/[pcre]/code/tags/pcre-8.33/pcre_compile.c

Diff of /code/tags/pcre-8.33/pcre_compile.c

Parent Directory | Revision Log | View Patch Patch

-revision 835 by ph10,
Wed Dec 28 16:10:09 2011 UTC
+revision 836 by ph10,
Wed Dec 28 17:16:11 2011 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)_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
-Line 88 
 so this number is very generous.
+Line 92 
 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. */
+ 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 (4096)
+ #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
+ #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
  /* The overrun tests check for a slightly smaller size so that they detect the
  overrun before it actually does run off the end of the data block. */
- #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
+ #define WORK_SIZE_SAFETY_MARGIN (100)
+ /* Private flags added to firstchar and reqchar. */
+ #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */
+ #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */
+ /* 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 231 
 static const char posix_names[] =
+Line 250 
 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 266 
 substitutes must be in the order of the
+Line 285 
 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 412 
 static const char error_texts[] =
+Line 485 
 static const char error_texts[] =
    "\\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 UTF-8/16 code point (>= 0xd800 && <= 0xdfff)\0"
    ;
  /* Table to identify digits and hex digits. This is used when compiling
-Line 430 
 For convenience, we use the same bit def
+Line 506 
 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 474 
 static const unsigned char digitab[] =
+Line 556 
 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 509 
 static const unsigned char digitab[] =
+Line 591 
 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 548 
 static const unsigned char ebcdic_charta
+Line 630 
 static const unsigned char ebcdic_charta
  /* Definition to allow mutual recursion */
  static BOOL
-   compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,
+   compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
      int *, int *, branch_chain *, compile_data *, int *);
-Line 580 
 return s;
+Line 662 
 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
+ Returns:  0 if all went well, else an error number
+ */
+ static int
+ expand_workspace(compile_data *cd)
+ {
+ pcre_uchar *newspace;
+ int newsize = cd->workspace_size * 2;
+ if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
+ if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
+     newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
+  return ERR72;
+ newspace = (PUBL(malloc))(IN_UCHARS(newsize));
+ if (newspace == NULL) return ERR21;
+ memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
+ cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
+ if (cd->workspace_size > COMPILE_WORK_SIZE)
+   (PUBL(free))((void *)cd->start_workspace);
+ cd->start_workspace = newspace;
+ cd->workspace_size = newsize;
+ return 0;
+ }
+ /*************************************************
  *            Check for counted repeat            *
  *************************************************/
-Line 595 
 Returns:    TRUE or FALSE
+Line 714 
 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 637 
 Returns:         zero or positive => a d
+Line 758 
 Returns:         zero or positive => a d
  */
  static int
- check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,
+ check_escape(const pcre_uchar **ptrptr, int *errorcodeptr, 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_int32 c;
+ int i;
  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
  ptr--;                            /* Set pointer back to the last byte */
-Line 656 
 in a table. A non-zero result is somethi
+Line 779 
 in a table. A non-zero result is somethi
  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 (c < CHAR_0 || c > CHAR_z) {}
  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
  #else           /* EBCDIC coding */
- else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
+ /* Not alphanumeric */
+ else if (c < 'a' || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
  else if ((i = escapes[c - 0x48]) != 0)  c = i;
  #endif
-Line 668 
 else if ((i = escapes[c - 0x48]) != 0)
+Line 793 
 else if ((i = escapes[c - 0x48]) != 0)
  else
    {
-   const uschar *oldptr;
+   const pcre_uchar *oldptr;
    BOOL braced, negated;
    switch (c)
-Line 686 
 else
+Line 811 
 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)
-Line 741 
 else
+Line 868 
 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++)
-         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)
          {
          c = -ESC_k;
-Line 761 
 else
+Line 888 
 else
        }
      else negated = FALSE;
+     /* The integer range is limited by the machine's int representation. */
      c = 0;
-     while ((digitab[ptr[1]] & ctype_digit) != 0)
+     while (IS_DIGIT(ptr[1]))
+       {
+       if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
+         {
+         c = -1;
+         break;
+         }
        c = c * 10 + *(++ptr) - CHAR_0;
+       }
-     if (c < 0)   /* Integer overflow */
+     if (((unsigned int)c) > INT_MAX) /* Integer overflow */
        {
+       while (IS_DIGIT(ptr[1]))
+         ptr++;
        *errorcodeptr = ERR61;
        break;
        }
-Line 814 
 else
+Line 950 
 else
      if (!isclass)
        {
        oldptr = ptr;
+       /* The integer range is limited by the machine's int representation. */
        c -= CHAR_0;
-       while ((digitab[ptr[1]] & ctype_digit) != 0)
+       while (IS_DIGIT(ptr[1]))
+         {
+         if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
+           {
+           c = -1;
+           break;
+           }
          c = c * 10 + *(++ptr) - CHAR_0;
-       if (c < 0)    /* Integer overflow */
+         }
+       if (((unsigned int)c) > INT_MAX) /* Integer overflow */
          {
+         while (IS_DIGIT(ptr[1]))
+           ptr++;
          *errorcodeptr = ERR61;
          break;
          }
-Line 851 
 else
+Line 997 
 else
      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;
+     if (!utf && c > 0xff) *errorcodeptr = ERR51;
      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)
-Line 883 
 else
+Line 1030 
 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)
+       while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)
          {
          register int cc = *pt++;
          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
-         count++;
  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
-Line 900 
 else
+Line 1045 
 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
+ #ifdef COMPILE_PCRE8
+         if (c > (utf ? 0x10ffff : 0xff)) { c = -1; break; }
+ #else
+ #ifdef COMPILE_PCRE16
+         if (c > (utf ? 0x10ffff : 0xffff)) { c = -1; break; }
+ #endif
+ #endif
+         }
+       if (c < 0)
+         {
+         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 916 
 else
+Line 1075 
 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 */
        cc = *(++ptr);                           /* ++ in initializers */
-Line 1014 
 Returns:         type value from ucp_typ
+Line 1173 
 Returns:         type value from ucp_typ
  */
  static int
- get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
+ get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
  {
  int c, i, bot, top;
- const uschar *ptr = *ptrptr;
+ const pcre_uchar *ptr = *ptrptr;
- char name[32];
+ pcre_uchar name[32];
  c = *(++ptr);
  if (c == 0) goto ERROR_RETURN;
-Line 1035 
 if (c == CHAR_LEFT_CURLY_BRACKET)
+Line 1194 
 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;
-Line 1059 
 else
+Line 1218 
 else
  /* Search for a recognized property name using binary chop */
  bot = 0;
- top = _pcre_utt_size;
+ top = PRIV(utt_size);
  while (bot < top)
    {
    i = (bot + top) >> 1;
-   c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
+   c = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
    if (c == 0)
      {
-     *dptr = _pcre_utt[i].value;
+     *dptr = PRIV(utt)[i].value;
-     return _pcre_utt[i].type;
+     return PRIV(utt)[i].type;
      }
    if (c > 0) bot = i + 1; else top = i;
    }
-Line 1106 
 Returns:         pointer to '}' on succe
+Line 1265 
 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 1115 
 int max = -1;
+Line 1274 
 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 + *p++ - CHAR_0;
  if (min < 0 || min > 65535)
    {
    *errorcodeptr = ERR5;
-Line 1130 
 if (*p == CHAR_RIGHT_CURLY_BRACKET) max
+Line 1289 
 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 + *p++ - CHAR_0;
      if (max < 0 || max > 65535)
        {
        *errorcodeptr = ERR5;
-Line 1185 
 Arguments:
+Line 1344 
 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 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 1262 
 if (ptr[0] == CHAR_LEFT_PARENTHESIS)
+Line 1421 
 if (ptr[0] == CHAR_LEFT_PARENTHESIS)
          ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
        {
        int term;
-       const uschar *thisname;
+       const pcre_uchar *thisname;
        *count += 1;
        if (name == NULL && *count == lorn) return *count;
        term = *ptr++;
-Line 1270 
 if (ptr[0] == CHAR_LEFT_PARENTHESIS)
+Line 1429 
 if (ptr[0] == CHAR_LEFT_PARENTHESIS)
        thisname = ptr;
        while (*ptr != term) ptr++;
        if (name != NULL && lorn == ptr - thisname &&
-           strncmp((const char *)name, (const char *)thisname, lorn) == 0)
+           STRNCMP_UC_UC(name, thisname, lorn) == 0)
          return *count;
        term++;
        }
-Line 1313 
 for (; ptr < cd->end_pattern; ptr++)
+Line 1472 
 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 1361 
 for (; ptr < cd->end_pattern; ptr++)
+Line 1520 
 for (; ptr < cd->end_pattern; ptr++)
        {
        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;
-Line 1373 
 for (; ptr < cd->end_pattern; ptr++)
+Line 1532 
 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;
      }
-Line 1419 
 Arguments:
+Line 1578 
 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 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 1439 
 matching closing parens. That is why we
+Line 1598 
 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;
    }
-Line 1466 
 Arguments:
+Line 1625 
 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 1478 
 for (;;)
+Line 1637 
 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 1492 
 for (;;)
+Line 1651 
 for (;;)
      case OP_RREF:
      case OP_NRREF:
      case OP_DEF:
-     code += _pcre_OP_lengths[*code];
+     code += PRIV(OP_lengths)[*code];
      break;
      default:
-Line 1522 
 and doing the check at the end; a flag s
+Line 1681 
 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 mode
    atend    TRUE if called when the pattern is complete
    cd       the "compile data" structure
-Line 1534 
 Returns:   the fixed length,
+Line 1693 
 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 1547 
 branch, check the length against that of
+Line 1706 
 branch, check the length against that of
  for (;;)
    {
    int d;
-   uschar *ce, *cs;
+   pcre_uchar *ce, *cs;
    register int op = *cc;
    switch (op)
      {
      /* We only need to continue for OP_CBRA (normal capturing bracket) and
-Line 1561 
 for (;;)
+Line 1721 
 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 1592 
 for (;;)
+Line 1752 
 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 1608 
 for (;;)
+Line 1768 
 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 1616 
 for (;;)
+Line 1777 
 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 1643 
 for (;;)
+Line 1804 
 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 1654 
 for (;;)
+Line 1815 
 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 1667 
 for (;;)
+Line 1828 
 for (;;)
      case OP_NOTEXACT:
      case OP_NOTEXACTI:
      branchlength += 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 += 2;
-     cc += 4;
+     cc += 1 + IMM2_SIZE + 1;
      break;
      /* Handle single-char matchers */
-Line 1702 
 for (;;)
+Line 1863 
 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 1710 
 for (;;)
+Line 1871 
 for (;;)
      /* Check a class for variable quantification */
- #ifdef SUPPORT_UTF8
+ #if defined SUPPORT_UTF || defined COMPILE_PCRE16
      case OP_XCLASS:
-     cc += GET(cc, 1) - 33;
+     cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];
      /* Fall through */
  #endif
      case OP_CLASS:
      case OP_NCLASS:
-     cc += 33;
+     cc += PRIV(OP_lengths)[OP_CLASS];
      switch (*cc)
        {
-Line 1732 
 for (;;)
+Line 1893 
 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);
-       cc += 5;
+       cc += 1 + 2 * IMM2_SIZE;
        break;
        default:
-Line 1849 
 length.
+Line 2010 
 length.
