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

Diff of /code/trunk/pcre_compile.c

Parent Directory | Revision Log | View Patch Patch

-revision 176 by ph10,
Mon Jun 11 13:48:37 2007 UTC
+revision 276 by ph10,
Wed Nov 21 15:39:20 2007 UTC
 Line 42 
 POSSIBILITY OF SUCH DAMAGE.
  supporting internal functions that are not used by other modules. */
+ #ifdef HAVE_CONFIG_H
+ #include "config.h"
+ #endif
  #define NLBLOCK cd             /* Block containing newline information */
  #define PSSTART start_pattern  /* Field containing processed string start */
  #define PSEND   end_pattern    /* Field containing processed string end */
  #include "pcre_internal.h"
-Line 58 
 used by pcretest. DEBUG is not defined w
+Line 61 
 used by pcretest. DEBUG is not defined w
  #endif
+ /* Macro for setting individual bits in class bitmaps. */
+ #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
+ /* Maximum length value to check against when making sure that the integer that
+ holds the compiled pattern length does not overflow. We make it a bit less than
+ INT_MAX to allow for adding in group terminating bytes, so that we don't have
+ to check them every time. */
+ #define OFLOW_MAX (INT_MAX - 20)
  /*************************************************
  *      Code parameters and static tables         *
  *************************************************/
-Line 87 
 static const short int escapes[] = {
+Line 102 
 static const short int escapes[] = {
 ,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */
 ,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */
     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */
-,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */
+ -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */
- -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */
+ -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */
  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */
     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */
-,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */
+ -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */
- -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */
+ -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */
 ,      0, -ESC_z                                            /* x - z */
  };
-Line 106 
 static const short int escapes[] = {
+Line 121 
 static const short int escapes[] = {
  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,
- /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,
+ /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
- /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,
+ /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
- /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,
+ /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
- /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,
+ /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,
  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
- /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,
+ /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
-Line 125 
 static const short int escapes[] = {
+Line 140 
 static const short int escapes[] = {
  #endif
- /* Tables of names of POSIX character classes and their lengths. The list is
+ /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
- terminated by a zero length entry. The first three must be alpha, lower, upper,
+ searched linearly. Put all the names into a single string, in order to reduce
- as this is assumed for handling case independence. */
+ the number of relocations when a shared library is dynamically linked. */
- static const char *const posix_names[] = {
+ typedef struct verbitem {
-   "alpha", "lower", "upper",
+   int   len;
-   "alnum", "ascii", "blank", "cntrl", "digit", "graph",
+   int   op;
-   "print", "punct", "space", "word",  "xdigit" };
+ } verbitem;
+ static const char verbnames[] =
+   "ACCEPT\0"
+   "COMMIT\0"
+   "F\0"
+   "FAIL\0"
+   "PRUNE\0"
+   "SKIP\0"
+   "THEN";
+ static verbitem verbs[] = {
+   { 6, OP_ACCEPT },
+   { 6, OP_COMMIT },
+   { 1, OP_FAIL },
+   { 4, OP_FAIL },
+   { 5, OP_PRUNE },
+   { 4, OP_SKIP  },
+   { 4, OP_THEN  }
+ };
+ static int verbcount = sizeof(verbs)/sizeof(verbitem);
+ /* Tables of names of POSIX character classes and their lengths. The names are
+ now all in a single string, to reduce the number of relocations when a shared
+ library is dynamically loaded. The list of lengths is terminated by a zero
+ length entry. The first three must be alpha, lower, upper, as this is assumed
+ for handling case independence. */
+ static const char posix_names[] =
+   "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"
+   "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"
+   "word\0"   "xdigit";
  static const uschar posix_name_lengths[] = {
 , 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
-Line 171 
 static const int posix_class_maps[] = {
+Line 219 
 static const int posix_class_maps[] = {
  /* The texts of compile-time error messages. These are "char *" because they
  are passed to the outside world. Do not ever re-use any error number, because
  they are documented. Always add a new error instead. Messages marked DEAD below
- are no longer used. */
+ are no longer used. This used to be a table of strings, but in order to reduce
+ the number of relocations needed when a shared library is loaded dynamically,
- static const char *error_texts[] = {
+ it is now one long string. We cannot use a table of offsets, because the
-   "no error",
+ lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
-   "\\ at end of pattern",
+ simply count through to the one we want - this isn't a performance issue
-   "\\c at end of pattern",
+ because these strings are used only when there is a compilation error. */
-   "unrecognized character follows \\",
-   "numbers out of order in {} quantifier",
+ static const char error_texts[] =
+   "no error\0"
+   "\\ at end of pattern\0"
+   "\\c at end of pattern\0"
+   "unrecognized character follows \\\0"
+   "numbers out of order in {} quantifier\0"
    /* 5 */
-   "number too big in {} quantifier",
+   "number too big in {} quantifier\0"
-   "missing terminating ] for character class",
+   "missing terminating ] for character class\0"
-   "invalid escape sequence in character class",
+   "invalid escape sequence in character class\0"
-   "range out of order in character class",
+   "range out of order in character class\0"
-   "nothing to repeat",
+   "nothing to repeat\0"
    /* 10 */
-   "operand of unlimited repeat could match the empty string",  /** DEAD **/
+   "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
-   "internal error: unexpected repeat",
+   "internal error: unexpected repeat\0"
-   "unrecognized character after (?",
+   "unrecognized character after (? or (?-\0"
-   "POSIX named classes are supported only within a class",
+   "POSIX named classes are supported only within a class\0"
-   "missing )",
+   "missing )\0"
    /* 15 */
-   "reference to non-existent subpattern",
+   "reference to non-existent subpattern\0"
-   "erroffset passed as NULL",
+   "erroffset passed as NULL\0"
-   "unknown option bit(s) set",
+   "unknown option bit(s) set\0"
-   "missing ) after comment",
+   "missing ) after comment\0"
-   "parentheses nested too deeply",  /** DEAD **/
+   "parentheses nested too deeply\0"  /** DEAD **/
    /* 20 */
-   "regular expression too large",
+   "regular expression is too large\0"
-   "failed to get memory",
+   "failed to get memory\0"
-   "unmatched parentheses",
+   "unmatched parentheses\0"
-   "internal error: code overflow",
+   "internal error: code overflow\0"
-   "unrecognized character after (?<",
+   "unrecognized character after (?<\0"
    /* 25 */
-   "lookbehind assertion is not fixed length",
+   "lookbehind assertion is not fixed length\0"
-   "malformed number or name after (?(",
+   "malformed number or name after (?(\0"
-   "conditional group contains more than two branches",
+   "conditional group contains more than two branches\0"
-   "assertion expected after (?(",
+   "assertion expected after (?(\0"
-   "(?R or (?[+-]digits must be followed by )",
+   "(?R or (?[+-]digits must be followed by )\0"
    /* 30 */
