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

Diff of /code/trunk/pcre_compile.c

Parent Directory | Revision Log | View Patch Patch

-revision 624 by ph10,
Tue Jul 19 10:43:28 2011 UTC
+revision 927 by ph10,
Wed Feb 22 15:15:08 2012 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 365 
 static const char error_texts[] =
+Line 438 
 static const char error_texts[] =
    /* 30 */
    "unknown POSIX class name\0"
    "POSIX collating elements are not supported\0"
-   "this version of PCRE is not compiled with PCRE_UTF8 support\0"
+   "this version of PCRE is compiled without UTF support\0"
    "spare error\0"  /** DEAD **/
    "character value in \\x{...} sequence is too large\0"
    /* 35 */
-Line 388 
 static const char error_texts[] =
+Line 461 
 static const char error_texts[] =
    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
    /* 50 */
    "repeated subpattern is too long\0"    /** DEAD **/
-   "octal value is greater than \\377 (not in UTF-8 mode)\0"
+   "octal value is greater than \\377 in 8-bit non-UTF-8 mode\0"
    "internal error: overran compiling workspace\0"
    "internal error: previously-checked referenced subpattern not found\0"
    "DEFINE group contains more than one branch\0"
    /* 55 */
-   "repeating a DEFINE group is not allowed\0"
+   "repeating a DEFINE group is not allowed\0"  /** DEAD **/
    "inconsistent NEWLINE options\0"
    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
    "a numbered reference must not be zero\0"
-Line 407 
 static const char error_texts[] =
+Line 480 
 static const char error_texts[] =
    /* 65 */
    "different names for subpatterns of the same number are not allowed\0"
    "(*MARK) must have an argument\0"
-   "this version of PCRE is not compiled with PCRE_UCP support\0"
+   "this version of PCRE is not compiled with Unicode property support\0"
    "\\c must be followed by an ASCII character\0"
+   "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
+   /* 70 */
+   "internal error: unknown opcode in find_fixedlength()\0"
+   "\\N is not supported in a class\0"
+   "too many forward references\0"
+   "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
+   "invalid UTF-16 string\0"
    ;
  /* Table to identify digits and hex digits. This is used when compiling
-Line 427 
 For convenience, we use the same bit def
+Line 507 
 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 471 
 static const unsigned char digitab[] =
+Line 557 
 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 506 
 static const unsigned char digitab[] =
+Line 592 
 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 545 
 static const unsigned char ebcdic_charta
+Line 631 
 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 *, branch_chain *, compile_data *, int *);
+     int *, int *, branch_chain *, compile_data *, int *);
-Line 577 
 return s;
+Line 663 
 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            *
+ *************************************************/
+ /* This function is called when a '{' is encountered in a place where it might
+ start a quantifier. It looks ahead to see if it really is a quantifier or not.
+ It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
+ where the ddds are digits.
+ Arguments:
+   p         pointer to the first char after '{'
+ Returns:    TRUE or FALSE
+ */
+ static BOOL
+ is_counted_repeat(const pcre_uchar *p)
+ {
+ if (!IS_DIGIT(*p)) return FALSE;
+ 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 (!IS_DIGIT(*p)) return FALSE;
+ p++;
+ while (IS_DIGIT(*p)) p++;
+ return (*p == CHAR_RIGHT_CURLY_BRACKET);
+ }
+ /*************************************************
  *            Handle escapes                      *
  *************************************************/
-Line 601 
 Returns:         zero or positive => a d
+Line 759 
 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 620 
 in a table. A non-zero result is somethi
+Line 780 
 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 632 
 else if ((i = escapes[c - 0x48]) != 0)
+Line 794 
 else if ((i = escapes[c - 0x48]) != 0)
  else
    {
-   const uschar *oldptr;
+   const pcre_uchar *oldptr;
    BOOL braced, negated;
    switch (c)
-Line 642 
 else
+Line 804 
 else
      case CHAR_l:
      case CHAR_L:
+     *errorcodeptr = ERR37;
+     break;
      case CHAR_u:
+     if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
+       {
+       /* In JavaScript, \u must be followed by four hexadecimal numbers.
+       Otherwise it is a lowercase u letter. */
+       if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
+         && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
+         && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
+         && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
+         {
+         c = 0;
+         for (i = 0; i < 4; ++i)
+           {
+           register int cc = *(++ptr);
+ #ifndef EBCDIC  /* ASCII/UTF-8 coding */
+           if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
+           c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
+ #else           /* EBCDIC coding */
+           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
+           }
+         }
+       }
+     else
+       *errorcodeptr = ERR37;
+     break;
      case CHAR_U:
-     *errorcodeptr = ERR37;
+     /* In JavaScript, \U is an uppercase U letter. */
+     if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
      break;
-     /* \g must be followed by one of a number of specific things:
+     /* In a character class, \g is just a literal "g". Outside a character
+     class, \g must be followed by one of a number of specific things:
      (1) A number, either plain or braced. If positive, it is an absolute
      backreference. If negative, it is a relative backreference. This is a Perl
-Line 664 
 else
+Line 858 
 else
      the -ESC_g code (cf \k). */
      case CHAR_g:
+     if (isclass) break;
      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
        {
        c = -ESC_g;
-Line 674 
 else
+Line 869 
 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 694 
 else
+Line 889 
 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 747 
 else
+Line 951 
 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 777 
 else
+Line 991 
 else
      /* \0 always starts an octal number, but we may drop through to here with a
      larger first octal digit. The original code used just to take the least
      significant 8 bits of octal numbers (I think this is what early Perls used
-     to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
+     to do). Nowadays we allow for larger numbers in UTF-8 mode and 16-bit mode,
-     than 3 octal digits. */
+     but no more than 3 octal digits. */
      case CHAR_0:
      c -= CHAR_0;
      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
          c = c * 8 + *(++ptr) - CHAR_0;
-     if (!utf8 && c > 255) *errorcodeptr = ERR51;
+ #ifdef COMPILE_PCRE8
+     if (!utf && c > 0xff) *errorcodeptr = ERR51;
+ #endif
      break;
      /* \x is complicated. \x{ddd} is a character number which can be greater
-     than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is
+     than 0xff in utf or non-8bit mode, but only if the ddd are hex digits.