  Arguments:
    code        points to start of expression
-   utf8        TRUE in UTF-8 mode
+   utf         TRUE in UTF-8 / UTF-16 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 (;;)
    {
-Line 1874 
 for (;;)
+Line 2035 
 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 1884 
 for (;;)
+Line 2045 
 for (;;)
             c == OP_CBRAPOS || c == OP_SCBRAPOS)
      {
      int n = 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 1913 
 for (;;)
+Line 2074 
 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:
-Line 1929 
 for (;;)
+Line 2091 
 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
+ #ifdef SUPPORT_UTF
-     if (utf8) switch(c)
+     if (utf) switch(c)
        {
        case OP_CHAR:
        case OP_CHARI:
-Line 1966 
 for (;;)
+Line 2128 
 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 1987 
 instance of OP_RECURSE.
+Line 2149 
 instance of OP_RECURSE.
  Arguments:
    code        points to start of expression
-   utf8        TRUE in UTF-8 mode
+   utf         TRUE in UTF-8 / UTF-16 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 (;;)
    {
-Line 2032 
 for (;;)
+Line 2194 
 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:
-Line 2048 
 for (;;)
+Line 2211 
 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
+ #ifdef SUPPORT_UTF
-     if (utf8) switch(c)
+     if (utf) switch(c)
        {
        case OP_CHAR:
        case OP_CHARI:
-Line 2085 
 for (;;)
+Line 2248 
 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 2112 
 bracket whose current branch will alread
+Line 2275 
 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 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;
- 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 2150 
 for (code = first_significant_code(code
+Line 2313 
 for (code = first_significant_code(code
    if (c == OP_RECURSE)
      {
-     const uschar *scode;
+     const pcre_uchar *scode;
      BOOL empty_branch;
      /* Test for forward reference */
-Line 2168 
 for (code = first_significant_code(code
+Line 2331 
 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 2186 
 for (code = first_significant_code(code
+Line 2349 
 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 2224 
 for (code = first_significant_code(code
+Line 2387 
 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 2242 
 for (code = first_significant_code(code
+Line 2405 
 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 2254 
 for (code = first_significant_code(code
+Line 2417 
 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 2329 
 for (code = first_significant_code(code
+Line 2492 
 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 2344 
 for (code = first_significant_code(code
+Line 2508 
 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
+ #ifdef SUPPORT_UTF
      case OP_STAR:
      case OP_STARI:
      case OP_MINSTAR:
-Line 2357 
 for (code = first_significant_code(code
+Line 2521 
 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 2366 
 for (code = first_significant_code(code
+Line 2530 
 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 2410 
 Arguments:
+Line 2574 
 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 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 2474 
 Returns:   TRUE or FALSE
+Line 2638 
 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 */
  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
-Line 2518 
 Returns:     a value representing the na
+Line 2682 
 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, len) == 0) return yield;
    pn += posix_name_lengths[yield] + 1;
    yield++;
    }
-Line 2557 
 value in the reference (which is a group
+Line 2721 
 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 mode
    cd         contains pointers to tables etc.
    save_hwm   the hwm forward reference pointer at the start of the group
-Line 2565 
 Returns:     nothing
+Line 2729 
 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. */
-Line 2618 
 Arguments:
+Line 2782 
 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 2647 
 Returns:             nothing
+Line 2811 
 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 2730 
 switch(ptype)
+Line 2894 
 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 2741 
 switch(ptype)
+Line 2905 
 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;
    }
  return FALSE;
-Line 2776 
 sense to automatically possessify the re
+Line 2940 
 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 mode
    ptr           next character in pattern
    options       options bits
    cd            contains pointers to tables etc.
-Line 2785 
 Returns:        TRUE if possessifying is
+Line 2949 
 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_int32 c, next;
  int op_code = *previous++;
  /* Skip whitespace and comments in extended mode */
-Line 2797 
 if ((options & PCRE_EXTENDED) != 0)
+Line 2961 
 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++;
-Line 2805 
 if ((options & PCRE_EXTENDED) != 0)
+Line 2969 
 if ((options & PCRE_EXTENDED) != 0)
          {
          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 2824 
 if (*ptr == CHAR_BACKSLASH)
+Line 2988 
 if (*ptr == CHAR_BACKSLASH)
    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
+ #ifdef SUPPORT_UTF
-   if (utf8) { GETCHARINC(next, ptr); } else
+   if (utf) { GETCHARINC(next, ptr); } else
  #endif
    next = *ptr++;
    }
  else return FALSE;
  /* Skip whitespace and comments in extended mode */
-Line 2841 
 if ((options & PCRE_EXTENDED) != 0)
+Line 3003 
 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++;
-Line 2849 
 if ((options & PCRE_EXTENDED) != 0)
+Line 3011 
 if ((options & PCRE_EXTENDED) != 0)
          {
          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 2861 
 if ((options & PCRE_EXTENDED) != 0)
+Line 3023 
 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
-Line 2870 
 the next item is a character. */
+Line 3032 
 the next item is a character. */
  if (next >= 0) switch(op_code)
    {
    case OP_CHAR:
- #ifdef SUPPORT_UTF8
+ #ifdef SUPPORT_UTF
    GETCHARTEST(c, previous);
  #else
    c = *previous;
-Line 2882 
 if (next >= 0) switch(op_code)
+Line 3044 
 if (next >= 0) switch(op_code)
    high-valued characters. */
    case OP_CHARI:
- #ifdef SUPPORT_UTF8
+ #ifdef SUPPORT_UTF
    GETCHARTEST(c, previous);
  #else
    c = *previous;
  #endif
    if (c == next) return FALSE;
- #ifdef SUPPORT_UTF8
+ #ifdef SUPPORT_UTF
-   if (utf8)
+   if (utf)
      {
      unsigned int othercase;
      if (next < 128) othercase = cd->fcc[next]; else
-Line 2901 
 if (next >= 0) switch(op_code)
+Line 3063 
 if (next >= 0) switch(op_code)
      return (unsigned int)c != othercase;
      }
    else
- #endif  /* SUPPORT_UTF8 */
+ #endif  /* SUPPORT_UTF */
-   return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
+   return (c != TABLE_GET(next, cd->fcc, next));  /* Non-UTF-8 mode */
    /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These
    opcodes are not used for multi-byte characters, because they are coded using
-Line 2913 
 if (next >= 0) switch(op_code)
+Line 3075 
 if (next >= 0) switch(op_code)
    case OP_NOTI:
    if ((c = *previous) == next) return TRUE;
- #ifdef SUPPORT_UTF8
+ #ifdef SUPPORT_UTF
-   if (utf8)
+   if (utf)
      {
      unsigned int othercase;
      if (next < 128) othercase = cd->fcc[next]; else
-Line 2926 
 if (next >= 0) switch(op_code)
+Line 3088 
 if (next >= 0) switch(op_code)
      return (unsigned int)c == othercase;
      }
    else
- #endif  /* SUPPORT_UTF8 */
+ #endif  /* SUPPORT_UTF */
-   return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
+   return (c == TABLE_GET(next, cd->fcc, next));  /* Non-UTF-8 mode */
    /* 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. */
-Line 3018 
 switch(op_code)
+Line 3180 
 switch(op_code)
    {
    case OP_CHAR:
    case OP_CHARI:
- #ifdef SUPPORT_UTF8
+ #ifdef SUPPORT_UTF
    GETCHARTEST(c, previous);
  #else
    c = *previous;
-Line 3123 
 switch(op_code)
+Line 3285 
 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. */
-Line 3201 
 Arguments:
+Line 3363 
 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   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
-   reqbyteptr     set to the last literal character required, else < 0
+   reqcharptr     set to the last literal character required, else < 0
    bcptr          points to current branch chain
    cond_depth     conditional nesting depth
    cd             contains pointers to tables etc.
-Line 3214 
 Returns:         TRUE on success
+Line 3376 
 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, pcre_int32 *firstcharptr,
-   int cond_depth, compile_data *cd, int *lengthptr)
+   pcre_int32 *reqcharptr, 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_int32 firstchar, reqchar;
- int zeroreqbyte, zerofirstbyte;
+ pcre_int32 zeroreqchar, zerofirstchar;
- int req_caseopt, reqvary, tempreqvary;
+ 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 uschar *code = *codeptr;
+ register pcre_uchar *code = *codeptr;
- uschar *last_code = code;
+ pcre_uchar *last_code = code;
- uschar *orig_code = code;
+ pcre_uchar *orig_code = code;
- uschar *tempcode;
+ 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_UTF16 has the same value as PCRE_UTF8. */
- BOOL utf8 = (options & PCRE_UTF8) != 0;
+ BOOL utf = (options & PCRE_UTF8) != 0;
- uschar *class_utf8data;
+ pcre_uchar utf_chars[6];
- uschar *class_utf8data_base;
- uschar utf8_char[6];
  #else
- BOOL utf8 = FALSE;
+ BOOL utf = FALSE;
+ #endif
+ /* Helper variables for OP_XCLASS opcode (for characters > 255). */
+ #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
+ BOOL xclass;
+ pcre_uchar *class_uchardata;
+ pcre_uchar *class_uchardata_base;
  #endif
  #ifdef PCRE_DEBUG
-Line 3268 
 greedy_non_default = greedy_default ^ 1;
+Line 3437 
 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 = 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 3295 
 for (;; ptr++)
+Line 3465 
 for (;; ptr++)
    BOOL is_quantifier;
    BOOL is_recurse;
    BOOL reset_bracount;
-   int class_charcount;
+   int class_has_8bitchar;
-   int class_lastchar;
+   int class_single_char;
    int newoptions;
    int recno;
    int refsign;
    int skipbytes;
-   int subreqbyte;
+   int subreqchar;
-   int subfirstbyte;
+   int subfirstchar;
    int terminator;
    int mclength;
    int tempbracount;
-   uschar mcbuffer[8];
+   pcre_uchar mcbuffer[8];
-   /* Get next byte in the pattern */
+   /* Get next character in the pattern */
    c = *ptr;
-Line 3330 
 for (;; ptr++)
+Line 3500 
 for (;; ptr++)
  #ifdef PCRE_DEBUG
      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
  #endif
-     if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
+     if (code > cd->start_workspace + cd->workspace_size -
+         WORK_SIZE_SAFETY_MARGIN)                       /* Check for overrun */
        {
        *errorcodeptr = ERR52;
        goto FAILED;
-Line 3353 
 for (;; ptr++)
+Line 3524 
 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,
      if "previous" is NULL, reset the current code pointer to the start. */
-Line 3364 
 for (;; ptr++)
+Line 3535 
 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 3380 
 for (;; ptr++)
+Line 3551 
 for (;; ptr++)
    /* In the real compile phase, just check the workspace used by the forward
    reference list. */
-   else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
+   else if (cd->hwm > cd->start_workspace + cd->workspace_size -
+            WORK_SIZE_SAFETY_MARGIN)
      {
      *errorcodeptr = ERR52;
      goto FAILED;
-Line 3432 
 for (;; ptr++)
+Line 3604 
 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++;
-Line 3440 
 for (;; ptr++)
+Line 3612 
 for (;; ptr++)
          {
          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;
-Line 3465 
 for (;; ptr++)
+Line 3637 
 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;
+     *reqcharptr = reqchar;
      *codeptr = code;
      *ptrptr = ptr;
      if (lengthptr != NULL)
-Line 3490 
 for (;; ptr++)
+Line 3662 
 for (;; ptr++)
      previous = NULL;
      if ((options & PCRE_MULTILINE) != 0)
        {
-       if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
+       if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
        *code++ = OP_CIRCM;
        }
      else *code++ = OP_CIRC;