-   "unknown POSIX class name",
+   "unknown POSIX class name\0"
-   "POSIX collating elements are not supported",
+   "POSIX collating elements are not supported\0"
-   "this version of PCRE is not compiled with PCRE_UTF8 support",
+   "this version of PCRE is not compiled with PCRE_UTF8 support\0"
-   "spare error",  /** DEAD **/
+   "spare error\0"  /** DEAD **/
-   "character value in \\x{...} sequence is too large",
+   "character value in \\x{...} sequence is too large\0"
    /* 35 */
-   "invalid condition (?(0)",
+   "invalid condition (?(0)\0"
-   "\\C not allowed in lookbehind assertion",
+   "\\C not allowed in lookbehind assertion\0"
-   "PCRE does not support \\L, \\l, \\N, \\U, or \\u",
+   "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"
-   "number after (?C is > 255",
+   "number after (?C is > 255\0"
-   "closing ) for (?C expected",
+   "closing ) for (?C expected\0"
    /* 40 */
-   "recursive call could loop indefinitely",
+   "recursive call could loop indefinitely\0"
-   "unrecognized character after (?P",
+   "unrecognized character after (?P\0"
-   "syntax error in subpattern name (missing terminator)",
+   "syntax error in subpattern name (missing terminator)\0"
-   "two named subpatterns have the same name",
+   "two named subpatterns have the same name\0"
-   "invalid UTF-8 string",
+   "invalid UTF-8 string\0"
    /* 45 */
-   "support for \\P, \\p, and \\X has not been compiled",
+   "support for \\P, \\p, and \\X has not been compiled\0"
-   "malformed \\P or \\p sequence",
+   "malformed \\P or \\p sequence\0"
-   "unknown property name after \\P or \\p",
+   "unknown property name after \\P or \\p\0"
-   "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",
+   "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
-   "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",
+   "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
    /* 50 */
-   "repeated subpattern is too long",
+   "repeated subpattern is too long\0"    /** DEAD **/
-   "octal value is greater than \\377 (not in UTF-8 mode)",
+   "octal value is greater than \\377 (not in UTF-8 mode)\0"
-   "internal error: overran compiling workspace",
+   "internal error: overran compiling workspace\0"
-   "internal error: previously-checked referenced subpattern not found",
+   "internal error: previously-checked referenced subpattern not found\0"
-   "DEFINE group contains more than one branch",
+   "DEFINE group contains more than one branch\0"
    /* 55 */
-   "repeating a DEFINE group is not allowed",
+   "repeating a DEFINE group is not allowed\0"
-   "inconsistent NEWLINE options",
+   "inconsistent NEWLINE options\0"
-   "\\g is not followed by a braced name or an optionally braced non-zero number",
+   "\\g is not followed by a braced name or an optionally braced non-zero number\0"
-   "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number"
+   "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number\0"
- };
+   "(*VERB) with an argument is not supported\0"
+   /* 60 */
+   "(*VERB) not recognized\0"
+   "number is too big\0"
+   "subpattern name expected\0"
+   "digit expected after (?+";
  /* Table to identify digits and hex digits. This is used when compiling
-Line 374 
 static const unsigned char ebcdic_charta
+Line 432 
 static const unsigned char ebcdic_charta
  /* Definition to allow mutual recursion */
  static BOOL
    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
      int *, int *, branch_chain *, compile_data *, int *);
  /*************************************************
+ *            Find an error text                  *
+ *************************************************/
+ /* The error texts are now all in one long string, to save on relocations. As
+ some of the text is of unknown length, we can't use a table of offsets.
+ Instead, just count through the strings. This is not a performance issue
+ because it happens only when there has been a compilation error.
+ Argument:   the error number
+ Returns:    pointer to the error string
+ */
+ static const char *
+ find_error_text(int n)
+ {
+ const char *s = error_texts;
+ for (; n > 0; n--) while (*s++ != 0);
+ return s;
+ }
+ /*************************************************
  *            Handle escapes                      *
  *************************************************/
-Line 400 
 Arguments:
+Line 480 
 Arguments:
  Returns:         zero or positive => a data character
                   negative => a special escape sequence
-                  on error, errorptr is set
+                  on error, errorcodeptr is set
  */
  static int
-Line 418 
 ptr--;                            /* Set
+Line 498 
 ptr--;                            /* Set
  if (c == 0) *errorcodeptr = ERR1;
- /* Non-alphamerics are literals. For digits or letters, do an initial lookup in
+ /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
- a table. A non-zero result is something that can be returned immediately.
+ in a table. A non-zero result is something that can be returned immediately.
  Otherwise further processing may be required. */
  #ifndef EBCDIC  /* ASCII coding */
- else if (c < '0' || c > 'z') {}                           /* Not alphameric */
+ else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */
  else if ((i = escapes[c - '0']) != 0) c = i;
  #else           /* EBCDIC coding */
- else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */
+ else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
  else if ((i = escapes[c - 0x48]) != 0)  c = i;
  #endif
-Line 485 
 else
+Line 565 
 else
      while ((digitab[ptr[1]] & ctype_digit) != 0)
        c = c * 10 + *(++ptr) - '0';
+     if (c < 0)
+       {
+       *errorcodeptr = ERR61;
+       break;
+       }
      if (c == 0 || (braced && *(++ptr) != '}'))
        {
        *errorcodeptr = ERR57;
-       return 0;
+       break;
        }
      if (negated)
-Line 496 
 else
+Line 582 
 else
        if (c > bracount)
          {
          *errorcodeptr = ERR15;
-         return 0;
+         break;
          }
        c = bracount - (c - 1);
        }
-Line 525 
 else
+Line 611 
 else
        c -= '0';
        while ((digitab[ptr[1]] & ctype_digit) != 0)
          c = c * 10 + *(++ptr) - '0';
+       if (c < 0)
+         {
+         *errorcodeptr = ERR61;
+         break;
+         }
        if (c < 10 || c <= bracount)
          {
          c = -(ESC_REF + c);
-Line 620 
 else
+Line 711 
 else
      if (c == 0)
        {
        *errorcodeptr = ERR2;
-       return 0;
+       break;
        }
  #ifndef EBCDIC  /* ASCII coding */
-Line 633 
 else
+Line 724 
 else
      break;
      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
-     other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,
+     other alphanumeric following \ is an error if PCRE_EXTRA was set;
-     for Perl compatibility, it is a literal. This code looks a bit odd, but
+     otherwise, for Perl compatibility, it is a literal. This code looks a bit
-     there used to be some cases other than the default, and there may be again
+     odd, but there used to be some cases other than the default, and there may
-     in future, so I haven't "optimized" it. */
+     be again in future, so I haven't "optimized" it. */
      default:
      if ((options & PCRE_EXTRA) != 0) switch(c)
-Line 696 
 if (c == '{')
+Line 787 
 if (c == '{')
      *negptr = TRUE;
      ptr++;
      }
-   for (i = 0; i < sizeof(name) - 1; i++)
+   for (i = 0; i < (int)sizeof(name) - 1; i++)
      {
      c = *(++ptr);
      if (c == 0) goto ERROR_RETURN;
-Line 725 
 top = _pcre_utt_size;
+Line 816 
 top = _pcre_utt_size;
  while (bot < top)
    {
    i = (bot + top) >> 1;
-   c = strcmp(name, _pcre_utt[i].name);
+   c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
    if (c == 0)
      {
      *dptr = _pcre_utt[i].value;
-Line 899 
 for (; *ptr != 0; ptr++)
+Line 990 
 for (; *ptr != 0; ptr++)
      {
      while (*(++ptr) != ']')
        {
+       if (*ptr == 0) return -1;
        if (*ptr == '\\')
          {
          if (*(++ptr) == 0) return -1;
-Line 926 
 for (; *ptr != 0; ptr++)
+Line 1018 
 for (; *ptr != 0; ptr++)
    /* An opening parens must now be a real metacharacter */
    if (*ptr != '(') continue;
-   if (ptr[1] != '?')