-     treated as a data character. */
+     If not, { is treated as a data character. */
      case CHAR_x:
+     if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
+       {
+       /* In JavaScript, \x must be followed by two hexadecimal numbers.
+       Otherwise it is a lowercase x letter. */
+       if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
+         && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
+         {
+         c = 0;
+         for (i = 0; i < 2; ++i)
+           {
+           register int cc = *(++ptr);
+ #ifndef EBCDIC  /* ASCII/UTF-8 coding */
+           if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
+           c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
+ #else           /* EBCDIC coding */
+           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
+           }
+         }
+       break;
+       }
      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 811 
 else
+Line 1048 
 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 827 
 else
+Line 1078 
 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 885 
 else
+Line 1136 
 else
    }
  /* Perl supports \N{name} for character names, as well as plain \N for "not
- newline". PCRE does not support \N{name}. */
+ newline". PCRE does not support \N{name}. However, it does support
+ quantification such as \N{2,3}. */
- if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)
+ if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
+      !is_counted_repeat(ptr+2))
    *errorcodeptr = ERR37;
  /* If PCRE_UCP is set, we change the values for \d etc. */
-Line 923 
 Returns:         type value from ucp_typ
+Line 1176 
 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 944 
 if (c == CHAR_LEFT_CURLY_BRACKET)
+Line 1197 
 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 968 
 else
+Line 1221 
 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 997 
 return -1;
+Line 1250 
 return -1;
  /*************************************************
- *            Check for counted repeat            *
- *************************************************/
- /* This function is called when a '{' is encountered in a place where it might
- start a quantifier. It looks ahead to see if it really is a quantifier or not.
- It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
- where the ddds are digits.
- Arguments:
-   p         pointer to the first char after '{'
- Returns:    TRUE or FALSE
- */
- static BOOL
- is_counted_repeat(const uschar *p)
- {
- if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
- while ((digitab[*p] & ctype_digit) != 0) 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;
- while ((digitab[*p] & ctype_digit) != 0) p++;
- return (*p == CHAR_RIGHT_CURLY_BRACKET);
- }
- /*************************************************
  *         Read repeat counts                     *
  *************************************************/
-Line 1048 
 Returns:         pointer to '}' on succe
+Line 1268 
 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 1057 
 int max = -1;
+Line 1277 
 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 1072 
 if (*p == CHAR_RIGHT_CURLY_BRACKET) max
+Line 1292 
 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 1127 
 Arguments:
+Line 1347 
 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 1204 
 if (ptr[0] == CHAR_LEFT_PARENTHESIS)
+Line 1424 
 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 1212 
 if (ptr[0] == CHAR_LEFT_PARENTHESIS)
+Line 1432 
 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 1255 
 for (; ptr < cd->end_pattern; ptr++)
+Line 1475 
 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 1303 
 for (; ptr < cd->end_pattern; ptr++)
+Line 1523 
 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 1315 
 for (; ptr < cd->end_pattern; ptr++)
+Line 1535 
 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 1361 
 Arguments:
+Line 1581 
 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 1381 
 matching closing parens. That is why we
+Line 1601 
 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 1408 
 Arguments:
+Line 1628 
 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 1420 
 for (;;)
+Line 1640 
 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 1434 
 for (;;)
+Line 1654 
 for (;;)
      case OP_RREF:
      case OP_NRREF:
      case OP_DEF:
-     code += _pcre_OP_lengths[*code];
+     code += PRIV(OP_lengths)[*code];
      break;
      default:
-Line 1464 
 and doing the check at the end; a flag s
+Line 1684 
 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
  Returns:   the fixed length,
               or -1 if there is no fixed length,
-              or -2 if \C was encountered
+              or -2 if \C was encountered (in UTF-8 mode only)
               or -3 if an OP_RECURSE item was encountered and atend is FALSE
+              or -4 if an unknown opcode was encountered (internal error)
  */
  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 1488 
 branch, check the length against that of
+Line 1709 
 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
      OP_BRA (normal non-capturing bracket) because the other variants of these
      opcodes are all concerned with unlimited repeated groups, which of course
-     are not of fixed length. They will cause a -1 response from the default
+     are not of fixed length. */
-     case of this switch. */
      case OP_CBRA:
      case OP_BRA:
      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);
      cc += 1 + LINK_SIZE;
      break;
-     /* Reached end of a branch; if it's a ket it is the end of a nested
+     /* Reached end of a branch; if it's a ket it is the end of a nested call.
-     call. If it's ALT it is an alternation in a nested call. If it is
+     If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
-     END it's the end of the outer call. All can be handled by the same code.
