (;; ptr++)
              case ESC_P:
                {
                BOOL negated;
-               int pdata;
+               unsigned int ptype = 0, pdata = 0;
-               int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
+               if (!get_ucp(&ptr, &negated, &ptype, &pdata, errorcodeptr))
-               if (ptype < 0) goto FAILED;
+                 goto FAILED;
-               class_utf8 = TRUE;
+               *class_uchardata++ = ((escape == ESC_p) != negated)?
-               *class_utf8data++ = ((-c == ESC_p) != negated)?
                  XCL_PROP : XCL_NOTPROP;
-               *class_utf8data++ = ptype;
+               *class_uchardata++ = ptype;
-               *class_utf8data++ = pdata;
+               *class_uchardata++ = pdata;
-               class_charcount -= 2;   /* Not a < 256 character */
+               class_has_8bitchar--;                /* Undo! */
                continue;
                }
  #endif
-Line 4031 
 for (;; ptr++)
+Line 4400 
 for (;; ptr++)
                *errorcodeptr = ERR7;
                goto FAILED;
                }
-             class_charcount -= 2;  /* Undo the default count from above */
+             class_has_8bitchar--;    /* Undo the speculative increase. */
-             c = *ptr;              /* Get the final character and fall through */
+             class_one_char -= 2;     /* Undo the speculative increase. */
+             c = *ptr;                /* Get the final character and fall through */
              break;
              }
            }
-         /* Fall through if we have a single character (c >= 0). This may be
+         /* Fall through if the escape just defined a single character (c >= 0).
-         greater than 256 in UTF-8 mode. */
+         This may be greater than 256. */
+         escape = 0;
          }   /* End of backslash handling */
-       /* A single character may be followed by '-' to form a range. However,
+       /* A character may be followed by '-' to form a range. However, Perl does
-       Perl does not permit ']' to be the end of the range. A '-' character
+       not permit ']' to be the end of the range. A '-' character at the end is
-       at the end is treated as a literal. Perl ignores orphaned \E sequences
+       treated as a literal. Perl ignores orphaned \E sequences entirely. The
-       entirely. The code for handling \Q and \E is messy. */
+       code for handling \Q and \E is messy. */
        CHECK_RANGE:
        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
-Line 4053 
 for (;; ptr++)
+Line 4425 
 for (;; ptr++)
          inescq = FALSE;
          ptr += 2;
          }
        oldptr = ptr;
-       /* Remember \r or \n */
+       /* Remember if \r or \n were explicitly used */
        if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
-Line 4064 
 for (;; ptr++)
+Line 4435 
 for (;; ptr++)
        if (!inescq && ptr[1] == CHAR_MINUS)
          {
-         int d;
+         pcre_uint32 d;
          ptr += 2;
          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
-Line 4080 
 for (;; ptr++)
+Line 4451 
 for (;; ptr++)
            break;
            }
-         if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
+         /* Minus (hyphen) at the end of a class is treated as a literal, so put
+         back the pointer and jump to handle the character that preceded it. */
+         if (*ptr == CHAR_NULL || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
            {
            ptr = oldptr;
-           goto LONE_SINGLE_CHARACTER;
+           goto CLASS_SINGLE_CHARACTER;
            }
- #ifdef SUPPORT_UTF8
+         /* Otherwise, we have a potential range; pick up the next character */
-         if (utf8)
+ #ifdef SUPPORT_UTF
+         if (utf)
            {                           /* Braces are required because the */
            GETCHARLEN(d, ptr, ptr);    /* macro generates multiple statements */
            }
-Line 4101 
 for (;; ptr++)
+Line 4477 
 for (;; ptr++)
          if (!inescq && d == CHAR_BACKSLASH)
            {
-           d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
+           int descape;
+           descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);
            if (*errorcodeptr != 0) goto FAILED;
-           /* \b is backspace; any other special means the '-' was literal */
+           /* \b is backspace; any other special means the '-' was literal. */
-           if (d < 0)
+           if (descape != 0)
              {
-             if (d == -ESC_b) d = CHAR_BS; else
+             if (descape == ESC_b) d = CHAR_BS; else
                {
                ptr = oldptr;
-               goto LONE_SINGLE_CHARACTER;  /* A few lines below */