+   if (ptr[1] != '?' && ptr[1] != '*')
      {
      count++;
      if (name == NULL && count == lorn) return count;
-Line 1054 
 for (;;)
+Line 1146 
 for (;;)
    {
    int d;
    register int op = *cc;
    switch (op)
      {
      case OP_CBRA:
-Line 1143 
 for (;;)
+Line 1234 
 for (;;)
      case OP_TYPEEXACT:
      branchlength += GET2(cc,1);
+     if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
      cc += 4;
      break;
-Line 1251 
 for (;;)
+Line 1343 
 for (;;)
      code += _pcre_OP_lengths[c];
      }
-   /* In UTF-8 mode, opcodes that are followed by a character may be followed by
+   /* Otherwise, we can get the item's length from the table, except that for
-   a multi-byte character. The length in the table is a minimum, so we have to
+   repeated character types, we have to test for \p and \P, which have an extra
-   arrange to skip the extra bytes. */
+   two bytes of parameters. */
    else
      {
+     switch(c)
+       {
+       case OP_TYPESTAR:
+       case OP_TYPEMINSTAR:
+       case OP_TYPEPLUS:
+       case OP_TYPEMINPLUS:
+       case OP_TYPEQUERY:
+       case OP_TYPEMINQUERY:
+       case OP_TYPEPOSSTAR:
+       case OP_TYPEPOSPLUS:
+       case OP_TYPEPOSQUERY:
+       if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
+       break;
+       case OP_TYPEUPTO:
+       case OP_TYPEMINUPTO:
+       case OP_TYPEEXACT:
+       case OP_TYPEPOSUPTO:
+       if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
+       break;
+       }
+     /* Add in the fixed length from the table */
      code += _pcre_OP_lengths[c];
+   /* In UTF-8 mode, opcodes that are followed by a character may be followed by
+   a multi-byte character. The length in the table is a minimum, so we have to
+   arrange to skip the extra bytes. */
  #ifdef SUPPORT_UTF8
      if (utf8) switch(c)
        {
-Line 1315 
 for (;;)
+Line 1436 
 for (;;)
    if (c == OP_XCLASS) code += GET(code, 1);
-   /* Otherwise, we get the item's length from the table. In UTF-8 mode, opcodes
+   /* Otherwise, we can get the item's length from the table, except that for
-   that are followed by a character may be followed by a multi-byte character.
+   repeated character types, we have to test for \p and \P, which have an extra
-   The length in the table is a minimum, so we have to arrange to skip the extra
+   two bytes of parameters. */
-   bytes. */
    else
      {
+     switch(c)
+       {
+       case OP_TYPESTAR:
+       case OP_TYPEMINSTAR:
+       case OP_TYPEPLUS:
+       case OP_TYPEMINPLUS:
+       case OP_TYPEQUERY:
+       case OP_TYPEMINQUERY:
+       case OP_TYPEPOSSTAR:
+       case OP_TYPEPOSPLUS:
+       case OP_TYPEPOSQUERY:
+       if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
+       break;
+       case OP_TYPEPOSUPTO:
+       case OP_TYPEUPTO:
+       case OP_TYPEMINUPTO:
+       case OP_TYPEEXACT:
+       if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
+       break;
+       }
+     /* Add in the fixed length from the table */
      code += _pcre_OP_lengths[c];
+     /* In UTF-8 mode, opcodes that are followed by a character may be followed
+     by a multi-byte character. The length in the table is a minimum, so we have
+     to arrange to skip the extra bytes. */
  #ifdef SUPPORT_UTF8
      if (utf8) switch(c)
        {
-Line 1394 
 for (code = first_significant_code(code
+Line 1543 
 for (code = first_significant_code(code
    /* For other groups, scan the branches. */
-   if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)
+   if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
      {
      BOOL empty_branch;
      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
-Line 1418 
 for (code = first_significant_code(code
+Line 1567 
 for (code = first_significant_code(code
    switch (c)
      {
-     /* Check for quantifiers after a class */
+     /* 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
+     actual length is stored in the compiled code, so we must update "code"
+     here. */
  #ifdef SUPPORT_UTF8
      case OP_XCLASS:
-     ccode = code + GET(code, 1);
+     ccode = code += GET(code, 1);
      goto CHECK_CLASS_REPEAT;
  #endif
-Line 1484 
 for (code = first_significant_code(code
+Line 1637 
 for (code = first_significant_code(code
      case OP_TYPEEXACT:
      return FALSE;
+     /* These are going to continue, as they may be empty, but we have to
+     fudge the length for the \p and \P cases. */
+     case OP_TYPESTAR:
+     case OP_TYPEMINSTAR:
+     case OP_TYPEPOSSTAR:
+     case OP_TYPEQUERY:
+     case OP_TYPEMINQUERY:
+     case OP_TYPEPOSQUERY:
+     if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
+     break;
+     /* Same for these */
+     case OP_TYPEUPTO:
+     case OP_TYPEMINUPTO:
+     case OP_TYPEPOSUPTO:
+     if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
+     break;
      /* End of branch */
      case OP_KET:
-Line 1600 
 Returns:     a value representing the na
+Line 1773 
 Returns:     a value representing the na
  static int
  check_posix_name(const uschar *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, posix_names[yield], len) == 0) return yield;
+     strncmp((const char *)ptr, pn, len) == 0) return yield;
+   pn += posix_name_lengths[yield] + 1;
    yield++;
    }
  return -1;
-Line 1646 
 adjust_recurse(uschar *group, int adjust
+Line 1821 
 adjust_recurse(uschar *group, int adjust
    uschar *save_hwm)
  {
  uschar *ptr = group;
  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
    {
    int offset;
-Line 1944 
 if (next >= 0) switch(op_code)
+Line 2120 
 if (next >= 0) switch(op_code)
    case OP_NOT_WORDCHAR:
    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
+   case OP_HSPACE:
+   case OP_NOT_HSPACE:
+   switch(next)
+     {
+     case 0x09:
+     case 0x20:
+     case 0xa0:
+     case 0x1680:
+     case 0x180e:
+     case 0x2000:
+     case 0x2001:
+     case 0x2002:
+     case 0x2003:
+     case 0x2004:
+     case 0x2005:
+     case 0x2006:
+     case 0x2007:
+     case 0x2008:
+     case 0x2009:
+     case 0x200A:
+     case 0x202f:
+     case 0x205f:
+     case 0x3000:
+     return op_code != OP_HSPACE;
+     default:
+     return op_code == OP_HSPACE;
+     }
+   case OP_VSPACE:
+   case OP_NOT_VSPACE:
+   switch(next)
+     {
+     case 0x0a:
+     case 0x0b:
+     case 0x0c:
+     case 0x0d:
+     case 0x85:
+     case 0x2028:
+     case 0x2029:
+     return op_code != OP_VSPACE;
+     default:
+     return op_code == OP_VSPACE;
+     }
    default:
    return FALSE;
    }
-Line 1978 
 switch(op_code)
+Line 2198 
 switch(op_code)
      case ESC_W:
      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
+     case ESC_h:
+     case ESC_H:
+     switch(item)
+       {
+       case 0x09:
+       case 0x20:
+       case 0xa0:
+       case 0x1680:
+       case 0x180e:
+       case 0x2000:
+       case 0x2001:
+       case 0x2002:
+       case 0x2003:
+       case 0x2004:
+       case 0x2005:
+       case 0x2006:
+       case 0x2007:
+       case 0x2008:
+       case 0x2009:
+       case 0x200A:
+       case 0x202f:
+       case 0x205f:
+       case 0x3000:
+       return -next != ESC_h;
+       default:
+       return -next == ESC_h;
+       }
+     case ESC_v:
+     case ESC_V:
+     switch(item)
+       {
+       case 0x0a:
+       case 0x0b:
+       case 0x0c:
+       case 0x0d:
+       case 0x85:
+       case 0x2028:
+       case 0x2029:
+       return -next != ESC_v;
+       default:
+       return -next == ESC_v;
+       }
      default:
      return FALSE;
      }
    case OP_DIGIT:
-   return next == -ESC_D || next == -ESC_s || next == -ESC_W;
+   return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
+          next == -ESC_h || next == -ESC_v;
    case OP_NOT_DIGIT:
    return next == -ESC_d;
-Line 1992 
 switch(op_code)
+Line 2257 
 switch(op_code)
    return next == -ESC_S || next == -ESC_d || next == -ESC_w;
    case OP_NOT_WHITESPACE:
-   return next == -ESC_s;
+   return next == -ESC_s || next == -ESC_h || next == -ESC_v;
+   case OP_HSPACE:
+   return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
+   case OP_NOT_HSPACE:
+   return next == -ESC_h;
+   /* Can't have \S in here because VT matches \S (Perl anomaly) */
+   case OP_VSPACE:
+   return next == -ESC_V || next == -ESC_d || next == -ESC_w;
+   case OP_NOT_VSPACE:
+   return next == -ESC_v;
    case OP_WORDCHAR:
-   return next == -ESC_W || next == -ESC_s;
+   return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
    case OP_NOT_WORDCHAR:
    return next == -ESC_w || next == -ESC_d;
-Line 2107 
 req_caseopt = ((options & PCRE_CASELESS)
+Line 2385 
 req_caseopt = ((options & PCRE_CASELESS)
  for (;; ptr++)
    {
    BOOL negate_class;
+   BOOL should_flip_negation;
    BOOL possessive_quantifier;
    BOOL is_quantifier;
    BOOL is_recurse;
    BOOL reset_bracount;
    int class_charcount;
    int class_lastchar;
    int newoptions;
-Line 2148 
 for (;; ptr++)
+Line 2427 
 for (;; ptr++)
      */
      if (code < last_code) code = last_code;
+     /* Paranoid check for integer overflow */
+     if (OFLOW_MAX - *lengthptr < code - last_code)
+       {
+       *errorcodeptr = ERR20;
+       goto FAILED;
+       }
      *lengthptr += code - last_code;
      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
-Line 2260 
 for (;; ptr++)
+Line 2548 
 for (;; ptr++)
      *ptrptr = ptr;
      if (lengthptr != NULL)
        {
+       if (OFLOW_MAX - *lengthptr < code - last_code)
+         {
+         *errorcodeptr = ERR20;
+         goto FAILED;
+         }
        *lengthptr += code - last_code;   /* To include callout length */
        DPRINTF((">> end branch\n"));
        }
-Line 2322 
 for (;; ptr++)
+Line 2615 
 for (;; ptr++)
        goto FAILED;
        }
-     /* If the first character is '^', set the negation flag and skip it. */
+     /* If the first character is '^', set the negation flag and skip it. Also,
+     if the first few characters (either before or after ^) are \Q\E or \E we
+     skip them too. This makes for compatibility with Perl. */
-     if ((c = *(++ptr)) == '^')
+     negate_class = FALSE;
+     for (;;)
        {
-       negate_class = TRUE;
        c = *(++ptr);
+       if (c == '\\')
+         {
+         if (ptr[1] == 'E') ptr++;
+           else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;
+             else break;
+         }
+       else if (!negate_class && c == '^')
+         negate_class = TRUE;
+       else break;
        }
-     else
-       {
+     /* If a class contains a negative special such as \S, we need to flip the
-       negate_class = FALSE;
+     negation flag at the end, so that support for characters > 255 works
-       }
+     correctly (they are all included in the class). */
+     should_flip_negation = FALSE;
      /* Keep a count of chars with values < 256 so that we can optimize the case
      of just a single character (as long as it's < 256). However, For higher
-Line 2406 
 for (;; ptr++)
+Line 2712 
 for (;; ptr++)
          if (*ptr == '^')
            {
            local_negate = TRUE;
+           should_flip_negation = TRUE;  /* Note negative special */
            ptr++;
            }
-Line 2472 
 for (;; ptr++)
+Line 2779 
 for (;; ptr++)
        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.