+     an ALT. If it is END it's the end of the outer call. All can be handled by
-     Note that we must not include the OP_KETRxxx opcodes here, because they
+     the same code. Note that we must not include the OP_KETRxxx opcodes here,
-     all imply an unlimited repeat. */
+     because they all imply an unlimited repeat. */
      case OP_ALT:
      case OP_KET:
      case OP_END:
+     case OP_ACCEPT:
+     case OP_ASSERT_ACCEPT:
      if (length < 0) length = branchlength;
        else if (length != branchlength) return -1;
      if (*cc != OP_ALT) return length;
-Line 1531 
 for (;;)
+Line 1755 
 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 1547 
 for (;;)
+Line 1771 
 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 */
-     case OP_REVERSE:
+     case OP_MARK:
-     case OP_CREF:
+     case OP_PRUNE_ARG:
-     case OP_NCREF:
+     case OP_SKIP_ARG:
-     case OP_RREF:
+     case OP_THEN_ARG:
-     case OP_NRREF:
+     cc += cc[1] + PRIV(OP_lengths)[*cc];
-     case OP_DEF:
+     break;
      case OP_CALLOUT:
-     case OP_SOD:
-     case OP_SOM:
-     case OP_SET_SOM:
-     case OP_EOD:
-     case OP_EODN:
      case OP_CIRC:
      case OP_CIRCM:
+     case OP_CLOSE:
+     case OP_COMMIT:
+     case OP_CREF:
+     case OP_DEF:
      case OP_DOLL:
      case OP_DOLLM:
+     case OP_EOD:
+     case OP_EODN:
+     case OP_FAIL:
+     case OP_NCREF:
+     case OP_NRREF:
      case OP_NOT_WORD_BOUNDARY:
+     case OP_PRUNE:
+     case OP_REVERSE:
+     case OP_RREF:
+     case OP_SET_SOM:
+     case OP_SKIP:
+     case OP_SOD:
+     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 1580 
 for (;;)
+Line 1818 
 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 1589 
 for (;;)
+Line 1827 
 for (;;)
      need to skip over a multibyte character in UTF8 mode.  */
      case OP_EXACT:
+     case OP_EXACTI:
+     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 1609 
 for (;;)
+Line 1850 
 for (;;)
      cc += 2;
      /* Fall through */
+     case OP_HSPACE:
+     case OP_VSPACE:
+     case OP_NOT_HSPACE:
+     case OP_NOT_VSPACE:
      case OP_NOT_DIGIT:
      case OP_DIGIT:
      case OP_NOT_WHITESPACE:
-Line 1621 
 for (;;)
+Line 1866 
 for (;;)
      cc++;
      break;
-     /* The single-byte matcher isn't allowed */
+     /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
+     otherwise \C is coded as OP_ALLANY. */
      case OP_ANYBYTE:
      return -2;
      /* 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)
        {
+       case OP_CRPLUS:
+       case OP_CRMINPLUS:
        case OP_CRSTAR:
        case OP_CRMINSTAR:
        case OP_CRQUERY:
-Line 1648 
 for (;;)
+Line 1896 
 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 1660 
 for (;;)
+Line 1908 
 for (;;)
      /* Anything else is variable length */
-     default:
+     case OP_ANYNL:
+     case OP_BRAMINZERO:
+     case OP_BRAPOS:
+     case OP_BRAPOSZERO:
+     case OP_BRAZERO:
+     case OP_CBRAPOS:
+     case OP_EXTUNI:
+     case OP_KETRMAX:
+     case OP_KETRMIN:
+     case OP_KETRPOS:
+     case OP_MINPLUS:
+     case OP_MINPLUSI:
+     case OP_MINQUERY:
+     case OP_MINQUERYI:
+     case OP_MINSTAR:
+     case OP_MINSTARI:
+     case OP_MINUPTO:
+     case OP_MINUPTOI:
+     case OP_NOTMINPLUS:
+     case OP_NOTMINPLUSI:
+     case OP_NOTMINQUERY:
+     case OP_NOTMINQUERYI:
+     case OP_NOTMINSTAR:
+     case OP_NOTMINSTARI:
+     case OP_NOTMINUPTO:
+     case OP_NOTMINUPTOI:
+     case OP_NOTPLUS:
+     case OP_NOTPLUSI:
+     case OP_NOTPOSPLUS:
+     case OP_NOTPOSPLUSI:
+     case OP_NOTPOSQUERY:
+     case OP_NOTPOSQUERYI:
+     case OP_NOTPOSSTAR:
+     case OP_NOTPOSSTARI:
+     case OP_NOTPOSUPTO:
+     case OP_NOTPOSUPTOI:
+     case OP_NOTQUERY:
+     case OP_NOTQUERYI:
+     case OP_NOTSTAR:
+     case OP_NOTSTARI:
+     case OP_NOTUPTO:
+     case OP_NOTUPTOI:
+     case OP_PLUS:
+     case OP_PLUSI:
+     case OP_POSPLUS:
+     case OP_POSPLUSI:
+     case OP_POSQUERY:
+     case OP_POSQUERYI:
+     case OP_POSSTAR:
+     case OP_POSSTARI:
+     case OP_POSUPTO:
+     case OP_POSUPTOI:
+     case OP_QUERY:
+     case OP_QUERYI:
+     case OP_REF:
+     case OP_REFI:
+     case OP_SBRA:
+     case OP_SBRAPOS:
+     case OP_SCBRA:
+     case OP_SCBRAPOS:
+     case OP_SCOND:
+     case OP_SKIPZERO:
+     case OP_STAR:
+     case OP_STARI:
+     case OP_TYPEMINPLUS:
+     case OP_TYPEMINQUERY:
+     case OP_TYPEMINSTAR:
+     case OP_TYPEMINUPTO:
+     case OP_TYPEPLUS:
+     case OP_TYPEPOSPLUS:
+     case OP_TYPEPOSQUERY:
+     case OP_TYPEPOSSTAR:
+     case OP_TYPEPOSUPTO:
+     case OP_TYPEQUERY:
+     case OP_TYPESTAR:
+     case OP_TYPEUPTO:
+     case OP_UPTO:
+     case OP_UPTOI:
      return -1;
+     /* Catch unrecognized opcodes so that when new ones are added they
+     are not forgotten, as has happened in the past. */
+     default:
+     return -4;
      }
    }
  /* Control never gets here */
-Line 1682 
 length.