-Line 3502 
 for (;; ptr++)
+Line 3674 
 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 (firstchar == REQ_UNSET) firstchar = REQ_NONE;
-     zerofirstbyte = firstbyte;
+     zerofirstchar = firstchar;
-     zeroreqbyte = reqbyte;
+     zeroreqchar = reqchar;
      previous = code;
      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
      break;
-Line 3562 
 for (;; ptr++)
+Line 3734 
 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 3582 
 for (;; ptr++)
+Line 3753 
 for (;; ptr++)
          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
        {
        *code++ = negate_class? OP_ALLANY : OP_FAIL;
-       if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
+       if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
-       zerofirstbyte = firstbyte;
+       zerofirstchar = firstchar;
        break;
        }
-Line 3593 
 for (;; ptr++)
+Line 3764 
 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. */
+     < 256 character. class_single_char will be 1 if the class contains only
+     a single character. */
-     class_charcount = 0;
+     class_has_8bitchar = 0;
-     class_lastchar = -1;
+     class_single_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
      than 256), 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;                           /* No chars >= 256 */
-     class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
+     class_uchardata = code + LINK_SIZE + 2;   /* For UTF-8 items */
-     class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
+     class_uchardata_base = class_uchardata;   /* For resetting in pass 1 */
  #endif
      /* Process characters until ] is reached. By writing this as a "do" it
-Line 3619 
 for (;; ptr++)
+Line 3791 
 for (;; ptr++)
      if (c != 0) 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). */
        if (lengthptr != NULL)
          {
-         *lengthptr += class_utf8data - class_utf8data_base;
+         *lengthptr += class_uchardata - class_uchardata_base;
-         class_utf8data = class_utf8data_base;
+         class_uchardata = class_uchardata_base;
          }
  #endif
        /* Inside \Q...\E everything is literal except \E */
-Line 3665 
 for (;; ptr++)
+Line 3838 
 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 3721 
 for (;; ptr++)
+Line 3894 
 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 3752 
 for (;; ptr++)
+Line 3925 
 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 characters. */
+         class_has_8bitchar = 1;
+         /* Every class contains at least two characters. */
+         class_single_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_single_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)
-Line 3770 
 for (;; ptr++)
+Line 3947 
 for (;; ptr++)
          if (*errorcodeptr != 0) goto FAILED;
          if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */
+         else if (-c == ESC_N)            /* \N is not supported in a class */
+           {
+           *errorcodeptr = ERR71;
+           goto FAILED;
+           }
          else if (-c == ESC_Q)            /* Handle start of quoted string */
            {
            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
-Line 3783 
 for (;; ptr++)
+Line 3965 
 for (;; ptr++)
          if (c < 0)
            {
-           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_single_char += 2;
            switch (-c)
              {
-Line 3797 
 for (;; ptr++)
+Line 3982 
 for (;; ptr++)
              case ESC_SU:
              nestptr = ptr;
              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
-             class_charcount -= 2;                /* Undo! */
+             class_has_8bitchar--;                /* Undo! */
              continue;
  #endif
              case ESC_d:
-Line 3838 
 for (;; ptr++)
+Line 4023 
 for (;; ptr++)
              SETBIT(classbits, 0x09); /* VT */
              SETBIT(classbits, 0x20); /* SPACE */
              SETBIT(classbits, 0xa0); /* NSBP */
- #ifdef SUPPORT_UTF8
+ #ifndef COMPILE_PCRE8
-             if (utf8)
+             xclass = TRUE;
+             *class_uchardata++ = XCL_SINGLE;
+             *class_uchardata++ = 0x1680;
+             *class_uchardata++ = XCL_SINGLE;
+             *class_uchardata++ = 0x180e;
+             *class_uchardata++ = XCL_RANGE;
+             *class_uchardata++ = 0x2000;
+             *class_uchardata++ = 0x200a;
+             *class_uchardata++ = XCL_SINGLE;
+             *class_uchardata++ = 0x202f;
+             *class_uchardata++ = XCL_SINGLE;
+             *class_uchardata++ = 0x205f;
+             *class_uchardata++ = XCL_SINGLE;
+             *class_uchardata++ = 0x3000;
+ #elif defined SUPPORT_UTF
+             if (utf)
                {
-               class_utf8 = TRUE;
+               xclass = TRUE;
-               *class_utf8data++ = XCL_SINGLE;
+               *class_uchardata++ = XCL_SINGLE;
-               class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(0x1680, class_uchardata);
-               *class_utf8data++ = XCL_SINGLE;
+               *class_uchardata++ = XCL_SINGLE;
-               class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(0x180e, class_uchardata);
-               *class_utf8data++ = XCL_RANGE;
+               *class_uchardata++ = XCL_RANGE;
-               class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(0x2000, class_uchardata);
-               class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(0x200a, class_uchardata);
-               *class_utf8data++ = XCL_SINGLE;
+               *class_uchardata++ = XCL_SINGLE;
-               class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(0x202f, class_uchardata);
-               *class_utf8data++ = XCL_SINGLE;
+               *class_uchardata++ = XCL_SINGLE;
-               class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(0x205f, class_uchardata);
-               *class_utf8data++ = XCL_SINGLE;
+               *class_uchardata++ = XCL_SINGLE;
-               class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(0x3000, class_uchardata);
                }
  #endif
              continue;
-Line 3872 
 for (;; ptr++)
+Line 4072 
 for (;; ptr++)
                  }
                classbits[c] |= x;
                }
+ #ifndef COMPILE_PCRE8
- #ifdef SUPPORT_UTF8
+             xclass = TRUE;
-             if (utf8)
+             *class_uchardata++ = XCL_RANGE;
+             *class_uchardata++ = 0x0100;
+             *class_uchardata++ = 0x167f;
+             *class_uchardata++ = XCL_RANGE;
+             *class_uchardata++ = 0x1681;
+             *class_uchardata++ = 0x180d;
+             *class_uchardata++ = XCL_RANGE;
+             *class_uchardata++ = 0x180f;
+             *class_uchardata++ = 0x1fff;
+             *class_uchardata++ = XCL_RANGE;
+             *class_uchardata++ = 0x200b;
+             *class_uchardata++ = 0x202e;
+             *class_uchardata++ = XCL_RANGE;
+             *class_uchardata++ = 0x2030;
+             *class_uchardata++ = 0x205e;
+             *class_uchardata++ = XCL_RANGE;
+             *class_uchardata++ = 0x2060;
+             *class_uchardata++ = 0x2fff;
+             *class_uchardata++ = XCL_RANGE;
+             *class_uchardata++ = 0x3001;
+ #ifdef SUPPORT_UTF
+             if (utf)
+               class_uchardata += PRIV(ord2utf)(0x10ffff, class_uchardata);
+             else
+ #endif
+               *class_uchardata++ = 0xffff;
+ #elif defined SUPPORT_UTF
+             if (utf)
                {
-               class_utf8 = TRUE;
+               xclass = TRUE;
-               *class_utf8data++ = XCL_RANGE;
+               *class_uchardata++ = XCL_RANGE;
-               class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(0x0100, class_uchardata);
-               class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(0x167f, class_uchardata);
-               *class_utf8data++ = XCL_RANGE;
+               *class_uchardata++ = XCL_RANGE;
-               class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(0x1681, class_uchardata);
-               class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(0x180d, class_uchardata);
-               *class_utf8data++ = XCL_RANGE;
+               *class_uchardata++ = XCL_RANGE;
-               class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(0x180f, class_uchardata);
-               class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(0x1fff, class_uchardata);
-               *class_utf8data++ = XCL_RANGE;
+               *class_uchardata++ = XCL_RANGE;
-               class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(0x200b, class_uchardata);
-               class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(0x202e, class_uchardata);
-               *class_utf8data++ = XCL_RANGE;
+               *class_uchardata++ = XCL_RANGE;
-               class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(0x2030, class_uchardata);
-               class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(0x205e, class_uchardata);
-               *class_utf8data++ = XCL_RANGE;
+               *class_uchardata++ = XCL_RANGE;
-               class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(0x2060, class_uchardata);
-               class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(0x2fff, class_uchardata);
-               *class_utf8data++ = XCL_RANGE;
+               *class_uchardata++ = XCL_RANGE;
-               class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(0x3001, class_uchardata);
-               class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(0x10ffff, class_uchardata);
                }
  #endif
              continue;
-Line 3908 
 for (;; ptr++)
+Line 4135 
 for (;; ptr++)
              SETBIT(classbits, 0x0c); /* FF */
              SETBIT(classbits, 0x0d); /* CR */
              SETBIT(classbits, 0x85); /* NEL */
- #ifdef SUPPORT_UTF8
+ #ifndef COMPILE_PCRE8
-             if (utf8)
+             xclass = TRUE;
+             *class_uchardata++ = XCL_RANGE;
+             *class_uchardata++ = 0x2028;
+             *class_uchardata++ = 0x2029;
+ #elif defined SUPPORT_UTF
+             if (utf)
                {
-               class_utf8 = TRUE;
+               xclass = TRUE;
-               *class_utf8data++ = XCL_RANGE;
+               *class_uchardata++ = XCL_RANGE;
-               class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(0x2028, class_uchardata);
-               class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(0x2029, class_uchardata);
                }
  #endif
              continue;
-Line 3936 
 for (;; ptr++)
+Line 4168 
 for (;; ptr++)
                classbits[c] |= x;
                }
- #ifdef SUPPORT_UTF8
+ #ifndef COMPILE_PCRE8
-             if (utf8)
+             xclass = TRUE;
+             *class_uchardata++ = XCL_RANGE;
+             *class_uchardata++ = 0x0100;
+             *class_uchardata++ = 0x2027;
+             *class_uchardata++ = XCL_RANGE;
+             *class_uchardata++ = 0x202a;
+ #ifdef SUPPORT_UTF
+             if (utf)
+               class_uchardata += PRIV(ord2utf)(0x10ffff, class_uchardata);
+             else
+ #endif
+               *class_uchardata++ = 0xffff;
+ #elif defined SUPPORT_UTF
+             if (utf)
                {
-               class_utf8 = TRUE;
+               xclass = TRUE;
-               *class_utf8data++ = XCL_RANGE;
+               *class_uchardata++ = XCL_RANGE;
-               class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(0x0100, class_uchardata);
-               class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(0x2027, class_uchardata);
-               *class_utf8data++ = XCL_RANGE;
+               *class_uchardata++ = XCL_RANGE;
-               class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(0x202a, class_uchardata);
-               class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(0x10ffff, class_uchardata);
                }
  #endif
              continue;
-Line 3958 
 for (;; ptr++)
+Line 4203 
 for (;; ptr++)
                int pdata;
                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
                if (ptype < 0) goto FAILED;
-               class_utf8 = TRUE;
+               xclass = TRUE;
-               *class_utf8data++ = ((-c == ESC_p) != negated)?