        Elsewhere it marks a word boundary. Other escapes have preset maps ready
-       to or into the one we are building. We assume they have more than one
+       to 'or' into the one we are building. We assume they have more than one
        character in them, so set class_charcount bigger than one. */
        if (c == '\\')
-Line 2480 
 for (;; ptr++)
+Line 2787 
 for (;; ptr++)
          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
          if (*errorcodeptr != 0) goto FAILED;
-         if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */
+         if (-c == ESC_b) c = '\b';       /* \b is backspace in a class */
          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */
          else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */
          else if (-c == ESC_Q)            /* Handle start of quoted string */
-Line 2492 
 for (;; ptr++)
+Line 2799 
 for (;; ptr++)
            else inescq = TRUE;
            continue;
            }
+         else if (-c == ESC_E) continue;  /* Ignore orphan \E */
          if (c < 0)
            {
-Line 2507 
 for (;; ptr++)
+Line 2815 
 for (;; ptr++)
              continue;
              case ESC_D:
+             should_flip_negation = TRUE;
              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
              continue;
-Line 2515 
 for (;; ptr++)
+Line 2824 
 for (;; ptr++)
              continue;
              case ESC_W:
+             should_flip_negation = TRUE;
              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
              continue;
-Line 2524 
 for (;; ptr++)
+Line 2834 
 for (;; ptr++)
              continue;
              case ESC_S:
+             should_flip_negation = TRUE;
              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
              continue;
-             case ESC_E: /* Perl ignores an orphan \E */
-             continue;
              default:    /* Not recognized; fall through */
              break;      /* Need "default" setting to stop compiler warning. */
              }
-Line 2540 
 for (;; ptr++)
+Line 2848 
 for (;; ptr++)
            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
+           /* We need to deal with \H, \h, \V, and \v in both phases because
+           they use extra memory. */
+           if (-c == ESC_h)
+             {
+             SETBIT(classbits, 0x09); /* VT */
+             SETBIT(classbits, 0x20); /* SPACE */
+             SETBIT(classbits, 0xa0); /* NSBP */
+ #ifdef SUPPORT_UTF8
+             if (utf8)
+               {
+               class_utf8 = TRUE;
+               *class_utf8data++ = XCL_SINGLE;
+               class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
+               *class_utf8data++ = XCL_SINGLE;
+               class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
+               *class_utf8data++ = XCL_RANGE;
+               class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
+               class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
+               *class_utf8data++ = XCL_SINGLE;
+               class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
+               *class_utf8data++ = XCL_SINGLE;
+               class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
+               *class_utf8data++ = XCL_SINGLE;
+               class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
+               }
+ #endif
+             continue;
+             }
+           if (-c == ESC_H)
+             {
+             for (c = 0; c < 32; c++)
+               {
+               int x = 0xff;
+               switch (c)
+                 {
+                 case 0x09/8: x ^= 1 << (0x09%8); break;
+                 case 0x20/8: x ^= 1 << (0x20%8); break;
+                 case 0xa0/8: x ^= 1 << (0xa0%8); break;
+                 default: break;
+                 }
+               classbits[c] |= x;
+               }
+ #ifdef SUPPORT_UTF8
+             if (utf8)
+               {
+               class_utf8 = TRUE;
+               *class_utf8data++ = XCL_RANGE;
+               class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
+               class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
+               *class_utf8data++ = XCL_RANGE;
+               class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
+               class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
+               *class_utf8data++ = XCL_RANGE;
+               class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
+               class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
+               *class_utf8data++ = XCL_RANGE;
+               class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
+               class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
+               *class_utf8data++ = XCL_RANGE;
+               class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
+               class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
+               *class_utf8data++ = XCL_RANGE;
+               class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
+               class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
+               *class_utf8data++ = XCL_RANGE;
+               class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
+               class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
+               }
+ #endif
+             continue;
+             }
+           if (-c == ESC_v)
+             {
+             SETBIT(classbits, 0x0a); /* LF */
+             SETBIT(classbits, 0x0b); /* VT */
+             SETBIT(classbits, 0x0c); /* FF */
+             SETBIT(classbits, 0x0d); /* CR */
+             SETBIT(classbits, 0x85); /* NEL */
+ #ifdef SUPPORT_UTF8
+             if (utf8)
+               {
+               class_utf8 = TRUE;
+               *class_utf8data++ = XCL_RANGE;
+               class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
+               class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
+               }
+ #endif
+             continue;
+             }
+           if (-c == ESC_V)
+             {
+             for (c = 0; c < 32; c++)
+               {
+               int x = 0xff;
+               switch (c)
+                 {
+                 case 0x0a/8: x ^= 1 << (0x0a%8);
+                              x ^= 1 << (0x0b%8);
+                              x ^= 1 << (0x0c%8);
+                              x ^= 1 << (0x0d%8);
+                              break;
+                 case 0x85/8: x ^= 1 << (0x85%8); break;
+                 default: break;
+                 }
+               classbits[c] |= x;
+               }
+ #ifdef SUPPORT_UTF8
+             if (utf8)
+               {
+               class_utf8 = TRUE;
+               *class_utf8data++ = XCL_RANGE;
+               class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
+               class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
+               *class_utf8data++ = XCL_RANGE;
+               class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
+               class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
+               }
+ #endif
+             continue;
+             }
            /* We need to deal with \P and \p in both phases. */
  #ifdef SUPPORT_UCP
-Line 2591 
 for (;; ptr++)
+Line 3026 
 for (;; ptr++)
        oldptr = ptr;
+       /* Remember \r or \n */
+       if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;
+       /* Check for range */
        if (!inescq && ptr[1] == '-')
          {
          int d;
-Line 2632 
 for (;; ptr++)
+Line 3073 
 for (;; ptr++)
            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
            if (*errorcodeptr != 0) goto FAILED;
-           /* \b is backslash; \X is literal X; \R is literal R; any other
+           /* \b is backspace; \X is literal X; \R is literal R; any other
            special means the '-' was literal */
            if (d < 0)
-Line 2658 
 for (;; ptr++)
+Line 3099 
 for (;; ptr++)
          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
+         /* Remember \r or \n */
+         if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;
          /* 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
-Line 2680 
 for (;; ptr++)
+Line 3125 
 for (;; ptr++)
              unsigned int origd = d;
              while (get_othercase_range(&cc, origd, &occ, &ocd))
                {
                if (occ >= (unsigned int)c &&
                    ocd <= (unsigned int)d)
                  continue;                          /* Skip embedded ranges */
                if (occ < (unsigned int)c  &&
                    ocd >= (unsigned int)c - 1)      /* Extend the basic range */
                  {                                  /* if there is overlap,   */
                  c = occ;                           /* noting that if occ < c */
                  continue;                          /* we can't have ocd > d  */
                  }                                  /* because a subrange is  */
                if (ocd > (unsigned int)d &&
                    occ <= (unsigned int)d + 1)      /* always shorter than    */
                  {                                  /* the basic range.       */
                  d = ocd;
-Line 2811 
 for (;; ptr++)
+Line 3256 
 for (;; ptr++)
        goto FAILED;
        }
+ /* This code has been disabled because it would mean that \s counts as
+ an explicit \r or \n reference, and that's not really what is wanted. Now
+ we set the flag only if there is a literal "\r" or "\n" in the class. */
+ #if 0
+     /* Remember whether \r or \n are in this class */
+     if (negate_class)
+       {
+       if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
+       }
+     else
+       {
+       if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
+       }
+ #endif
      /* If class_charcount is 1, we saw precisely one character whose value is
-     less than 256. In non-UTF-8 mode we can always optimize. In UTF-8 mode, we
+     less than 256. As long as there were no characters >= 128 and there was no
-     can optimize the negative case only if there were no characters >= 128
+     use of \p or \P, in other words, no use of any XCLASS features, we can
-     because OP_NOT and the related opcodes like OP_NOTSTAR operate on
+     optimize.
-     single-bytes only. This is an historical hangover. Maybe one day we can
-     tidy these opcodes to handle multi-byte characters.