+Line 2013 
 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 1707 
 for (;;)
+Line 2038 
 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 1717 
 for (;;)
+Line 2048 
 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 1746 
 for (;;)
+Line 2077 
 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 1756 
 for (;;)
+Line 2088 
 for (;;)
        break;
        case OP_THEN_ARG:
-       code += code[1+LINK_SIZE];
+       code += code[1];
        break;
        }
      /* 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 1799 
 for (;;)
+Line 2131 
 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 1820 
 instance of OP_RECURSE.
+Line 2152 
 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 1865 
 for (;;)
+Line 2197 
 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 1875 
 for (;;)
+Line 2208 
 for (;;)
        break;
        case OP_THEN_ARG:
-       code += code[1+LINK_SIZE];
+       code += code[1];
        break;
        }
      /* 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:
+       case OP_NOT:
+       case OP_NOTI:
        case OP_EXACT:
        case OP_EXACTI:
+       case OP_NOTEXACT:
+       case OP_NOTEXACTI:
        case OP_UPTO:
        case OP_UPTOI:
+       case OP_NOTUPTO:
+       case OP_NOTUPTOI:
        case OP_MINUPTO:
        case OP_MINUPTOI:
+       case OP_NOTMINUPTO:
+       case OP_NOTMINUPTOI:
        case OP_POSUPTO:
        case OP_POSUPTOI:
+       case OP_NOTPOSUPTO:
+       case OP_NOTPOSUPTOI:
        case OP_STAR:
        case OP_STARI:
+       case OP_NOTSTAR:
+       case OP_NOTSTARI:
        case OP_MINSTAR:
        case OP_MINSTARI:
+       case OP_NOTMINSTAR:
+       case OP_NOTMINSTARI:
        case OP_POSSTAR:
        case OP_POSSTARI:
+       case OP_NOTPOSSTAR:
+       case OP_NOTPOSSTARI:
        case OP_PLUS:
        case OP_PLUSI:
+       case OP_NOTPLUS:
+       case OP_NOTPLUSI:
        case OP_MINPLUS:
        case OP_MINPLUSI:
+       case OP_NOTMINPLUS:
+       case OP_NOTMINPLUSI:
        case OP_POSPLUS:
        case OP_POSPLUSI:
+       case OP_NOTPOSPLUS:
+       case OP_NOTPOSPLUSI:
        case OP_QUERY:
        case OP_QUERYI:
+       case OP_NOTQUERY:
+       case OP_NOTQUERYI:
        case OP_MINQUERY:
        case OP_MINQUERYI:
+       case OP_NOTMINQUERY:
+       case OP_NOTMINQUERYI:
        case OP_POSQUERY:
        case OP_POSQUERYI:
-       if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
+       case OP_NOTPOSQUERY:
+       case OP_NOTPOSQUERYI:
+       if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
        break;
        }
  #else
-     (void)(utf8);  /* Keep compiler happy by referencing function argument */
+     (void)(utf);  /* Keep compiler happy by referencing function argument */
  #endif
      }
    }
-Line 1945 
 bracket whose current branch will alread
+Line 2306 
 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 1977 
 for (code = first_significant_code(code
+Line 2338 
 for (code = first_significant_code(code
    /* For a recursion/subroutine call, if its end has been reached, which
    implies a backward reference subroutine call, we can scan it. If it's a
    forward reference subroutine call, we can't. To detect forward reference
    we have to scan up the list that is kept in the workspace. This function is
    called only when doing the real compile, not during the pre-compile that
    measures the size of the compiled pattern. */
    if (c == OP_RECURSE)
      {
-     const uschar *scode;
+     const pcre_uchar *scode;
      BOOL empty_branch;
      /* Test for forward reference */
      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
        if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
      /* Not a forward reference, test for completed backward reference */
      empty_branch = FALSE;
      scode = cd->start_code + GET(code, 1);
      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
      /* Completed backwards reference */
      do
        {
-       if (could_be_empty_branch(scode, endcode, utf8, cd))
+       if (could_be_empty_branch(scode, endcode, utf, cd))
          {
          empty_branch = TRUE;
          break;
-Line 2009 
 for (code = first_significant_code(code
+Line 2370 
 for (code = first_significant_code(code
        scode += GET(scode, 1);
        }
      while (*scode == OP_ALT);
      if (!empty_branch) return FALSE;  /* All branches are non-empty */
      continue;
      }
-Line 2019 
 for (code = first_significant_code(code
+Line 2380 
 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 2040 
 for (code = first_significant_code(code
+Line 2401 
 for (code = first_significant_code(code
    if (c == OP_BRA  || c == OP_BRAPOS ||
        c == OP_CBRA || c == OP_CBRAPOS ||
-       c == OP_ONCE || c == OP_COND)
+       c == OP_ONCE || c == OP_ONCE_NC ||
+       c == OP_COND)
      {
      BOOL empty_branch;
      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
-Line 2056 
 for (code = first_significant_code(code
+Line 2418 
 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 2074 
 for (code = first_significant_code(code
+Line 2436 
 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 2086 
 for (code = first_significant_code(code
+Line 2448 
 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 2161 
 for (code = first_significant_code(code
+Line 2523 
 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 2176 
 for (code = first_significant_code(code
+Line 2539 
 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 2189 
 for (code = first_significant_code(code
+Line 2552 
 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 2198 
 for (code = first_significant_code(code
+Line 2561 
 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 2212 
 for (code = first_significant_code(code
+Line 2575 
 for (code = first_significant_code(code
      break;
      case OP_THEN_ARG:
-     code += code[1+LINK_SIZE];
+     code += code[1];
      break;
      /* None of the remaining opcodes are required to match a character. */
-Line 2235 
 return TRUE;
+Line 2598 
 return TRUE;
  the current branch of the current pattern to see if it could match the empty
  string. If it could, we must look outwards for branches at other levels,
  stopping when we pass beyond the bracket which is the subject of the recursion.