+               goto CLASS_SINGLE_CHARACTER;  /* A few lines below */
                }
              }
            }
          /* Check that the two values are in the correct order. Optimize
-         one-character ranges */
+         one-character ranges. */
          if (d < c)
            {
            *errorcodeptr = ERR8;
            goto FAILED;
            }
+         if (d == c) goto CLASS_SINGLE_CHARACTER;  /* A few lines below */
-         if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
+         /* We have found a character range, so single character optimizations
+         cannot be done anymore. Any value greater than 1 indicates that there
+         is more than one character. */
-         /* Remember \r or \n */
+         class_one_char = 2;
-         if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
+         /* Remember an explicit \r or \n, and add the range to the class. */
-         /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless
+         if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
-         matching, we have to use an XCLASS with extra data items. Caseless
-         matching for characters > 127 is available only if UCP support is
-         available. */
- #ifdef SUPPORT_UTF8
-         if (utf8 && (d > 255 || ((options & PCRE_CASELESS) != 0 && d > 127)))
-           {
-           class_utf8 = TRUE;
-           /* With UCP support, we can find the other case equivalents of
-           the relevant characters. There may be several ranges. Optimize how
-           they fit with the basic range. */
- #ifdef SUPPORT_UCP
+         class_has_8bitchar +=
-           if ((options & PCRE_CASELESS) != 0)
+           add_to_class(classbits, &class_uchardata, options, cd, c, d);
-             {
-             unsigned int occ, ocd;
-             unsigned int cc = c;
-             unsigned int origd = d;
-             while (get_othercase_range(&cc, origd, &occ, &ocd))
-               {
-               if (occ >= (unsigned int)c &&
-                   ocd <= (unsigned int)d)
-                 continue;                          /* Skip embedded ranges */
-               if (occ < (unsigned int)c  &&
-                   ocd >= (unsigned int)c - 1)      /* Extend the basic range */
-                 {                                  /* if there is overlap,   */
-                 c = occ;                           /* noting that if occ < c */
-                 continue;                          /* we can't have ocd > d  */
-                 }                                  /* because a subrange is  */
-               if (ocd > (unsigned int)d &&
-                   occ <= (unsigned int)d + 1)      /* always shorter than    */
-                 {                                  /* the basic range.       */
-                 d = ocd;
-                 continue;
-                 }
-               if (occ == ocd)
+         continue;   /* Go get the next char in the class */
-                 {
+         }
-                 *class_utf8data++ = XCL_SINGLE;
-                 }
-               else
-                 {
-                 *class_utf8data++ = XCL_RANGE;
-                 class_utf8data += _pcre_ord2utf8(occ, class_utf8data);
-                 }
-               class_utf8data += _pcre_ord2utf8(ocd, class_utf8data);
-               }
-             }
- #endif  /* SUPPORT_UCP */
-           /* Now record the original range, possibly modified for UCP caseless
+       /* Handle a single character - we can get here for a normal non-escape
-           overlapping ranges. */
+       char, or after \ that introduces a single character or for an apparent
+       range that isn't. Only the value 1 matters for class_one_char, so don't
+       increase it if it is already 2 or more ... just in case there's a class
+       with a zillion characters in it. */
+       CLASS_SINGLE_CHARACTER:
+       if (class_one_char < 2) class_one_char++;
+       /* If class_one_char is 1, we have the first single character in the
+       class, and there have been no prior ranges, or XCLASS items generated by
+       escapes. If this is the final character in the class, we can optimize by
+       turning the item into a 1-character OP_CHAR[I] if it's positive, or
+       OP_NOT[I] if it's negative. In the positive case, it can cause firstchar
+       to be set. Otherwise, there can be no first char if this item is first,
+       whatever repeat count may follow. In the case of reqchar, save the
+       previous value for reinstating. */
-           *class_utf8data++ = XCL_RANGE;
+       if (class_one_char == 1 && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
-           class_utf8data += _pcre_ord2utf8(c, class_utf8data);