+               *class_uchardata++ = ((-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 3977 
 for (;; ptr++)
+Line 4222 
 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_single_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
-         greater than 256 in UTF-8 mode. */
+         greater than 256. */
          }   /* End of backslash handling */
-Line 4032 
 for (;; ptr++)
+Line 4278 
 for (;; ptr++)
            goto LONE_SINGLE_CHARACTER;
            }
- #ifdef SUPPORT_UTF8
+ #ifdef SUPPORT_UTF
-         if (utf8)
+         if (utf)
            {                           /* Braces are required because the */
            GETCHARLEN(d, ptr, ptr);    /* macro generates multiple statements */
            }
-Line 4077 
 for (;; ptr++)
+Line 4323 
 for (;; ptr++)
          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
+         /* Since we found a character range, single character optimizations
+         cannot be done anymore. */
+         class_single_char = 2;
          /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless
          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 defined SUPPORT_UTF && !(defined COMPILE_PCRE8)
-         if (utf8 && (d > 255 || ((options & PCRE_CASELESS) != 0 && d > 127)))
+         if ((d > 255) || (utf && ((options & PCRE_CASELESS) != 0 && d > 127)))
+ #elif defined  SUPPORT_UTF
+         if (utf && (d > 255 || ((options & PCRE_CASELESS) != 0 && d > 127)))
+ #elif !(defined COMPILE_PCRE8)
+         if (d > 255)
+ #endif
+ #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
            {
-           class_utf8 = TRUE;
+           xclass = 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
+ #ifndef COMPILE_PCRE8
+           if (utf && (options & PCRE_CASELESS) != 0)
+ #else
            if ((options & PCRE_CASELESS) != 0)
+ #endif
              {
              unsigned int occ, ocd;
              unsigned int cc = c;
-Line 4118 
 for (;; ptr++)
+Line 4378 
 for (;; ptr++)
                if (occ == ocd)
                  {
-                 *class_utf8data++ = XCL_SINGLE;
+                 *class_uchardata++ = XCL_SINGLE;
                  }
                else
                  {
-                 *class_utf8data++ = XCL_RANGE;
+                 *class_uchardata++ = XCL_RANGE;
-                 class_utf8data += _pcre_ord2utf8(occ, class_utf8data);
+                 class_uchardata += PRIV(ord2utf)(occ, class_uchardata);
                  }
-               class_utf8data += _pcre_ord2utf8(ocd, class_utf8data);
+               class_uchardata += PRIV(ord2utf)(ocd, class_uchardata);
                }
              }
  #endif  /* SUPPORT_UCP */
-Line 4133 
 for (;; ptr++)
+Line 4393 
 for (;; ptr++)
            /* Now record the original range, possibly modified for UCP caseless
            overlapping ranges. */
-           *class_utf8data++ = XCL_RANGE;
+           *class_uchardata++ = XCL_RANGE;
-           class_utf8data += _pcre_ord2utf8(c, class_utf8data);
+ #ifdef SUPPORT_UTF
-           class_utf8data += _pcre_ord2utf8(d, class_utf8data);
+ #ifndef COMPILE_PCRE8
+           if (utf)
+             {
+             class_uchardata += PRIV(ord2utf)(c, class_uchardata);
+             class_uchardata += PRIV(ord2utf)(d, class_uchardata);
+             }
+           else
+             {
+             *class_uchardata++ = c;
+             *class_uchardata++ = d;
+             }
+ #else
+           class_uchardata += PRIV(ord2utf)(c, class_uchardata);
+           class_uchardata += PRIV(ord2utf)(d, class_uchardata);
+ #endif
+ #else /* SUPPORT_UTF */
+           *class_uchardata++ = c;
+           *class_uchardata++ = d;
+ #endif /* SUPPORT_UTF */
            /* With UCP support, we are done. Without UCP support, there is no
-           caseless matching for UTF-8 characters > 127; we can use the bit map
+           caseless matching for UTF characters > 127; we can use the bit map
-           for the smaller ones. */
+           for the smaller ones. As for 16 bit characters without UTF, we
+           can still use  */
  #ifdef SUPPORT_UCP
-           continue;    /* With next character in the class */
+ #ifndef COMPILE_PCRE8
- #else
+           if (utf)
-           if ((options & PCRE_CASELESS) == 0 || c > 127) continue;
+ #endif
+             continue;    /* With next character in the class */
+ #endif  /* SUPPORT_UCP */
+ #if defined SUPPORT_UTF && !defined(SUPPORT_UCP) && !(defined COMPILE_PCRE8)
+           if (utf)
+             {
+             if ((options & PCRE_CASELESS) == 0 || c > 127) continue;
+             /* Adjust upper limit and fall through to set up the map */
+             d = 127;
+             }
+           else
+             {
+             if (c > 255) continue;
+             /* Adjust upper limit and fall through to set up the map */
+             d = 255;
+             }
+ #elif defined SUPPORT_UTF && !defined(SUPPORT_UCP)
+           if ((options & PCRE_CASELESS) == 0 || c > 127) continue;
            /* Adjust upper limit and fall through to set up the map */
            d = 127;
+ #else
- #endif  /* SUPPORT_UCP */
+           if (c > 255) continue;
+           /* Adjust upper limit and fall through to set up the map */
+           d = 255;
+ #endif  /* SUPPORT_UTF && !SUPPORT_UCP && !COMPILE_PCRE8 */
            }
- #endif  /* SUPPORT_UTF8 */
+ #endif  /* SUPPORT_UTF || !COMPILE_PCRE8 */
-         /* We use the bit map for all cases when not in UTF-8 mode; else
+         /* We use the bit map for 8 bit mode, or when the characters fall
-         ranges that lie entirely within 0-127 when there is UCP support; else
+         partially or entirely to [0-255] ([0-127] for UCP) ranges. */
-         for partial ranges without UCP support. */
-         class_charcount += d - c + 1;
+         class_has_8bitchar = 1;
-         class_lastchar = d;
          /* We can save a bit of time by skipping this in the pre-compile. */
-Line 4168 
 for (;; ptr++)
+Line 4464 
 for (;; ptr++)
            classbits[c/8] |= (1 << (c&7));
            if ((options & PCRE_CASELESS) != 0)
              {
-             int uc = cd->fcc[c];           /* flip case */
+             int uc = cd->fcc[c]; /* flip case */
              classbits[uc/8] |= (1 << (uc&7));
              }
            }
-Line 4182 
 for (;; ptr++)
+Line 4478 
 for (;; ptr++)
        LONE_SINGLE_CHARACTER:
-       /* Handle a character that cannot go in the bit map */
+       /* Only the value of 1 matters for class_single_char. */
+       if (class_single_char < 2) class_single_char++;
- #ifdef SUPPORT_UTF8
+       /* If class_charcount is 1, we saw precisely one character. As long as
-       if (utf8 && (c > 255 || ((options & PCRE_CASELESS) != 0 && c > 127)))
+       there were no negated characters >= 128 and there was no use of \p or \P,
+       in other words, no use of any XCLASS features, we can optimize.
+       In UTF-8 mode, we can optimize the negative case only if there were no
+       characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
+       operate on single-bytes characters only. This is an historical hangover.
+       Maybe one day we can tidy these opcodes to handle multi-byte characters.
+       The optimization throws away the bit map. We turn the item into a
+-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
+       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. */
+ #ifdef SUPPORT_UTF
+       if (class_single_char == 1 && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET
+         && (!utf || !negate_class || c < (MAX_VALUE_FOR_SINGLE_CHAR + 1)))
+ #else
+       if (class_single_char == 1 && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
+ #endif
          {
-         class_utf8 = TRUE;
+         ptr++;
-         *class_utf8data++ = XCL_SINGLE;
+         zeroreqchar = reqchar;
-         class_utf8data += _pcre_ord2utf8(c, class_utf8data);
+         /* The OP_NOT[I] opcodes work on single characters only. */
+         if (negate_class)
+           {
+           if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
+           zerofirstchar = firstchar;
+           *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
+           *code++ = c;
+           goto NOT_CHAR;
+           }
+         /* For a single, positive character, get the value into mcbuffer, and
+         then we can handle this with the normal one-character code. */
+ #ifdef SUPPORT_UTF
+         if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
+           mclength = PRIV(ord2utf)(c, mcbuffer);
+         else
+ #endif
+           {
+           mcbuffer[0] = c;
+           mclength = 1;
+           }
+         goto ONE_CHAR;
+         }       /* End of 1-char optimization */
+       /* Handle a character that cannot go in the bit map. */
+ #if defined SUPPORT_UTF && !(defined COMPILE_PCRE8)
+       if ((c > 255) || (utf && ((options & PCRE_CASELESS) != 0 && c > 127)))
+ #elif defined SUPPORT_UTF
+       if (utf && (c > 255 || ((options & PCRE_CASELESS) != 0 && c > 127)))
+ #elif !(defined COMPILE_PCRE8)
+       if (c > 255)
+ #endif
+ #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
+         {
+         xclass = TRUE;
+         *class_uchardata++ = XCL_SINGLE;
+ #ifdef SUPPORT_UTF
+ #ifndef COMPILE_PCRE8
+         /* In non 8 bit mode, we can get here even if we are not in UTF mode. */
+         if (!utf)
+           *class_uchardata++ = c;
+         else
+ #endif
+           class_uchardata += PRIV(ord2utf)(c, class_uchardata);
+ #else /* SUPPORT_UTF */
+         *class_uchardata++ = c;
+ #endif /* SUPPORT_UTF */
  #ifdef SUPPORT_UCP
+ #ifdef COMPILE_PCRE8
          if ((options & PCRE_CASELESS) != 0)
+ #else
+         /* In non 8 bit mode, we can get here even if we are not in UTF mode. */
+         if (utf && (options & PCRE_CASELESS) != 0)
+ #endif
            {
            unsigned int othercase;
            if ((othercase = UCD_OTHERCASE(c)) != c)
              {
-             *class_utf8data++ = XCL_SINGLE;
+             *class_uchardata++ = XCL_SINGLE;
-             class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
+             class_uchardata += PRIV(ord2utf)(othercase, class_uchardata);
              }
            }
  #endif  /* SUPPORT_UCP */
          }
        else
- #endif  /* SUPPORT_UTF8 */
+ #endif  /* SUPPORT_UTF || COMPILE_PCRE16 */
        /* Handle a single-byte character */
          {
+         class_has_8bitchar = 1;
          classbits[c/8] |= (1 << (c&7));
          if ((options & PCRE_CASELESS) != 0)
            {
-           c = cd->fcc[c];   /* flip case */
+           c = cd->fcc[c]; /* flip case */
            classbits[c/8] |= (1 << (c&7));
            }
-         class_charcount++;
-         class_lastchar = c;
          }
        }
-Line 4237 
 for (;; ptr++)
+Line 4609 
 for (;; ptr++)
        goto FAILED;
        }
-     /* If class_charcount is 1, we saw precisely one character whose value is
+     /* If this is the first thing in the branch, there can be no first char
-     less than 256. As long as there were no characters >= 128 and there was no
+     setting, whatever the repeat count. Any reqchar setting must remain
-     use of \p or \P, in other words, no use of any XCLASS features, we can
+     unchanged after any kind of repeat. */
-     optimize.
+     if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
-     In UTF-8 mode, we can optimize the negative case only if there were no
+     zerofirstchar = firstchar;
-     characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
+     zeroreqchar = reqchar;
-     operate on single-bytes characters only. This is an historical hangover.
-     Maybe one day we can tidy these opcodes to handle multi-byte characters.