  This function is called only during the real compile, not during the
  pre-compile.
  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 2287 
 where Perl recognizes it as the POSIX cl
+Line 2650 
 where Perl recognizes it as the POSIX cl
  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
  I think.
+ A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
+ It seems that the appearance of a nested POSIX class supersedes an apparent
+ external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
+ a digit.
+ In Perl, unescaped square brackets may also appear as part of class names. For
+ example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
+ [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
+ seem right at all. PCRE does not allow closing square brackets in POSIX class
+ names.
  Arguments:
    ptr      pointer to the initial [
    endptr   where to return the end pointer
-Line 2295 
 Returns:   TRUE or FALSE
+Line 2669 
 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. */
  for (++ptr; *ptr != 0; ptr++)
    {
-   if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
+   if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
+     ptr++;
+   else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
+   else
      {
-     if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
        {
        *endptr = ptr;
        return TRUE;
        }
+     if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
+          (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
+           ptr[1] == CHAR_EQUALS_SIGN) &&
+         check_posix_syntax(ptr, endptr))
+       return FALSE;
      }
    }
  return FALSE;
-Line 2332 
 Returns:     a value representing the na
+Line 2713 
 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 2371 
 value in the reference (which is a group
+Line 2752 
 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 2379 
 Returns:     nothing
+Line 2760 
 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 2432 
 Arguments:
+Line 2813 
 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 2461 
 Returns:             nothing
+Line 2842 
 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 2544 
 switch(ptype)
+Line 2925 
 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 2555 
 switch(ptype)
+Line 2936 
 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 2590 
 sense to automatically possessify the re
+Line 2971 
 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 2599 
 Returns:        TRUE if possessifying is
+Line 2980 
 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 2611 
 if ((options & PCRE_EXTENDED) != 0)
+Line 2992 
 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 2619 
 if ((options & PCRE_EXTENDED) != 0)
+Line 3000 
 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 2638 
 if (*ptr == CHAR_BACKSLASH)
+Line 3019 
 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 2655 
 if ((options & PCRE_EXTENDED) != 0)
+Line 3034 
 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 2663 
 if ((options & PCRE_EXTENDED) != 0)
+Line 3042 
 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 2675 
 if ((options & PCRE_EXTENDED) != 0)
+Line 3054 
 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 2684 
 the next item is a character. */
+Line 3063 
 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 2696 
 if (next >= 0) switch(op_code)
+Line 3075 
 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 2715 
 if (next >= 0) switch(op_code)
+Line 3094 
 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((unsigned int)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
-   an XCLASS instead. */
    case OP_NOT:
-   return (c = *previous) == next;
+ #ifdef SUPPORT_UTF
+   GETCHARTEST(c, previous);
+ #else
+   c = *previous;
+ #endif
+   return c == next;
    case OP_NOTI:
-   if ((c = *previous) == next) return TRUE;
+ #ifdef SUPPORT_UTF
- #ifdef SUPPORT_UTF8
+   GETCHARTEST(c, previous);
-   if (utf8)
+ #else
+   c = *previous;
+ #endif
+   if (c == next) return TRUE;
+ #ifdef SUPPORT_UTF
+   if (utf)
      {
      unsigned int othercase;
      if (next < 128) othercase = cd->fcc[next]; else
  #ifdef SUPPORT_UCP
-     othercase = UCD_OTHERCASE(next);
+     othercase = UCD_OTHERCASE((unsigned int)next);
  #else
      othercase = NOTACHAR;
  #endif
      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((unsigned int)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 2832 
 switch(op_code)
+Line 3217 
 switch(op_code)
    {
    case OP_CHAR:
    case OP_CHARI:
- #ifdef SUPPORT_UTF8
+ #ifdef SUPPORT_UTF
    GETCHARTEST(c, previous);
  #else
    c = *previous;
-Line 2937 
 switch(op_code)
+Line 3322 
 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 3015 
 Arguments:
+Line 3400 
 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.