+     In UTF-8 mode, we can optimize the negative case only if there were no
+     characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
+     operate on single-bytes 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[NC] if it's positive, or OP_NOT if it's negative. Note
-Line 2826 
 for (;; ptr++)
+Line 3293 
 for (;; ptr++)
      reqbyte, save the previous value for reinstating. */
  #ifdef SUPPORT_UTF8
-     if (class_charcount == 1 &&
+     if (class_charcount == 1 && !class_utf8 &&
-           (!utf8 ||
+       (!utf8 || !negate_class || class_lastchar < 128))
-           (!class_utf8 && (!negate_class || class_lastchar < 128))))
  #else
      if (class_charcount == 1)
  #endif
-Line 2872 
 for (;; ptr++)
+Line 3337 
 for (;; ptr++)
      zeroreqbyte = reqbyte;
      /* If there are characters with values > 255, we have to compile an
-     extended class, with its own opcode. If there are no characters < 256,
+     extended class, with its own opcode, unless there was a negated special
-     we can omit the bitmap in the actual compiled code. */
+     such as \S in the class, because in that case all characters > 255 are in
+     the class, so any that were explicitly given as well can be ignored. If
+     (when there are explicit characters > 255 that must be listed) there are no
+     characters < 256, we can omit the bitmap in the actual compiled code. */
  #ifdef SUPPORT_UTF8
-     if (class_utf8)
+     if (class_utf8 && !should_flip_negation)
        {
        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
        *code++ = OP_XCLASS;
-Line 2902 
 for (;; ptr++)
+Line 3370 
 for (;; ptr++)
        }
  #endif
-     /* If there are no characters > 255, negate the 32-byte map if necessary,
+     /* If there are no characters > 255, set the opcode to OP_CLASS or
-     and copy it into the code vector. If this is the first thing in the branch,
+     OP_NCLASS, depending on whether the whole class was negated and whether
-     there can be no first char setting, whatever the repeat count. Any reqbyte
+     there were negative specials such as \S in the class. Then copy the 32-byte
-     setting must remain unchanged after any kind of repeat. */
+     map into the code vector, negating it if necessary. */
+     *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
      if (negate_class)
        {
-       *code++ = OP_NCLASS;
        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
        }
      else
        {
-       *code++ = OP_CLASS;
        memcpy(code, classbits, 32);
        }
      code += 32;
-Line 3101 
 for (;; ptr++)
+Line 3568 
 for (;; ptr++)
        /* All real repeats make it impossible to handle partial matching (maybe
        one day we will be able to remove this restriction). */
-       if (repeat_max != 1) cd->nopartial = TRUE;
+       if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
        /* Combine the op_type with the repeat_type */
-Line 3251 
 for (;; ptr++)
+Line 3718 
 for (;; ptr++)
        /* All real repeats make it impossible to handle partial matching (maybe
        one day we will be able to remove this restriction). */
-       if (repeat_max != 1) cd->nopartial = TRUE;
+       if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
        if (repeat_min == 0 && repeat_max == -1)
          *code++ = OP_CRSTAR + repeat_type;
-Line 3287 
 for (;; ptr++)
+Line 3754 
 for (;; ptr++)
          goto FAILED;
          }
-       /* This is a paranoid check to stop integer overflow later on */
-       if (len > MAX_DUPLENGTH)
-         {
-         *errorcodeptr = ERR50;
-         goto FAILED;
-         }
        /* If the maximum repeat count is unlimited, find the end of the bracket
        by scanning through from the start, and compute the offset back to it
        from the current code pointer. There may be an OP_OPT setting following
-Line 3383 
 for (;; ptr++)
+Line 3842 
 for (;; ptr++)
          if (repeat_min > 1)
            {
            /* In the pre-compile phase, we don't actually do the replication. We
-           just adjust the length as if we had. */
+           just adjust the length as if we had. Do some paranoid checks for
+           potential integer overflow. */
            if (lengthptr != NULL)
-             *lengthptr += (repeat_min - 1)*length_prevgroup;
+             {
+             int delta = (repeat_min - 1)*length_prevgroup;
+             if ((double)(repeat_min - 1)*(double)length_prevgroup >
+                                                             (double)INT_MAX ||
+                 OFLOW_MAX - *lengthptr < delta)
+               {
+               *errorcodeptr = ERR20;
+               goto FAILED;
+               }
+             *lengthptr += delta;
+             }
            /* This is compiling for real */
-Line 3424 
 for (;; ptr++)
+Line 3894 
 for (;; ptr++)
          /* In the pre-compile phase, we don't actually do the replication. We
          just adjust the length as if we had. For each repetition we must add 1
          to the length for BRAZERO and for all but the last repetition we must
-         add 2 + 2*LINKSIZE to allow for the nesting that occurs. */
+         add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
+         paranoid checks to avoid integer overflow. */
          if (lengthptr != NULL && repeat_max > 0)
-           *lengthptr += repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
+           {
-- 2*LINK_SIZE;  /* Last one doesn't nest */
+           int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
+- 2*LINK_SIZE;   /* Last one doesn't nest */
+           if ((double)repeat_max *
+                 (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
+                   > (double)INT_MAX ||
+               OFLOW_MAX - *lengthptr < delta)
+             {
+             *errorcodeptr = ERR20;
+             goto FAILED;
+             }
+           *lengthptr += delta;
+           }
          /* This is compiling for real */
-Line 3580 
 for (;; ptr++)
+Line 4062 
 for (;; ptr++)
      /* ===================================================================*/
      /* Start of nested parenthesized sub-expression, or comment or lookahead or
      lookbehind or option setting or condition or all the other extended
-     parenthesis forms. First deal with the specials; all are introduced by ?,
+     parenthesis forms.  */
-     and the appearance of any of them means that this is not a capturing
-     group. */
      case '(':
      newoptions = options;
      skipbytes = 0;
      bravalue = OP_CBRA;
      save_hwm = cd->hwm;
      reset_bracount = FALSE;
+     /* First deal with various "verbs" that can be introduced by '*'. */
+     if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
+       {
+       int i, namelen;
+       const char *vn = verbnames;
+       const uschar *name = ++ptr;
+       previous = NULL;
+       while ((cd->ctypes[*++ptr] & ctype_letter) != 0);
+       if (*ptr == ':')
+         {
+         *errorcodeptr = ERR59;   /* Not supported */
+         goto FAILED;
+         }
+       if (*ptr != ')')
+         {
+         *errorcodeptr = ERR60;
+         goto FAILED;
+         }
+       namelen = ptr - name;
+       for (i = 0; i < verbcount; i++)
+         {
+         if (namelen == verbs[i].len &&
+             strncmp((char *)name, vn, namelen) == 0)
+           {
+           *code = verbs[i].op;
+           if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;
+           break;
+           }
+         vn += verbs[i].len + 1;
+         }
+       if (i < verbcount) continue;
+       *errorcodeptr = ERR60;
+       goto FAILED;
+       }
-     if (*(++ptr) == '?')
+     /* Deal with the extended parentheses; all are introduced by '?', and the
+     appearance of any of them means that this is not a capturing group. */
+     else if (*ptr == '?')
        {
        int i, set, unset, namelen;
        int *optset;
-Line 3614 
 for (;; ptr++)
+Line 4133 
 for (;; ptr++)
          /* ------------------------------------------------------------ */
          case '|':                 /* Reset capture count for each branch */
          reset_bracount = TRUE;
          /* Fall through */
          /* ------------------------------------------------------------ */
          case ':':                 /* Non-capturing bracket */
-Line 3726 
 for (;; ptr++)
+Line 4245 
 for (;; ptr++)
              *errorcodeptr = ERR58;
              goto FAILED;
              }
-           if (refsign == '-')
+           recno = (refsign == '-')?