-     The optimization throws away the bit map. We turn the item into a
--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
-     case, it can cause firstbyte to be set. Otherwise, there can be no first
-     char if this item is first, whatever repeat count may follow. In the case
-     of reqbyte, save the previous value for reinstating. */
- #ifdef SUPPORT_UTF8
-     if (class_charcount == 1 && !class_utf8 &&
-       (!utf8 || !negate_class || class_lastchar < 128))
- #else
-     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 opcode, unless there was a negated special
-Line 4306 
 for (;; ptr++)
+Line 4625 
 for (;; ptr++)
      be listed) there are no characters < 256, we can omit the bitmap in the
      actual compiled code. */
- #ifdef SUPPORT_UTF8
+ #ifdef SUPPORT_UTF
-     if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
+     if (xclass && (!should_flip_negation || (options & PCRE_UCP) != 0))
+ #elif !defined COMPILE_PCRE8
+     if (xclass && !should_flip_negation)
+ #endif
+ #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
        {
-       *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
+       *class_uchardata++ = XCL_END;    /* Marks the end of extra data */
        *code++ = OP_XCLASS;
        code += LINK_SIZE;
-       *code = negate_class? XCL_NOT : 0;
+       *code = negate_class? XCL_NOT:0;
        /* If the map is required, move up the extra data to make room for it;
        otherwise just move the code pointer to the end of the extra data. */
-       if (class_charcount > 0)
+       if (class_has_8bitchar > 0)
          {
          *code++ |= XCL_MAP;
-         memmove(code + 32, code, class_utf8data - code);
+         memmove(code + (32 / sizeof(pcre_uchar)), code,
+           IN_UCHARS(class_uchardata - code));
          memcpy(code, classbits, 32);
-         code = class_utf8data + 32;
+         code = class_uchardata + (32 / sizeof(pcre_uchar));
          }
-       else code = class_utf8data;
+       else code = class_uchardata;
        /* Now fill in the complete length of the item */
-       PUT(previous, 1, code - previous);
+       PUT(previous, 1, (int)(code - previous));
        break;   /* End of class handling */
        }
  #endif
-Line 4340 
 for (;; ptr++)
+Line 4664 
 for (;; ptr++)
      negating it if necessary. */
      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
-     if (negate_class)
+     if (lengthptr == NULL)    /* Save time in the pre-compile phase */
-       {
-       if (lengthptr == NULL)    /* Save time in the pre-compile phase */
-         for (c = 0; c < 32; c++) code[c] = ~classbits[c];
-       }
-     else
        {
+       if (negate_class)
+         for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
        memcpy(code, classbits, 32);
        }
-     code += 32;
+     code += 32 / sizeof(pcre_uchar);
+     NOT_CHAR:
      break;
-Line 4386 
 for (;; ptr++)
+Line 4708 
 for (;; ptr++)
      if (repeat_min == 0)
        {
-       firstbyte = zerofirstbyte;    /* Adjust for zero repeat */
+       firstchar = zerofirstchar;    /* Adjust for zero repeat */
-       reqbyte = zeroreqbyte;        /* Ditto */
+       reqchar = zeroreqchar;        /* Ditto */
        }
      /* Remember whether this is a variable length repeat */
-Line 4426 
 for (;; ptr++)
+Line 4748 
 for (;; ptr++)
      past, but it no longer happens for non-repeated recursions. In fact, the
      repeated ones could be re-implemented independently so as not to need this,
      but for the moment we rely on the code for repeating groups. */
      if (*previous == OP_RECURSE)
        {
-       memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
+       memmove(previous + 1 + LINK_SIZE, previous, IN_UCHARS(1 + LINK_SIZE));
        *previous = OP_ONCE;
        PUT(previous, 1, 2 + 2*LINK_SIZE);
        previous[2 + 2*LINK_SIZE] = OP_KET;
-Line 4452 
 for (;; ptr++)
+Line 4774 
 for (;; ptr++)
      /* If previous was a character match, abolish the item and generate a
      repeat item instead. If a char item has a minumum of more than one, ensure
-     that it is set in reqbyte - it might not be if a sequence such as x{3} is
+     that it is set in reqchar - it might not be if a sequence such as x{3} is
-     the first thing in a branch because the x will have gone into firstbyte
+     the first thing in a branch because the x will have gone into firstchar
      instead.  */
      if (*previous == OP_CHAR || *previous == OP_CHARI)
        {
        op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
-       /* Deal with UTF-8 characters that take up more than one byte. It's
+       /* Deal with UTF characters that take up more than one character. It's
        easier to write this out separately than try to macrify it. Use c to
-       hold the length of the character in bytes, plus 0x80 to flag that it's a
+       hold the length of the character in bytes, plus UTF_LENGTH to flag that
-       length rather than a small character. */
+       it's a length rather than a small character. */
- #ifdef SUPPORT_UTF8
+ #ifdef SUPPORT_UTF
-       if (utf8 && (code[-1] & 0x80) != 0)
+       if (utf && NOT_FIRSTCHAR(code[-1]))
          {
-         uschar *lastchar = code - 1;
+         pcre_uchar *lastchar = code - 1;
-         while((*lastchar & 0xc0) == 0x80) lastchar--;
+         BACKCHAR(lastchar);
-         c = code - lastchar;            /* Length of UTF-8 character */
+         c = (int)(code - lastchar);     /* Length of UTF-8 character */
-         memcpy(utf8_char, lastchar, c); /* Save the char */
+         memcpy(utf_chars, lastchar, IN_UCHARS(c)); /* Save the char */
-         c |= 0x80;                      /* Flag c as a length */
+         c |= UTF_LENGTH;                /* Flag c as a length */
          }
        else
- #endif
+ #endif /* SUPPORT_UTF */
-       /* Handle the case of a single byte - either with no UTF8 support, or
-       with UTF-8 disabled, or for a UTF-8 character < 128. */
+       /* Handle the case of a single charater - either with no UTF support, or
+       with UTF disabled, or for a single character UTF character. */
          {
          c = code[-1];
-         if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;
+         if (repeat_min > 1) reqchar = c | req_caseopt | cd->req_varyopt;
          }
        /* If the repetition is unlimited, it pays to see if the next thing on
-Line 4492 
 for (;; ptr++)
+Line 4813 
 for (;; ptr++)
        if (!possessive_quantifier &&
            repeat_max < 0 &&
-           check_auto_possessive(previous, utf8, ptr + 1, options, cd))
+           check_auto_possessive(previous, utf, ptr + 1, options, cd))
          {
          repeat_type = 0;    /* Force greedy */
          possessive_quantifier = TRUE;
-Line 4513 
 for (;; ptr++)
+Line 4834 
 for (;; ptr++)
        c = previous[1];
        if (!possessive_quantifier &&
            repeat_max < 0 &&
-           check_auto_possessive(previous, utf8, ptr + 1, options, cd))
+           check_auto_possessive(previous, utf, ptr + 1, options, cd))
          {
          repeat_type = 0;    /* Force greedy */
          possessive_quantifier = TRUE;
-Line 4530 
 for (;; ptr++)
+Line 4851 
 for (;; ptr++)
      else if (*previous < OP_EODN)
        {
-       uschar *oldcode;
+       pcre_uchar *oldcode;
        int prop_type, prop_value;
        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
        c = *previous;
        if (!possessive_quantifier &&
            repeat_max < 0 &&
-           check_auto_possessive(previous, utf8, ptr + 1, options, cd))
+           check_auto_possessive(previous, utf, ptr + 1, options, cd))
          {
          repeat_type = 0;    /* Force greedy */
          possessive_quantifier = TRUE;
-Line 4617 
 for (;; ptr++)
+Line 4938 
 for (;; ptr++)
          we have to insert the character for the previous code. For a repeated
          Unicode property match, there are two extra bytes that define the
          required property. In UTF-8 mode, long characters have their length in
-         c, with the 0x80 bit as a flag. */
+         c, with the UTF_LENGTH bit as a flag. */
          if (repeat_max < 0)
            {
- #ifdef SUPPORT_UTF8
+ #ifdef SUPPORT_UTF
-           if (utf8 && c >= 128)
+           if (utf && (c & UTF_LENGTH) != 0)
              {
-             memcpy(code, utf8_char, c & 7);
+             memcpy(code, utf_chars, IN_UCHARS(c & 7));
              code += c & 7;
              }
            else
-Line 4646 
 for (;; ptr++)
+Line 4967 
 for (;; ptr++)
          else if (repeat_max != repeat_min)
            {
- #ifdef SUPPORT_UTF8
+ #ifdef SUPPORT_UTF
-           if (utf8 && c >= 128)
+           if (utf && (c & UTF_LENGTH) != 0)
              {
-             memcpy(code, utf8_char, c & 7);
+             memcpy(code, utf_chars, IN_UCHARS(c & 7));
              code += c & 7;
              }
            else
-Line 4676 
 for (;; ptr++)
+Line 4997 
 for (;; ptr++)
        /* The character or character type itself comes last in all cases. */
- #ifdef SUPPORT_UTF8
+ #ifdef SUPPORT_UTF
-       if (utf8 && c >= 128)
+       if (utf && (c & UTF_LENGTH) != 0)
          {
-         memcpy(code, utf8_char, c & 7);
+         memcpy(code, utf_chars, IN_UCHARS(c & 7));
          code += c & 7;
          }
        else
-Line 4703 
 for (;; ptr++)
+Line 5024 
 for (;; ptr++)
      else if (*previous == OP_CLASS ||
               *previous == OP_NCLASS ||
- #ifdef SUPPORT_UTF8
+ #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
               *previous == OP_XCLASS ||
  #endif
               *previous == OP_REF ||
-Line 4752 
 for (;; ptr++)
+Line 5073 
 for (;; ptr++)
        {
        register int i;
        int len = (int)(code - previous);
-       uschar *bralink = NULL;
+       pcre_uchar *bralink = NULL;
-       uschar *brazeroptr = NULL;
+       pcre_uchar *brazeroptr = NULL;
        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
        we just ignore the repeat. */
-Line 4806 
 for (;; ptr++)
+Line 5127 
 for (;; ptr++)
          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
            {
            *code = OP_END;
-           adjust_recurse(previous, 1, utf8, cd, save_hwm);
+           adjust_recurse(previous, 1, utf, cd, save_hwm);
-           memmove(previous+1, previous, len);
+           memmove(previous + 1, previous, IN_UCHARS(len));
            code++;
            if (repeat_max == 0)
              {
-Line 4830 
 for (;; ptr++)
+Line 5151 
 for (;; ptr++)
            {
            int offset;
            *code = OP_END;
-           adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd, save_hwm);
+           adjust_recurse(previous, 2 + LINK_SIZE, utf, cd, save_hwm);
-           memmove(previous + 2 + LINK_SIZE, previous, len);
+           memmove(previous + 2 + LINK_SIZE, previous, IN_UCHARS(len));
            code += 2 + LINK_SIZE;
            *previous++ = OP_BRAZERO + repeat_type;
            *previous++ = OP_BRA;
-Line 4877 
 for (;; ptr++)
+Line 5198 
 for (;; ptr++)
              *lengthptr += delta;
              }
-           /* This is compiling for real */
+           /* This is compiling for real. If there is a set first byte for
+           the group, and we have not yet set a "required byte", set it. Make
+           sure there is enough workspace for copying forward references before
+           doing the copy. */
            else
              {
-             if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;
+             if (groupsetfirstchar && reqchar < 0) reqchar = firstchar;
              for (i = 1; i < repeat_min; i++)
                {
-               uschar *hc;
+               pcre_uchar *hc;
-               uschar *this_hwm = cd->hwm;
+               pcre_uchar *this_hwm = cd->hwm;
-               memcpy(code, previous, len);
+               memcpy(code, previous, IN_UCHARS(len));
+               while (cd->hwm > cd->start_workspace + cd->workspace_size -
+                      WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
+                 {
+                 int save_offset = save_hwm - cd->start_workspace;
+                 int this_offset = this_hwm - cd->start_workspace;
+                 *errorcodeptr = expand_workspace(cd);
+                 if (*errorcodeptr != 0) goto FAILED;
+                 save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
+                 this_hwm = (pcre_uchar *)cd->start_workspace + this_offset;
+                 }
                for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
                  {
                  PUT(cd->hwm, 0, GET(hc, 0) + len);
-Line 4936 
 for (;; ptr++)
+Line 5273 
 for (;; ptr++)
          else for (i = repeat_max - 1; i >= 0; i--)
            {
-           uschar *hc;
+           pcre_uchar *hc;
-           uschar *this_hwm = cd->hwm;
+           pcre_uchar *this_hwm = cd->hwm;
            *code++ = OP_BRAZERO + repeat_type;
-Line 4953 
 for (;; ptr++)
+Line 5290 
 for (;; ptr++)
              PUTINC(code, 0, offset);
              }
-           memcpy(code, previous, len);
+           memcpy(code, previous, IN_UCHARS(len));
+           /* Ensure there is enough workspace for forward references before
+           copying them. */
+           while (cd->hwm > cd->start_workspace + cd->workspace_size -
+                  WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
+             {
+             int save_offset = save_hwm - cd->start_workspace;
+             int this_offset = this_hwm - cd->start_workspace;
+             *errorcodeptr = expand_workspace(cd);
+             if (*errorcodeptr != 0) goto FAILED;
+             save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
+             this_hwm = (pcre_uchar *)cd->start_workspace + this_offset;
+             }
            for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
              {
              PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
-Line 4970 
 for (;; ptr++)
+Line 5322 
 for (;; ptr++)
            {
            int oldlinkoffset;
            int offset = (int)(code - bralink + 1);
-           uschar *bra = code - offset;
+           pcre_uchar *bra = code - offset;
            oldlinkoffset = GET(bra, 1);
            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
            *code++ = OP_KET;
-Line 4984 
 for (;; ptr++)
+Line 5336 
 for (;; ptr++)
        ONCE brackets can be converted into non-capturing brackets, as the
        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
        deal with possessive ONCEs specially.