    lengthptr      NULL during the real compile phase
                   points to length accumulator during pre-compile phase
-Line 3027 
 Returns:         TRUE on success
+Line 3413 
 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,
+   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;
+ 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];
- #ifdef SUPPORT_UTF8
+ /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
- BOOL class_utf8;
+ must not do this for other options (e.g. PCRE_EXTENDED) because they may change
- BOOL utf8 = (options & PCRE_UTF8) != 0;
+ dynamically as we process the pattern. */
- uschar *class_utf8data;
- uschar *class_utf8data_base;
+ #ifdef SUPPORT_UTF
- uschar utf8_char[6];
+ /* PCRE_UTF16 has the same value as PCRE_UTF8. */
+ BOOL utf = (options & PCRE_UTF8) != 0;
+ pcre_uchar utf_chars[6];
  #else
- BOOL utf8 = FALSE;
+ BOOL utf = FALSE;
- uschar *utf8_char = NULL;
+ #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 3078 
 greedy_non_default = greedy_default ^ 1;
+Line 3474 
 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 3105 
 for (;; ptr++)
+Line 3502 
 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;
-   uschar mcbuffer[8];
+   int tempbracount;
+   pcre_uchar mcbuffer[8];
-   /* Get next byte in the pattern */
+   /* Get next character in the pattern */
    c = *ptr;
-Line 3139 
 for (;; ptr++)
+Line 3537 
 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 3162 
 for (;; ptr++)
+Line 3561 
 for (;; ptr++)
        }
      *lengthptr += (int)(code - last_code);
-     DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
+     DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
+       (int)(code - last_code), c, c));
      /* If "previous" is set and it is not at the start of the work space, move
      it back to there, in order to avoid filling up the work space. Otherwise,
-Line 3172 
 for (;; ptr++)
+Line 3572 
 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 3188 
 for (;; ptr++)
+Line 3588 
 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 3236 
 for (;; ptr++)
+Line 3637 
 for (;; ptr++)
      previous_callout = NULL;
      }
-   /* In extended mode, skip white space and comments */
+   /* In extended mode, skip white space and comments. */
    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 3248 
 for (;; ptr++)
+Line 3649 
 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 3273 
 for (;; ptr++)
+Line 3674 
 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 3298 
 for (;; ptr++)
+Line 3699 
 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 3310 
 for (;; ptr++)
+Line 3711 
 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 3370 
 for (;; ptr++)
+Line 3771 
 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 3390 
 for (;; ptr++)
+Line 3790 
 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 3401 
 for (;; ptr++)
+Line 3801 
 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 3427 
 for (;; ptr++)
+Line 3828 
 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 3473 
 for (;; ptr++)
+Line 3875 
 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 3529 
 for (;; ptr++)
+Line 3931 
 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 3560 
 for (;; ptr++)
+Line 3962 
 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 3578 
 for (;; ptr++)
+Line 3984 
 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 3591 
 for (;; ptr++)
+Line 4002 
 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 3605 
 for (;; ptr++)
+Line 4019 
 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 3646 
 for (;; ptr++)
+Line 4060 
 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 3680 
 for (;; ptr++)
+Line 4109 
 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 3716 
 for (;; ptr++)
+Line 4172 
 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 3744 
 for (;; ptr++)
+Line 4205 
 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 3766 
 for (;; ptr++)
+Line 4240 
 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 3785 
 for (;; ptr++)
+Line 4259 
 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 3840 
 for (;; ptr++)
+Line 4315 
 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 3885 
 for (;; ptr++)
+Line 4360 
 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 3926 
 for (;; ptr++)
+Line 4415 
 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 3941 
 for (;; ptr++)
+Line 4430 
 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 3976 
 for (;; ptr++)
+Line 4501 
 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 3990 
 for (;; ptr++)
+Line 4515 
 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++;
+       /* If class_charcount is 1, we saw precisely one character. As long as
+       there was no use of \p or \P, in other words, no use of any XCLASS
+       features, we can optimize.
+       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.