+             cd->bracount - recno + 1 : recno +cd->bracount;
+           if (recno <= 0 || recno > cd->final_bracount)
              {
-             recno = cd->bracount - recno + 1;
+             *errorcodeptr = ERR15;
-             if (recno <= 0)
+             goto FAILED;
-               {
-               *errorcodeptr = ERR15;
-               goto FAILED;
-               }
              }
-           else recno += cd->bracount;
            PUT2(code, 2+LINK_SIZE, recno);
            break;
            }
-Line 3807 
 for (;; ptr++)
+Line 4323 
 for (;; ptr++)
            skipbytes = 1;
            }
-         /* Check for the "name" actually being a subpattern number. */
+         /* Check for the "name" actually being a subpattern number. We are
+         in the second pass here, so final_bracount is set. */
-         else if (recno > 0)
+         else if (recno > 0 && recno <= cd->final_bracount)
            {
            PUT2(code, 2+LINK_SIZE, recno);
            }
-Line 3833 
 for (;; ptr++)
+Line 4350 
 for (;; ptr++)
          /* ------------------------------------------------------------ */
          case '!':                 /* Negative lookahead */
-         bravalue = OP_ASSERT_NOT;
          ptr++;
+         if (*ptr == ')')          /* Optimize (?!) */
+           {
+           *code++ = OP_FAIL;
+           previous = NULL;
+           continue;
+           }
+         bravalue = OP_ASSERT_NOT;
          break;
-Line 3997 
 for (;; ptr++)
+Line 4520 
 for (;; ptr++)
          /* We come here from the Python syntax above that handles both
          references (?P=name) and recursion (?P>name), as well as falling
-         through from the Perl recursion syntax (?&name). */
+         through from the Perl recursion syntax (?&name). We also come here from
+         the Perl \k<name> or \k'name' back reference syntax and the \k{name}
+         .NET syntax. */
          NAMED_REF_OR_RECURSE:
          name = ++ptr;
-Line 4009 
 for (;; ptr++)
+Line 4534 
 for (;; ptr++)
          if (lengthptr != NULL)
            {
+           if (namelen == 0)
+             {
+             *errorcodeptr = ERR62;
+             goto FAILED;
+             }
            if (*ptr != terminator)
              {
              *errorcodeptr = ERR42;
-Line 4022 
 for (;; ptr++)
+Line 4552 
 for (;; ptr++)
            recno = 0;
            }
-         /* In the real compile, seek the name in the table */
+         /* In the real compile, seek the name in the table. We check the name
+         first, and then check that we have reached the end of the name in the
+         table. That way, if the name that is longer than any in the table,
+         the comparison will fail without reading beyond the table entry. */
          else
            {
            slot = cd->name_table;
            for (i = 0; i < cd->names_found; i++)
              {
-             if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;
+             if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&
+                 slot[2+namelen] == 0)
+               break;
              slot += cd->name_entry_size;
              }
-Line 4066 
 for (;; ptr++)
+Line 4601 
 for (;; ptr++)
            {
            const uschar *called;
-           if ((refsign = *ptr) == '+') ptr++;
+           if ((refsign = *ptr) == '+')
+             {
+             ptr++;
+             if ((digitab[*ptr] & ctype_digit) == 0)
+               {
+               *errorcodeptr = ERR63;
+               goto FAILED;
+               }
+             }
            else if (refsign == '-')
              {
              if ((digitab[ptr[1]] & ctype_digit) == 0)
-Line 4192 
 for (;; ptr++)
+Line 4735 
 for (;; ptr++)
              case 'J':    /* Record that it changed in the external options */
              *optset |= PCRE_DUPNAMES;
-             cd->external_options |= PCRE_JCHANGED;
+             cd->external_flags |= PCRE_JCHANGED;
              break;
              case 'i': *optset |= PCRE_CASELESS; break;
-Line 4315 
 for (;; ptr++)
+Line 4858 
 for (;; ptr++)
           errorcodeptr,                 /* Where to put an error message */
           (bravalue == OP_ASSERTBACK ||
            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
           reset_bracount,               /* True if (?| group */
           skipbytes,                    /* Skip over bracket number */
           &subfirstbyte,                /* For possible first char */
           &subreqbyte,                  /* For possible last char */
-Line 4383 
 for (;; ptr++)
+Line 4926 
 for (;; ptr++)
        goto FAILED;
        }
-     /* In the pre-compile phase, update the length by the length of the nested
+     /* In the pre-compile phase, update the length by the length of the group,
-     group, less the brackets at either end. Then reduce the compiled code to
+     less the brackets at either end. Then reduce the compiled code to just a
-     just the brackets so that it doesn't use much memory if it is duplicated by
+     set of non-capturing brackets so that it doesn't use much memory if it is
-     a quantifier. */
+     duplicated by a quantifier.*/
      if (lengthptr != NULL)
        {
+       if (OFLOW_MAX - *lengthptr < length_prevgroup - 2 - 2*LINK_SIZE)
+         {
+         *errorcodeptr = ERR20;
+         goto FAILED;
+         }
        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
-       code++;
+       *code++ = OP_BRA;
        PUTINC(code, 0, 1 + LINK_SIZE);
        *code++ = OP_KET;
        PUTINC(code, 0, 1 + LINK_SIZE);
+       break;    /* No need to waste time with special character handling */
        }
      /* Otherwise update the main code pointer to the end of the group. */
-     else code = tempcode;
+     code = tempcode;
      /* For a DEFINE group, required and first character settings are not
      relevant. */
-Line 4603 
 for (;; ptr++)
+Line 5152 
 for (;; ptr++)
      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;
      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
+     /* Remember if \r or \n were seen */
+     if (mcbuffer[0] == '\r' || mcbuffer[0] == '\n')
+       cd->external_flags |= PCRE_HASCRORLF;
      /* Set the first and required bytes appropriately. If no previous first
      byte, set it from this character, but revert to none on a zero repeat.