        Otherwise, when we are doing the actual compile phase, check to see
        whether this group is one that could match an empty string. If so,
        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
        that runtime checking can be done. [This check is also applied to ONCE
        groups at runtime, but in a different way.]
        Then, if the quantifier was possessive and the bracket is not a
        conditional, we convert the BRA code to the POS form, and the KET code to
        KETRPOS. (It turns out to be convenient at runtime to detect this kind of
        subpattern at both the start and at the end.) The use of special opcodes
        makes it possible to reduce greatly the stack usage in pcre_exec(). If
        the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
        Then, if the minimum number of matches is 1 or 0, cancel the possessive
        flag so that the default action below, of wrapping everything inside
        atomic brackets, does not happen. When the minimum is greater than 1,
        there will be earlier copies of the group, and so we still have to wrap
        the whole thing. */
        else
          {
-         uschar *ketcode = code - 1 - LINK_SIZE;
+         pcre_uchar *ketcode = code - 1 - LINK_SIZE;
-         uschar *bracode = ketcode - GET(ketcode, 1);
+         pcre_uchar *bracode = ketcode - GET(ketcode, 1);
          /* Convert possessive ONCE brackets to non-capturing */
          if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
              possessive_quantifier) *bracode = OP_BRA;
          /* For non-possessive ONCE brackets, all we need to do is to
          set the KET. */
          if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
            *ketcode = OP_KETRMAX + repeat_type;
          /* Handle non-ONCE brackets and possessive ONCEs (which have been
          converted to non-capturing above). */
          else
            {
            /* In the compile phase, check for empty string matching. */
            if (lengthptr == NULL)
              {
-             uschar *scode = bracode;
+             pcre_uchar *scode = bracode;
              do
                {
-               if (could_be_empty_branch(scode, ketcode, utf8, cd))
+               if (could_be_empty_branch(scode, ketcode, utf, cd))
                  {
                  *bracode += OP_SBRA - OP_BRA;
                  break;
-Line 5041 
 for (;; ptr++)
+Line 5393 
 for (;; ptr++)
                }
              while (*scode == OP_ALT);
              }
            /* Handle possessive quantifiers. */
            if (possessive_quantifier)
-Line 5050 
 for (;; ptr++)
+Line 5402 
 for (;; ptr++)
              repeated non-capturing bracket, because we have not invented POS
              versions of the COND opcodes. Because we are moving code along, we
              must ensure that any pending recursive references are updated. */
              if (*bracode == OP_COND || *bracode == OP_SCOND)
                {
                int nlen = (int)(code - bracode);
                *code = OP_END;
-               adjust_recurse(bracode, 1 + LINK_SIZE, utf8, cd, save_hwm);
+               adjust_recurse(bracode, 1 + LINK_SIZE, utf, cd, save_hwm);
-               memmove(bracode + 1+LINK_SIZE, bracode, nlen);
+               memmove(bracode + 1 + LINK_SIZE, bracode, IN_UCHARS(nlen));
                code += 1 + LINK_SIZE;
                nlen += 1 + LINK_SIZE;
                *bracode = OP_BRAPOS;
                *code++ = OP_KETRPOS;
                PUTINC(code, 0, nlen);
                PUT(bracode, 1, nlen);
                }
              /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
              else
                {
                *bracode += 1;              /* Switch to xxxPOS opcodes */
                *ketcode = OP_KETRPOS;
                }
              /* If the minimum is zero, mark it as possessive, then unset the
              possessive flag when the minimum is 0 or 1. */
              if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
              if (repeat_min < 2) possessive_quantifier = FALSE;
              }
            /* Non-possessive quantifier */
            else *ketcode = OP_KETRMAX + repeat_type;
            }
          }
-Line 5125 
 for (;; ptr++)
+Line 5477 
 for (;; ptr++)
        int len;
        if (*tempcode == OP_TYPEEXACT)
-         tempcode += _pcre_OP_lengths[*tempcode] +
+         tempcode += PRIV(OP_lengths)[*tempcode] +
-           ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
+           ((tempcode[1 + IMM2_SIZE] == OP_PROP
+           || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
          {
-         tempcode += _pcre_OP_lengths[*tempcode];
+         tempcode += PRIV(OP_lengths)[*tempcode];
- #ifdef SUPPORT_UTF8
+ #ifdef SUPPORT_UTF
-         if (utf8 && tempcode[-1] >= 0xc0)
+         if (utf && HAS_EXTRALEN(tempcode[-1]))
-           tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
+           tempcode += GET_EXTRALEN(tempcode[-1]);
  #endif
          }
-Line 5170 
 for (;; ptr++)
+Line 5523 
 for (;; ptr++)
          default:
          *code = OP_END;
-         adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
+         adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm);
-         memmove(tempcode + 1+LINK_SIZE, tempcode, len);
+         memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
          code += 1 + LINK_SIZE;
          len += 1 + LINK_SIZE;
          tempcode[0] = OP_ONCE;
-Line 5183 
 for (;; ptr++)
+Line 5536 
 for (;; ptr++)
        }
      /* In all case we no longer have a previous item. We also set the
-     "follows varying string" flag for subsequently encountered reqbytes if
+     "follows varying string" flag for subsequently encountered reqchars if
      it isn't already set and we have just passed a varying length item. */
      END_REPEAT:
-Line 5206 
 for (;; ptr++)
+Line 5559 
 for (;; ptr++)
      /* First deal with various "verbs" that can be introduced by '*'. */
-     if (*(++ptr) == CHAR_ASTERISK &&
+     ptr++;
-          ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
+     if (ptr[0] == CHAR_ASTERISK && (ptr[1] == ':'
+          || (MAX_255(ptr[1]) && ((cd->ctypes[ptr[1]] & ctype_letter) != 0))))
        {
        int i, namelen;
        int arglen = 0;
        const char *vn = verbnames;
-       const uschar *name = ptr + 1;
+       const pcre_uchar *name = ptr + 1;
-       const uschar *arg = NULL;
+       const pcre_uchar *arg = NULL;
        previous = NULL;
-       while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
+       ptr++;
+       while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;
        namelen = (int)(ptr - name);
        /* It appears that Perl allows any characters whatsoever, other than
-Line 5240 
 for (;; ptr++)
+Line 5595 
 for (;; ptr++)
        for (i = 0; i < verbcount; i++)
          {
          if (namelen == verbs[i].len &&
-             strncmp((char *)name, vn, namelen) == 0)
+             STRNCMP_UC_C8(name, vn, namelen) == 0)
            {
            /* Check for open captures before ACCEPT and convert it to
            ASSERT_ACCEPT if in an assertion. */
-Line 5261 
 for (;; ptr++)
+Line 5616 
 for (;; ptr++)
                }
              *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
-             /* Do not set firstbyte after *ACCEPT */
+             /* Do not set firstchar after *ACCEPT */
-             if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
+             if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
              }
            /* Handle other cases with/without an argument */
-Line 5288 
 for (;; ptr++)
+Line 5643 
 for (;; ptr++)
              *code = verbs[i].op_arg;
              if (*code++ == OP_THEN_ARG) cd->external_flags |= PCRE_HASTHEN;
              *code++ = arglen;
-             memcpy(code, arg, arglen);
+             memcpy(code, arg, IN_UCHARS(arglen));
              code += arglen;
              *code++ = 0;
              }
-Line 5311 
 for (;; ptr++)
+Line 5666 
 for (;; ptr++)
        {
        int i, set, unset, namelen;
        int *optset;
-       const uschar *name;
+       const pcre_uchar *name;
-       uschar *slot;
+       pcre_uchar *slot;
        switch (*(++ptr))
          {
-Line 5365 
 for (;; ptr++)
+Line 5720 
 for (;; ptr++)
            break;
          /* Most other conditions use OP_CREF (a couple change to OP_RREF
-         below), and all need to skip 3 bytes at the start of the group. */
+         below), and all need to skip 1+IMM2_SIZE bytes at the start of the group. */
          code[1+LINK_SIZE] = OP_CREF;
-         skipbytes = 3;
+         skipbytes = 1+IMM2_SIZE;
          refsign = -1;
          /* Check for a test for recursion in a named group. */
-Line 5401 
 for (;; ptr++)
+Line 5756 
 for (;; ptr++)
          /* We now expect to read a name; any thing else is an error */
-         if ((cd->ctypes[ptr[1]] & ctype_word) == 0)
+         if (!MAX_255(ptr[1]) || (cd->ctypes[ptr[1]] & ctype_word) == 0)
            {
            ptr += 1;  /* To get the right offset */
            *errorcodeptr = ERR28;
-Line 5412 
 for (;; ptr++)
+Line 5767 
 for (;; ptr++)
          recno = 0;
          name = ++ptr;
-         while ((cd->ctypes[*ptr] & ctype_word) != 0)
+         while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0)
            {
            if (recno >= 0)
-             recno = ((digitab[*ptr] & ctype_digit) != 0)?