+         {
-           class_utf8data += _pcre_ord2utf8(d, class_utf8data);
+         ptr++;
+         zeroreqchar = reqchar;
-           /* With UCP support, we are done. Without UCP support, there is no
+         zeroreqcharflags = reqcharflags;
-           caseless matching for UTF-8 characters > 127; we can use the bit map
-           for the smaller ones. */
+         if (negate_class)
+           {
  #ifdef SUPPORT_UCP
-           continue;    /* With next character in the class */
+           int d;
- #else
+ #endif
-           if ((options & PCRE_CASELESS) == 0 || c > 127) continue;
+           if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
+           zerofirstchar = firstchar;
-           /* Adjust upper limit and fall through to set up the map */
+           zerofirstcharflags = firstcharflags;
-           d = 127;
+           /* For caseless UTF-8 mode when UCP support is available, check
+           whether this character has more than one other case. If so, generate
- #endif  /* SUPPORT_UCP */
+           a special OP_NOTPROP item instead of OP_NOTI. */
-           }
- #endif  /* SUPPORT_UTF8 */
-         /* We use the bit map for all cases when not in UTF-8 mode; else
-         ranges that lie entirely within 0-127 when there is UCP support; else
-         for partial ranges without UCP support. */
-         class_charcount += d - c + 1;
-         class_lastchar = d;
-         /* We can save a bit of time by skipping this in the pre-compile. */
-         if (lengthptr == NULL) for (; c <= d; c++)
+ #ifdef SUPPORT_UCP
-           {
+           if (utf && (options & PCRE_CASELESS) != 0 &&
-           classbits[c/8] |= (1 << (c&7));
+               (d = UCD_CASESET(c)) != 0)
-           if ((options & PCRE_CASELESS) != 0)
              {
-             int uc = cd->fcc[c];           /* flip case */
+             *code++ = OP_NOTPROP;
-             classbits[uc/8] |= (1 << (uc&7));
+             *code++ = PT_CLIST;
+             *code++ = d;
              }
-           }
+           else
+ #endif
-         continue;   /* Go get the next char in the class */
+           /* Char has only one other case, or UCP not available */
-         }
-       /* Handle a lone single character - we can get here for a normal
+             {
-       non-escape char, or after \ that introduces a single character or for an
+             *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
-       apparent range that isn't. */
+ #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
+             if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
+               code += PRIV(ord2utf)(c, code);
+             else
+ #endif
+               *code++ = c;
+             }
-       LONE_SINGLE_CHARACTER:
+           /* We are finished with this character class */
-       /* Handle a character that cannot go in the bit map */
+           goto END_CLASS;
+           }
- #ifdef SUPPORT_UTF8
+         /* For a single, positive character, get the value into mcbuffer, and
-       if (utf8 && (c > 255 || ((options & PCRE_CASELESS) != 0 && c > 127)))
+         then we can handle this with the normal one-character code. */
-         {
-         class_utf8 = TRUE;
-         *class_utf8data++ = XCL_SINGLE;
-         class_utf8data += _pcre_ord2utf8(c, class_utf8data);
- #ifdef SUPPORT_UCP
+ #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
-         if ((options & PCRE_CASELESS) != 0)
+         if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
+           mclength = PRIV(ord2utf)(c, mcbuffer);
+         else
+ #endif
            {
-           unsigned int othercase;
+           mcbuffer[0] = c;
-           if ((othercase = UCD_OTHERCASE(c)) != c)
+           mclength = 1;
-             {
-             *class_utf8data++ = XCL_SINGLE;
-             class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
-             }
            }
- #endif  /* SUPPORT_UCP */
+         goto ONE_CHAR;
+         }       /* End of 1-char optimization */
-         }
+       /* There is more than one character in the class, or an XCLASS item
-       else
+       has been generated. Add this character to the class. */
- #endif  /* SUPPORT_UTF8 */
-       /* Handle a single-byte character */
+       class_has_8bitchar +=
-         {
+         add_to_class(classbits, &class_uchardata, options, cd, c, c);
-         classbits[c/8] |= (1 << (c&7));
-         if ((options & PCRE_CASELESS) != 0)
-           {
-           c = cd->fcc[c];   /* flip case */
-           classbits[c/8] |= (1 << (c&7));
-           }
-         class_charcount++;
-         class_lastchar = c;
-         }
        }
      /* Loop until ']' reached. This "while" is the end of the "do" far above.