+       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_UTF8
+       if (class_single_char == 1 && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
-       if (utf8 && (c > 255 || ((options & PCRE_CASELESS) != 0 && c > 127)))
          {
-         class_utf8 = TRUE;
+         ptr++;
-         *class_utf8data++ = XCL_SINGLE;
+         zeroreqchar = reqchar;
-         class_utf8data += _pcre_ord2utf8(c, class_utf8data);
+         if (negate_class)
+           {
+           if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
+           zerofirstchar = firstchar;
+           *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
+ #ifdef SUPPORT_UTF
+           if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
+             code += PRIV(ord2utf)(c, code);
+           else
+ #endif
+             *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)
+           if ((int)(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 4045 
 for (;; ptr++)
+Line 4639 
 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 4114 
 for (;; ptr++)
+Line 4655 
 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 4148 
 for (;; ptr++)
+Line 4694 
 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 4194 
 for (;; ptr++)
+Line 4738 
 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 4205 
 for (;; ptr++)
+Line 4749 
 for (;; ptr++)
      op_type = 0;                    /* Default single-char op codes */
      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
-     /* Save start of previous item, in case we have to move it up to make space
+     /* Save start of previous item, in case we have to move it up in order to
-     for an inserted OP_ONCE for the additional '+' extension. */
+     insert something before it. */
      tempcode = previous;
-Line 4228 
 for (;; ptr++)
+Line 4772 
 for (;; ptr++)
        ptr++;
        }
      else repeat_type = greedy_default;
      /* If previous was a recursion call, wrap it in atomic brackets so that
      previous becomes the atomic group. All recursions were so wrapped in the
      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;
        PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
        code += 2 + 2 * LINK_SIZE;
        length_prevgroup = 3 + 3*LINK_SIZE;
        /* When actually compiling, we need to check whether this was a forward
        reference, and if so, adjust the offset. */
        if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
          {
          int offset = GET(cd->hwm, -LINK_SIZE);
          if (offset == previous + 1 - cd->start_code)
            PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
          }
        }
      /* Now handle repetition for the different types of item. */
-     /* If previous was a character match, abolish the item and generate a
+     /* If previous was a character or negated character match, abolish the item
-     repeat item instead. If a char item has a minumum of more than one, ensure
+     and generate a repeat item instead. If a char item has a minimum of more
-     that it is set in reqbyte - it might not be if a sequence such as x{3} is
+     than one, ensure that it is set in reqchar - it might not be if a sequence
-     the first thing in a branch because the x will have gone into firstbyte
+     such as x{3} is the first thing in a branch because the x will have gone
-     instead.  */
+     into firstchar instead.  */
-     if (*previous == OP_CHAR || *previous == OP_CHARI)
+     if (*previous == OP_CHAR || *previous == OP_CHARI
+         || *previous == OP_NOT || *previous == OP_NOTI)
        {
-       op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
+       switch (*previous)
+         {
+         default: /* Make compiler happy. */
+         case OP_CHAR:  op_type = OP_STAR - OP_STAR; break;
+         case OP_CHARI: op_type = OP_STARI - OP_STAR; break;
+         case OP_NOT:   op_type = OP_NOTSTAR - OP_STAR; break;
+         case OP_NOTI:  op_type = OP_NOTSTARI - OP_STAR; break;
+         }
-       /* 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 (*previous <= OP_CHARI && 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 4300 
 for (;; ptr++)
+Line 4852 
 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 4309 
 for (;; ptr++)
+Line 4861 
 for (;; ptr++)
        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
        }
-     /* If previous was a single negated character ([^a] or similar), we use
-     one of the special opcodes, replacing it. The code is shared with single-
-     character repeats by setting opt_type to add a suitable offset into
-     repeat_type. We can also test for auto-possessification. OP_NOT and OP_NOTI
-     are currently used only for single-byte chars. */
-     else if (*previous == OP_NOT || *previous == OP_NOTI)
-       {
-       op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
-       c = previous[1];
-       if (!possessive_quantifier &&
-           repeat_max < 0 &&
-           check_auto_possessive(previous, utf8, ptr + 1, options, cd))
-         {
-         repeat_type = 0;    /* Force greedy */
-         possessive_quantifier = TRUE;
-         }
-       goto OUTPUT_SINGLE_REPEAT;
-       }
      /* If previous was a character type match (\d or similar), abolish it and
      create a suitable repeat item. The code is shared with single-character
      repeats by setting op_type to add a suitable offset into repeat_type. Note
-Line 4338 
 for (;; ptr++)
+Line 4870 
 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 4425 
 for (;; ptr++)
+Line 4957 
 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 4454 
 for (;; ptr++)
+Line 4986 
 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 4484 
 for (;; ptr++)
+Line 5016 
 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 4511 
 for (;; ptr++)
+Line 5043 
 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 4549 
 for (;; ptr++)
+Line 5081 
 for (;; ptr++)
        }
      /* If previous was a bracket group, we may have to replicate it in certain
-     cases. Note that at this point we can encounter only the "basic" BRA and
+     cases. Note that at this point we can encounter only the "basic" bracket
-     KET opcodes, as this is the place where they get converted into the more
+     opcodes such as BRA and CBRA, as this is the place where they get converted
-     special varieties. */
+     into the more special varieties such as BRAPOS and SBRA. A test for >=
+     OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
+     ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
+     repetition of assertions, but now it does, for Perl compatibility. */
-     else if (*previous == OP_BRA  || *previous == OP_CBRA ||
+     else if (*previous >= OP_ASSERT && *previous <= OP_COND)
-              *previous == OP_ONCE || *previous == OP_COND)
        {
        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 */
+       /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
+       we just ignore the repeat. */
        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
+         goto END_REPEAT;
+       /* There is no sense in actually repeating assertions. The only potential
+       use of repetition is in cases when the assertion is optional. Therefore,
+       if the minimum is greater than zero, just ignore the repeat. If the
+       maximum is not not zero or one, set it to 1. */
+       if (*previous < OP_ONCE)    /* Assertion */
          {
-         *errorcodeptr = ERR55;
+         if (repeat_min > 0) goto END_REPEAT;
-         goto FAILED;
+         if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
          }
        /* The case of a zero minimum is special because of the need to stick
-Line 4587 
 for (;; ptr++)
+Line 5130 
 for (;; ptr++)
          **   goto END_REPEAT;
          **   }
-         However, that fails when a group is referenced as a subroutine from
+         However, that fails when a group or a subgroup within it is referenced
-         elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
+         as a subroutine from elsewhere in the pattern, so now we stick in
-         so that it is skipped on execution. As we don't have a list of which
+         OP_SKIPZERO in front of it so that it is skipped on execution. As we
-         groups are referenced, we cannot do this selectively.
+         don't have a list of which groups are referenced, we cannot do this
+         selectively.