+             recno = (IS_DIGIT(*ptr))? recno * 10 + *ptr - CHAR_0 : -1;
-               recno * 10 + *ptr - CHAR_0 : -1;
            ptr++;
            }
          namelen = (int)(ptr - name);
-Line 5464 
 for (;; ptr++)
+Line 5818 
 for (;; ptr++)
          slot = cd->name_table;
          for (i = 0; i < cd->names_found; i++)
            {
-           if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;
+           if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) == 0) break;
            slot += cd->name_entry_size;
            }
-Line 5480 
 for (;; ptr++)
+Line 5834 
 for (;; ptr++)
          /* Search the pattern for a forward reference */
          else if ((i = find_parens(cd, name, namelen,
-                         (options & PCRE_EXTENDED) != 0, utf8)) > 0)
+                         (options & PCRE_EXTENDED) != 0, utf)) > 0)
            {
            PUT2(code, 2+LINK_SIZE, i);
            code[1+LINK_SIZE]++;
-Line 5506 
 for (;; ptr++)
+Line 5860 
 for (;; ptr++)
            recno = 0;
            for (i = 1; i < namelen; i++)
              {
-             if ((digitab[name[i]] & ctype_digit) == 0)
+             if (!IS_DIGIT(name[i]))
                {
                *errorcodeptr = ERR15;
                goto FAILED;
-Line 5521 
 for (;; ptr++)
+Line 5875 
 for (;; ptr++)
          /* Similarly, check for the (?(DEFINE) "condition", which is always
          false. */
-         else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
+         else if (namelen == 6 && STRNCMP_UC_C8(name, STRING_DEFINE, 6) == 0)
            {
            code[1+LINK_SIZE] = OP_DEF;
            skipbytes = 1;
-Line 5584 
 for (;; ptr++)
+Line 5938 
 for (;; ptr++)
            break;
            default:                /* Could be name define, else bad */
-           if ((cd->ctypes[ptr[1]] & ctype_word) != 0) goto DEFINE_NAME;
+           if (MAX_255(ptr[1]) && (cd->ctypes[ptr[1]] & ctype_word) != 0)
+             goto DEFINE_NAME;
            ptr++;                  /* Correct offset for error */
            *errorcodeptr = ERR24;
            goto FAILED;
-Line 5606 
 for (;; ptr++)
+Line 5961 
 for (;; ptr++)
          *code++ = OP_CALLOUT;
            {
            int n = 0;
-           while ((digitab[*(++ptr)] & ctype_digit) != 0)
+           ptr++;
-             n = n * 10 + *ptr - CHAR_0;
+           while(IS_DIGIT(*ptr))
+             n = n * 10 + *ptr++ - CHAR_0;
            if (*ptr != CHAR_RIGHT_PARENTHESIS)
              {
              *errorcodeptr = ERR39;
-Line 5652 
 for (;; ptr++)
+Line 6008 
 for (;; ptr++)
              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
            name = ++ptr;
-           while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
+           while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
            namelen = (int)(ptr - name);
            /* In the pre-compile phase, just do a syntax check. */
-Line 5669 
 for (;; ptr++)
+Line 6025 
 for (;; ptr++)
                *errorcodeptr = ERR49;
                goto FAILED;
                }
-             if (namelen + 3 > cd->name_entry_size)
+             if (namelen + IMM2_SIZE + 1 > cd->name_entry_size)
                {
-               cd->name_entry_size = namelen + 3;
+               cd->name_entry_size = namelen + IMM2_SIZE + 1;
                if (namelen > MAX_NAME_SIZE)
                  {
                  *errorcodeptr = ERR48;
-Line 5700 
 for (;; ptr++)
+Line 6056 
 for (;; ptr++)
              for (i = 0; i < cd->names_found; i++)
                {
-               int crc = memcmp(name, slot+2, namelen);
+               int crc = memcmp(name, slot+IMM2_SIZE, IN_UCHARS(namelen));
                if (crc == 0)
                  {
-                 if (slot[2+namelen] == 0)
+                 if (slot[IMM2_SIZE+namelen] == 0)
                    {
                    if (GET2(slot, 0) != cd->bracount + 1 &&
                        (options & PCRE_DUPNAMES) == 0)
-Line 5724 
 for (;; ptr++)
+Line 6080 
 for (;; ptr++)
                if (crc < 0)
                  {
                  memmove(slot + cd->name_entry_size, slot,
-                   (cd->names_found - i) * cd->name_entry_size);
+                   IN_UCHARS((cd->names_found - i) * cd->name_entry_size));
                  break;
                  }
-Line 5738 
 for (;; ptr++)
+Line 6094 
 for (;; ptr++)
              if (!dupname)
                {
-               uschar *cslot = cd->name_table;
+               pcre_uchar *cslot = cd->name_table;
                for (i = 0; i < cd->names_found; i++)
                  {
                  if (cslot != slot)
-Line 5755 
 for (;; ptr++)
+Line 6111 
 for (;; ptr++)
                }
              PUT2(slot, 0, cd->bracount + 1);
-             memcpy(slot + 2, name, namelen);
+             memcpy(slot + IMM2_SIZE, name, IN_UCHARS(namelen));
-             slot[2+namelen] = 0;
+             slot[IMM2_SIZE + namelen] = 0;
              }
            }
-Line 5782 
 for (;; ptr++)
+Line 6138 
 for (;; ptr++)
          NAMED_REF_OR_RECURSE:
          name = ++ptr;
-         while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
+         while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
          namelen = (int)(ptr - name);
          /* In the pre-compile phase, do a syntax check. We used to just set
-Line 5794 
 for (;; ptr++)
+Line 6150 
 for (;; ptr++)
          if (lengthptr != NULL)
            {
-           const uschar *temp;
+           const pcre_uchar *temp;
            if (namelen == 0)
              {
-Line 5824 
 for (;; ptr++)
+Line 6180 
 for (;; ptr++)
            temp = cd->end_pattern;
            cd->end_pattern = ptr;
            recno = find_parens(cd, name, namelen,
-             (options & PCRE_EXTENDED) != 0, utf8);
+             (options & PCRE_EXTENDED) != 0, utf);
            cd->end_pattern = temp;
            if (recno < 0) recno = 0;    /* Forward ref; set dummy number */
            }
-Line 5839 
 for (;; ptr++)
+Line 6195 
 for (;; ptr++)
            slot = cd->name_table;
            for (i = 0; i < cd->names_found; i++)
              {
-             if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&
+             if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) == 0 &&
-                 slot[2+namelen] == 0)
+                 slot[IMM2_SIZE+namelen] == 0)
                break;
              slot += cd->name_entry_size;
              }
-Line 5851 
 for (;; ptr++)
+Line 6207 
 for (;; ptr++)
              }
            else if ((recno =                /* Forward back reference */
                      find_parens(cd, name, namelen,
-                       (options & PCRE_EXTENDED) != 0, utf8)) <= 0)
+                       (options & PCRE_EXTENDED) != 0, utf)) <= 0)
              {
              *errorcodeptr = ERR15;
              goto FAILED;
-Line 5876 
 for (;; ptr++)
+Line 6232 
 for (;; ptr++)
          case CHAR_0: case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4:
          case CHAR_5: case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
            {
-           const uschar *called;
+           const pcre_uchar *called;
            terminator = CHAR_RIGHT_PARENTHESIS;
            /* Come here from the \g<...> and \g'...' code (Oniguruma
-Line 5890 
 for (;; ptr++)
+Line 6246 
 for (;; ptr++)
            if ((refsign = *ptr) == CHAR_PLUS)
              {
              ptr++;
-             if ((digitab[*ptr] & ctype_digit) == 0)
+             if (!IS_DIGIT(*ptr))
                {
                *errorcodeptr = ERR63;
                goto FAILED;
-Line 5898 
 for (;; ptr++)
+Line 6254 
 for (;; ptr++)
              }
            else if (refsign == CHAR_MINUS)
              {
-             if ((digitab[ptr[1]] & ctype_digit) == 0)
+             if (!IS_DIGIT(ptr[1]))
                goto OTHER_CHAR_AFTER_QUERY;
              ptr++;
              }
            recno = 0;
-           while((digitab[*ptr] & ctype_digit) != 0)
+           while(IS_DIGIT(*ptr))
              recno = recno * 10 + *ptr++ - CHAR_0;
            if (*ptr != terminator)
-Line 5955 
 for (;; ptr++)
+Line 6311 
 for (;; ptr++)
              {
              *code = OP_END;
              if (recno != 0)
-               called = _pcre_find_bracket(cd->start_code, utf8, recno);
+               called = PRIV(find_bracket)(cd->start_code, utf, recno);
              /* Forward reference */
              if (called == NULL)
                {
                if (find_parens(cd, NULL, recno,
-                     (options & PCRE_EXTENDED) != 0, utf8) < 0)
+                     (options & PCRE_EXTENDED) != 0, utf) < 0)
                  {
                  *errorcodeptr = ERR15;
                  goto FAILED;
-Line 5973 
 for (;; ptr++)
+Line 6329 
 for (;; ptr++)
                of the group. Then remember the forward reference. */
                called = cd->start_code + recno;
+               if (cd->hwm >= cd->start_workspace + cd->workspace_size -
+                   WORK_SIZE_SAFETY_MARGIN)
+                 {
+                 *errorcodeptr = expand_workspace(cd);
+                 if (*errorcodeptr != 0) goto FAILED;
+                 }
                PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
                }
-Line 5986 
 for (;; ptr++)
+Line 6348 
 for (;; ptr++)
              conditional subpatterns will be picked up then. */
              else if (GET(called, 1) == 0 && cond_depth <= 0 &&
-                      could_be_empty(called, code, bcptr, utf8, cd))
+                      could_be_empty(called, code, bcptr, utf, cd))
                {
                *errorcodeptr = ERR40;
                goto FAILED;
                }
              }
-           /* Insert the recursion/subroutine item. */
+           /* Insert the recursion/subroutine item. It does not have a set first
+           character (relevant if it is repeated, because it will then be
+           wrapped with ONCE brackets). */
            *code = OP_RECURSE;
            PUT(code, 1, (int)(called - cd->start_code));
            code += 1 + LINK_SIZE;
+           groupsetfirstchar = FALSE;
            }
          /* Can't determine a first byte now */
-         if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
+         if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
          continue;
-Line 6059 
 for (;; ptr++)
+Line 6424 
 for (;; ptr++)
          both phases.