      If we are at the end of an internal nested string, revert to the outer
      string. */
-     while (((c = *(++ptr)) != 0 ||
+     while (((c = *(++ptr)) != CHAR_NULL ||
             (nestptr != NULL &&
-              (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
+              (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != CHAR_NULL)) &&
             (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
      /* Check for missing terminating ']' */
-     if (c == 0)
+     if (c == CHAR_NULL)
        {
        *errorcodeptr = ERR6;
        goto FAILED;
        }
-     /* If class_charcount is 1, we saw precisely one character whose value is
+     /* We will need an XCLASS if data has been placed in class_uchardata. In
-     less than 256. As long as there were no characters >= 128 and there was no
+     the second phase this is a sufficient test. However, in the pre-compile
-     use of \p or \P, in other words, no use of any XCLASS features, we can
+     phase, class_uchardata gets emptied to prevent workspace overflow, so it
-     optimize.
+     only if the very last character in the class needs XCLASS will it contain
+     anything at this point. For this reason, xclass gets set TRUE above when
-     In UTF-8 mode, we can optimize the negative case only if there were no
+     uchar_classdata is emptied, and that's why this code is the way it is here
-     characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
+     instead of just doing a test on class_uchardata below. */
-     operate on single-bytes characters only. This is an historical hangover.
-     Maybe one day we can tidy these opcodes to handle multi-byte characters.
+ #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
+     if (class_uchardata > class_uchardata_base) xclass = TRUE;
-     The optimization throws away the bit map. We turn the item into a
+ #endif
--character OP_CHAR[I] if it's positive, or OP_NOT[I] if it's negative.
-     Note that OP_NOT[I] does not support multibyte characters. In the positive
+     /* If this is the first thing in the branch, there can be no first char
-     case, it can cause firstbyte to be set. Otherwise, there can be no first
+     setting, whatever the repeat count. Any reqchar setting must remain
-     char if this item is first, whatever repeat count may follow. In the case
+     unchanged after any kind of repeat. */
-     of reqbyte, save the previous value for reinstating. */
+     if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
- #ifdef SUPPORT_UTF8
+     zerofirstchar = firstchar;
-     if (class_charcount == 1 && !class_utf8 &&
+     zerofirstcharflags = firstcharflags;
-       (!utf8 || !negate_class || class_lastchar < 128))
+     zeroreqchar = reqchar;
- #else
+     zeroreqcharflags = reqcharflags;
-     if (class_charcount == 1)
- #endif
-       {
-       zeroreqbyte = reqbyte;
-       /* The OP_NOT[I] opcodes work on one-byte characters only. */
-       if (negate_class)
-         {
-         if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
-         zerofirstbyte = firstbyte;
-         *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
-         *code++ = class_lastchar;
-         break;
-         }
-       /* For a single, positive character, get the value into mcbuffer, and
-       then we can handle this with the normal one-character code. */
- #ifdef SUPPORT_UTF8
-       if (utf8 && class_lastchar > 127)
-         mclength = _pcre_ord2utf8(class_lastchar, mcbuffer);
-       else
- #endif
-         {
-         mcbuffer[0] = class_lastchar;
-         mclength = 1;
-         }
-       goto ONE_CHAR;
-       }       /* End of 1-char optimization */
-     /* The general case - not the one-char optimization. If this is the first
-     thing in the branch, there can be no first char setting, whatever the
-     repeat count. Any reqbyte setting must remain unchanged after any kind of
-     repeat. */
-     if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
-     zerofirstbyte = firstbyte;
-     zeroreqbyte = reqbyte;
      /* If there are characters with values > 255, we have to compile an
      extended class, with its own opco