          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
          and do no more at this point. However, we do need to adjust any
-Line 4602 
 for (;; ptr++)
+Line 5146 
 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 4626 
 for (;; ptr++)
+Line 5170 
 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 4673 
 for (;; ptr++)
+Line 5217 
 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 4732 
 for (;; ptr++)
+Line 5292 
 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 4749 
 for (;; ptr++)
+Line 5309 
 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 4766 
 for (;; ptr++)
+Line 5341 
 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 4781 
 for (;; ptr++)
+Line 5356 
 for (;; ptr++)
        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
        deal with possessive ONCEs specially.
-       Otherwise, if the quantifier was possessive, we convert the BRA code to
+       Otherwise, when we are doing the actual compile phase, check to see
-       the POS form, and the KET code to KETRPOS. (It turns out to be convenient
+       whether this group is one that could match an empty string. If so,
-       at runtime to detect this kind of subpattern at both the start and at the
+       convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
-       end.) The use of special opcodes makes it possible to reduce greatly the
+       that runtime checking can be done. [This check is also applied to ONCE
-       stack usage in pcre_exec(). If the group is preceded by OP_BRAZERO,
+       groups at runtime, but in a different way.]
-       convert this to OP_BRAPOSZERO. Then cancel the possessive flag so that
-       the default action below, of wrapping everything inside atomic brackets,
+       Then, if the quantifier was possessive and the bracket is not a
-       does not happen.
+       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
-       Then, when we are doing the actual compile phase, check to see whether
+       subpattern at both the start and at the end.) The use of special opcodes
-       this group is one that could match an empty string. If so, convert the
+       makes it possible to reduce greatly the stack usage in pcre_exec(). If
-       initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so that runtime
+       the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
-       checking can be done. [This check is also applied to ONCE groups at
-       runtime, but in a different way.] */
+       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);
-         if (*bracode == OP_ONCE && possessive_quantifier) *bracode = OP_BRA;
+         /* Convert possessive ONCE brackets to non-capturing */
-         if (*bracode == OP_ONCE)
+         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
            {
-           if (possessive_quantifier)
+           /* In the compile phase, check for empty string matching. */
-             {
-             *bracode += 1;                   /* Switch to xxxPOS opcodes */
-             *ketcode = OP_KETRPOS;
-             if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
-             possessive_quantifier = FALSE;
-             }
-           else *ketcode = OP_KETRMAX + repeat_type;
            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 4829 
 for (;; ptr++)
+Line 5412 
 for (;; ptr++)
                }
              while (*scode == OP_ALT);
              }
+           /* Handle possessive quantifiers. */
+           if (possessive_quantifier)
+             {
+             /* For COND brackets, we wrap the whole thing in a possessively
+             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, utf, cd, save_hwm);
+               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 4853 
 for (;; ptr++)
+Line 5478 
 for (;; ptr++)
      there are special alternative opcodes for this case. For anything else, we
      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
      notation is just syntactic sugar, taken from Sun's Java package, but the
      special opcodes can optimize it.
-     Possessively repeated subpatterns have already been handled in the code
+     Some (but not all) possessively repeated subpatterns have already been
-     just above, so possessive_quantifier is always FALSE for them at this
+     completely handled in the code just above. For them, possessive_quantifier
-     stage.
+     is always FALSE at this stage.
      Note that the repeated item starts at tempcode, not at previous, which
      might be the first part of a string whose (former) last char we repeated.
-Line 4871 
 for (;; ptr++)
+Line 5496 
 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 4916 
 for (;; ptr++)
+Line 5542 
 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 4929 
 for (;; ptr++)
+Line 5555 
 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 4952 
 for (;; ptr++)
+Line 5578 
 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
+       a closing parenthesis, to appear in arguments, so we no longer insist on
+       letters, digits, and underscores. */
        if (*ptr == CHAR_COLON)
          {
          arg = ++ptr;
-         while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
+         while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
-           || *ptr == '_') ptr++;
          arglen = (int)(ptr - arg);
          }
-Line 4983 
 for (;; ptr++)
+Line 5614 
 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. */
            if (verbs[i].op == OP_ACCEPT)
-Line 4995 
 for (;; ptr++)
+Line 5626 
 for (;; ptr++)
                {
                *errorcodeptr = ERR59;
                goto FAILED;
                }
              cd->had_accept = TRUE;
              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
                {
-Line 5003 
 for (;; ptr++)
+Line 5634 
 for (;; ptr++)
                PUT2INC(code, 0, oc->number);
                }
              *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
+             /* Do not set firstchar after *ACCEPT */
+             if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
              }
            /* Handle other cases with/without an argument */
-Line 5015 
 for (;; ptr++)
+Line 5649 
 for (;; ptr++)
                goto FAILED;
                }
              *code = verbs[i].op;
-             if (*code++ == OP_THEN)
+             if (*code++ == OP_THEN) cd->external_flags |= PCRE_HASTHEN;
-               {
-               PUT(code, 0, code - bcptr->current_branch - 1);
-               code += LINK_SIZE;
-               }
              }
            else
-Line 5030 
 for (;; ptr++)
+Line 5660 
 for (;; ptr++)
                goto FAILED;
                }
              *code = verbs[i].op_arg;
-             if (*code++ == OP_THEN_ARG)
+             if (*code++ == OP_THEN_ARG) cd->external_flags |= PCRE_HASTHEN;
-               {
-               PUT(code, 0, code - bcptr->current_branch - 1);
-               code += LINK_SIZE;
-               }
              *code++ = arglen;
-             memcpy(code, arg, arglen);
+             memcpy(code, arg, IN_UCHARS(arglen));
              code += arglen;
              *code++ = 0;
              }