          If we are not at the pattern start, reset the greedy defaults and the
-         case value for firstbyte and reqbyte. */
+         case value for firstchar and reqchar. */
          if (*ptr == CHAR_RIGHT_PARENTHESIS)
            {
-Line 6072 
 for (;; ptr++)
+Line 6437 
 for (;; ptr++)
              {
              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
              greedy_non_default = greedy_default ^ 1;
-             req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
+             req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
              }
            /* Change options at this level, and pass them back for use
-Line 6109 
 for (;; ptr++)
+Line 6474 
 for (;; ptr++)
        NUMBERED_GROUP:
        cd->bracount += 1;
        PUT2(code, 1+LINK_SIZE, cd->bracount);
-       skipbytes = 2;
+       skipbytes = IMM2_SIZE;
        }
      /* Process nested bracketed regex. Assertions used not to be repeatable,
-Line 6135 
 for (;; ptr++)
+Line 6500 
 for (;; ptr++)
           skipbytes,                       /* Skip over bracket number */
           cond_depth +
             ((bravalue == OP_COND)?1:0),   /* Depth of condition subpatterns */
-          &subfirstbyte,                   /* For possible first char */
+          &subfirstchar,                   /* For possible first char */
-          &subreqbyte,                     /* For possible last char */
+          &subreqchar,                     /* For possible last char */
           bcptr,                           /* Current branch chain */
           cd,                              /* Tables block */
           (lengthptr == NULL)? NULL :      /* Actual compile phase */
-Line 6164 
 for (;; ptr++)
+Line 6529 
 for (;; ptr++)
      if (bravalue == OP_COND && lengthptr == NULL)
        {
-       uschar *tc = code;
+       pcre_uchar *tc = code;
        int condcount = 0;
        do {
-Line 6187 
 for (;; ptr++)
+Line 6552 
 for (;; ptr++)
          }
        /* A "normal" conditional group. If there is just one branch, we must not
-       make use of its firstbyte or reqbyte, because this is equivalent to an
+       make use of its firstchar or reqchar, because this is equivalent to an
        empty second branch. */
        else
-Line 6197 
 for (;; ptr++)
+Line 6562 
 for (;; ptr++)
            *errorcodeptr = ERR27;
            goto FAILED;
            }
-         if (condcount == 1) subfirstbyte = subreqbyte = REQ_NONE;
+         if (condcount == 1) subfirstchar = subreqchar = REQ_NONE;
          }
        }
-Line 6241 
 for (;; ptr++)
+Line 6606 
 for (;; ptr++)
      /* Handle updating of the required and first characters for other types of
      group. Update for normal brackets of all kinds, and conditions with two
      branches (see code above). If the bracket is followed by a quantifier with
-     zero repeat, we have to back off. Hence the definition of zeroreqbyte and
+     zero repeat, we have to back off. Hence the definition of zeroreqchar and
-     zerofirstbyte outside the main loop so that they can be accessed for the
+     zerofirstchar outside the main loop so that they can be accessed for the
      back off. */
-     zeroreqbyte = reqbyte;
+     zeroreqchar = reqchar;
-     zerofirstbyte = firstbyte;
+     zerofirstchar = firstchar;
-     groupsetfirstbyte = FALSE;
+     groupsetfirstchar = FALSE;
      if (bravalue >= OP_ONCE)
        {
-       /* If we have not yet set a firstbyte in this branch, take it from the
+       /* If we have not yet set a firstchar in this branch, take it from the
        subpattern, remembering that it was set here so that a repeat of more
-       than one can replicate it as reqbyte if necessary. If the subpattern has
+       than one can replicate it as reqchar if necessary. If the subpattern has
-       no firstbyte, set "none" for the whole branch. In both cases, a zero
+       no firstchar, set "none" for the whole branch. In both cases, a zero
-       repeat forces firstbyte to "none". */
+       repeat forces firstchar to "none". */
-       if (firstbyte == REQ_UNSET)
+       if (firstchar == REQ_UNSET)
          {
-         if (subfirstbyte >= 0)
+         if (subfirstchar >= 0)
            {
-           firstbyte = subfirstbyte;
+           firstchar = subfirstchar;
-           groupsetfirstbyte = TRUE;
+           groupsetfirstchar = TRUE;
            }
-         else firstbyte = REQ_NONE;
+         else firstchar = REQ_NONE;
-         zerofirstbyte = REQ_NONE;
+         zerofirstchar = REQ_NONE;
          }
-       /* If firstbyte was previously set, convert the subpattern's firstbyte
+       /* If firstchar was previously set, convert the subpattern's firstchar
-       into reqbyte if there wasn't one, using the vary flag that was in
+       into reqchar if there wasn't one, using the vary flag that was in
        existence beforehand. */
-       else if (subfirstbyte >= 0 && subreqbyte < 0)
+       else if (subfirstchar >= 0 && subreqchar < 0)
-         subreqbyte = subfirstbyte | tempreqvary;
+         subreqchar = subfirstchar | tempreqvary;
        /* If the subpattern set a required byte (or set a first byte that isn't
        really the first byte - see above), set it. */
-       if (subreqbyte >= 0) reqbyte = subreqbyte;
+       if (subreqchar >= 0) reqchar = subreqchar;
        }
-     /* For a forward assertion, we take the reqbyte, if set. This can be
+     /* For a forward assertion, we take the reqchar, if set. This can be
      helpful if the pattern that follows the assertion doesn't set a different
-     char. For example, it's useful for /(?=abcde).+/. We can't set firstbyte
+     char. For example, it's useful for /(?=abcde).+/. We can't set firstchar
      for an assertion, however because it leads to incorrect effect for patterns
-     such as /(?=a)a.+/ when the "real" "a" would then become a reqbyte instead
+     such as /(?=a)a.+/ when the "real" "a" would then become a reqchar instead
-     of a firstbyte. This is overcome by a scan at the end if there's no
+     of a firstchar. This is overcome by a scan at the end if there's no
-     firstbyte, looking for an asserted first char. */
+     firstchar, looking for an asserted first char. */
-     else if (bravalue == OP_ASSERT && subreqbyte >= 0) reqbyte = subreqbyte;
+     else if (bravalue == OP_ASSERT && subreqchar >= 0) reqchar = subreqchar;
      break;     /* End of processing '(' */
-Line 6322 
 for (;; ptr++)
+Line 6687 
 for (;; ptr++)
        /* For metasequences that actually match a character, we disable the
        setting of a first character if it hasn't already been set. */
-       if (firstbyte == REQ_UNSET && -c > ESC_b && -c < ESC_Z)
+       if (firstchar == REQ_UNSET && -c > ESC_b && -c < ESC_Z)
-         firstbyte = REQ_NONE;
+         firstchar = REQ_NONE;
        /* Set values to reset to if this is followed by a zero repeat. */
-       zerofirstbyte = firstbyte;
+       zerofirstchar = firstchar;
-       zeroreqbyte = reqbyte;
+       zeroreqchar = reqchar;
        /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
        is a subroutine call by number (Oniguruma syntax). In fact, the value
-Line 6339 
 for (;; ptr++)
+Line 6704 
 for (;; ptr++)
        if (-c == ESC_g)
          {
-         const uschar *p;
+         const pcre_uchar *p;
          save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
            CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
-Line 6356 
 for (;; ptr++)
+Line 6721 
 for (;; ptr++)
          if (ptr[1] != CHAR_PLUS && ptr[1] != CHAR_MINUS)
            {
-           BOOL isnumber = TRUE;
+           BOOL is_a_number = TRUE;
            for (p = ptr + 1; *p != 0 && *p != terminator; p++)
              {
-             if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;
+             if (!MAX_255(*p)) { is_a_number = FALSE; break; }
+             if ((cd->ctypes[*p] & ctype_digit) == 0) is_a_number = FALSE;
              if ((cd->ctypes[*p] & ctype_word) == 0) break;
              }
            if (*p != terminator)
-Line 6367 
 for (;; ptr++)
+Line 6733 
 for (;; ptr++)
              *errorcodeptr = ERR57;
              break;
              }
-           if (isnumber)
+           if (is_a_number)
              {
              ptr++;
              goto HANDLE_NUMERICAL_RECURSION;
-Line 6379 
 for (;; ptr++)
+Line 6745 
 for (;; ptr++)
          /* Test a signed number in angle brackets or quotes. */
          p = ptr + 2;
-         while ((digitab[*p] & ctype_digit) != 0) p++;
+         while (IS_DIGIT(*p)) p++;
          if (*p != terminator)
            {
            *errorcodeptr = ERR57;
-Line 6407 
 for (;; ptr++)
+Line 6773 
 for (;; ptr++)
          goto NAMED_REF_OR_RECURSE;
          }
-       /* Back references are handled specially; must disable firstbyte if
+       /* Back references are handled specially; must disable firstchar if
        not set to cope with cases like (?=(\w+))\1: which would otherwise set
        ':' later. */
-Line 6417 
 for (;; ptr++)
+Line 6783 
 for (;; ptr++)
          recno = -c - ESC_REF;
          HANDLE_REFERENCE:    /* Come here from named backref handling */
-         if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
+         if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
          previous = code;
          *code++ = ((options & PCRE_CASELESS) != 0)? OP_REFI : OP_REF;
          PUT2INC(code, 0, recno);
-Line 6481 
 for (;; ptr++)
+Line 6847 
 for (;; ptr++)
  #endif
          /* In non-UTF-8 mode, we turn \C into OP_ALLANY instead of OP_ANYBYTE
          so that it works in DFA mode and in lookbehinds. */
            {
            previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;
-           *code++ = (!utf8 && c == -ESC_C)? OP_ALLANY : -c;
+           *code++ = (!utf && c == -ESC_C)? OP_ALLANY : -c;
            }
          }
        continue;
-Line 6494 
 for (;; ptr++)
+Line 6860 
 for (;; ptr++)
      a value > 127. We set its representation in the length/buffer, and then
      handle it as a data character. */
- #ifdef SUPPORT_UTF8
+ #ifdef SUPPORT_UTF
-     if (utf8 && c > 127)
+     if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
-       mclength = _pcre_ord2utf8(c, mcbuffer);
+       mclength = PRIV(ord2utf)(c, mcbuffer);
      else
  #endif
-Line 6517 
 for (;; ptr++)
+Line 6883 
 for (;; ptr++)
      mclength = 1;
      mcbuffer[0] = c;
- #ifdef SUPPORT_UTF8
+ #ifdef SUPPORT_UTF
-     if (utf8 && c >= 0xc0)
+     if (utf && HAS_EXTRALEN(c))
-       {
+       ACROSSCHAR(TRUE, ptr[1], mcbuffer[mclength++] = *(++ptr));
-       while ((ptr[1] & 0xc0) == 0x80)
-         mcbuffer[mclength++] = *(++ptr);
-       }
  #endif
      /* At this point we have the character's bytes in mcbuffer, and the length
-Line 6540 
 for (;; ptr++)
+Line 6903 
 for (;; ptr++)
      /* Set the first and required bytes appropriately. If no previous first
      byte, set it from this character, but revert to none on a zero repeat.
-     Otherwise, leave the firstbyte value alone, and don't change it on a zero
+     Otherwise, leave the firstchar value alone, and don't change it on a zero
      repeat. */
-     if (firstbyte == REQ_UNSET)
+     if (firstchar == REQ_UNSET)
        {
-       zerofirstbyte = REQ_NONE;
+       zerofirstchar = REQ_NONE;
-       zeroreqbyte = reqbyte;
+       zeroreqchar = reqchar;
-       /* If the character is more than one byte long, we can set firstbyte
+       /* If the character is more than one byte long, we can set firstchar
        only if it is not to be matched caselessly. */
        if (mclength == 1 || req_caseopt == 0)
          {
-         firstbyte = mcbuffer[0] | req_caseopt;
+         firstchar = mcbuffer[0] | req_caseopt;
-         if (mclength != 1) reqbyte = code[-1] | cd->req_varyopt;
+         if (mclength != 1) reqchar = code[-1] | cd->req_varyopt;
          }
-       else firstbyte = reqbyte = REQ_NONE;
+       else firstchar = reqchar = REQ_NONE;