      Otherwise, leave the firstbyte value alone, and don't change it on a zero
-Line 4675 
 Arguments:
+Line 5229 
 Arguments:
    ptrptr         -> the address of the current pattern pointer
    errorcodeptr   -> pointer to error code variable
    lookbehind     TRUE if this is a lookbehind assertion
    reset_bracount TRUE to reset the count for each branch
    skipbytes      skip this many bytes at start (for brackets and OP_COND)
    firstbyteptr   place to put the first required character, or a negative number
    reqbyteptr     place to put the last required character, or a negative number
-Line 4689 
 Returns:         TRUE on success
+Line 5243 
 Returns:         TRUE on success
  static BOOL
  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,
    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,
    int *lengthptr)
  {
  const uschar *ptr = *ptrptr;
-Line 4735 
 orig_bracount = max_bracount = cd->braco
+Line 5289 
 orig_bracount = max_bracount = cd->braco
  for (;;)
    {
    /* For a (?| group, reset the capturing bracket count so that each branch
    uses the same numbers. */
    if (reset_bracount) cd->bracount = orig_bracount;
    /* Handle a change of ims options at the start of the branch */
    if ((options & PCRE_IMS) != oldims)
-Line 4767 
 for (;;)
+Line 5321 
 for (;;)
      *ptrptr = ptr;
      return FALSE;
      }
    /* Keep the highest bracket count in case (?| was used and some branch
    has fewer than the rest. */
    if (cd->bracount > max_bracount) max_bracount = cd->bracount;
    /* In the real compile phase, there is some post-processing to be done. */
-Line 4874 
 for (;;)
+Line 5428 
 for (;;)
        *code++ = oldims;
        length += 2;
        }
      /* Retain the highest bracket number, in case resetting was used. */
      cd->bracount = max_bracount;
      /* Set values to pass back */
-Line 4885 
 for (;;)
+Line 5439 
 for (;;)
      *ptrptr = ptr;
      *firstbyteptr = firstbyte;
      *reqbyteptr = reqbyte;
-     if (lengthptr != NULL) *lengthptr += length;
+     if (lengthptr != NULL)
+       {
+       if (OFLOW_MAX - *lengthptr < length)
+         {
+         *errorcodeptr = ERR20;
+         return FALSE;
+         }
+       *lengthptr += length;
+       }
      return TRUE;
      }
-Line 5194 
 real_pcre *re;
+Line 5756 
 real_pcre *re;
  int length = 1;  /* For final END opcode */
  int firstbyte, reqbyte, newline;
  int errorcode = 0;
+ int skipatstart = 0;
  #ifdef SUPPORT_UTF8
  BOOL utf8;
  #endif
-Line 5272 
 cd->fcc = tables + fcc_offset;
+Line 5835 
 cd->fcc = tables + fcc_offset;
  cd->cbits = tables + cbits_offset;
  cd->ctypes = tables + ctypes_offset;
+ /* Check for global one-time settings at the start of the pattern, and remember
+ the offset for later. */
+ while (ptr[skipatstart] == '(' && ptr[skipatstart+1] == '*')
+   {
+   int newnl = 0;
+   int newbsr = 0;
+   if (strncmp((char *)(ptr+skipatstart+2), "CR)", 3) == 0)
+     { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
+   else if (strncmp((char *)(ptr+skipatstart+2), "LF)", 3)  == 0)
+     { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }
+   else if (strncmp((char *)(ptr+skipatstart+2), "CRLF)", 5)  == 0)
+     { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }
+   else if (strncmp((char *)(ptr+skipatstart+2), "ANY)", 4) == 0)
+     { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }
+   else if (strncmp((char *)(ptr+skipatstart+2), "ANYCRLF)", 8)  == 0)
+     { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }
+   else if (strncmp((char *)(ptr+skipatstart+2), "BSR_ANYCRLF)", 12) == 0)
+     { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }
+   else if (strncmp((char *)(ptr+skipatstart+2), "BSR_UNICODE)", 12) == 0)
+     { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }
+   if (newnl != 0)
+     options = (options & ~PCRE_NEWLINE_BITS) | newnl;
+   else if (newbsr != 0)
+     options = (options & ~(PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) | newbsr;
+   else break;
+   }
+ /* Check validity of \R options. */
+ switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
+   {
+   case 0:
+   case PCRE_BSR_ANYCRLF:
+   case PCRE_BSR_UNICODE:
+   break;
+   default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;
+   }
  /* Handle different types of newline. The three bits give seven cases. The
  current code allows for fixed one- or two-byte sequences, plus "any" and
  "anycrlf". */
- switch (options & (PCRE_NEWLINE_CRLF | PCRE_NEWLINE_ANY))
+ switch (options & PCRE_NEWLINE_BITS)
    {
-   case 0: newline = NEWLINE; break;   /* Compile-time default */
+   case 0: newline = NEWLINE; break;   /* Build-time default */
    case PCRE_NEWLINE_CR: newline = '\r'; break;
    case PCRE_NEWLINE_LF: newline = '\n'; break;
    case PCRE_NEWLINE_CR+
-Line 5331 
 to compile parts of the pattern into; th
+Line 5936 
 to compile parts of the pattern into; th
  no longer needed, so hopefully this workspace will never overflow, though there
  is a test for its doing so. */
- cd->bracount = 0;
+ cd->bracount = cd->final_bracount = 0;
  cd->names_found = 0;
  cd->name_entry_size = 0;
  cd->name_table = NULL;
-Line 5341 
 cd->hwm = cworkspace;
+Line 5946 
 cd->hwm = cworkspace;
  cd->start_pattern = (const uschar *)pattern;
  cd->end_pattern = (const uschar *)(pattern + strlen(pattern));
  cd->req_varyopt = 0;
- cd->nopartial = FALSE;
  cd->external_options = options;
+ cd->external_flags = 0;
  /* Now do the pre-compile. On error, errorcode will be set non-zero, so we
  don't need to look at the result of the function here. The initial options have
-Line 5350 
 been put into the cd block so that they
+Line 5955 
 been put into the cd block so that they
  found within the regex right at the beginning. Bringing initial option settings
  outside can help speed up starting point checks. */
+ ptr += skipatstart;
  code = cworkspace;
  *code = OP_BRA;
  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,
    &code, &ptr, &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd,
    &length);
  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
-Line 5380 
 if (re == NULL)
+Line 5986 
 if (re == NULL)
    goto PCRE_EARLY_ERROR_RETURN;
    }
- /* Put in the magic number, and save the sizes, initial options, and character
+ /* Put in the magic number, and save the sizes, initial options, internal
- table pointer. NULL is used for the default character tables. The nullpad field
+ flags, and character table pointer. NULL is used for the default character
- is at the end; it's there to help in the case when a regex compiled on a system
+ tables. The nullpad field is at the end; it's there to help in the case when a
- with 4-byte pointers is run on another with 8-byte pointers. */
+ regex compiled on a system with 4-byte pointers is run on another with 8-byte
+ pointers. */
  re->magic_number = MAGIC_NUMBER;
  re->size = size;
  re->options = cd->external_options;
+ re->flags = cd->external_flags;
  re->dummy1 = 0;
  re->first_byte = 0;
  re->req_byte = 0;
-Line 5405 
 field. Reset the bracket count and the n
+Line 6013 
 field. Reset the bracket count and the n
  field; this time it's used for remembering forward references to subpatterns.
  */
+ cd->final_bracount = cd->bracount;  /* Save for checking forward references */
  cd->bracount = 0;
  cd->names_found = 0;
  cd->name_table = (uschar *)re + re->name_table_offset;
-Line 5412 
 codestart = cd->name_table + re->name_en
+Line 6021 
 codestart = cd->name_table + re->name_en
  cd->start_code = codestart;
  cd->hwm = cworkspace;
  cd->req_varyopt = 0;
- cd->nopartial = FALSE;
+ cd->had_accept = FALSE;
  /* Set up a starting, non-extracting bracket, then compile the expression. On
  error, errorcode will be set non-zero, so we don't need to look at the result
  of the function here. */
- ptr = (const uschar *)pattern;
+ ptr = (const uschar *)pattern + skipatstart;
  code = (uschar *)codestart;
  *code = OP_BRA;
  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,
    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);
  re->top_bracket = cd->bracount;
  re->top_backref = cd->top_backref;
+ re->flags = cd->external_flags;
- if (cd->nopartial) re->options |= PCRE_NOPARTIAL;
+ if (cd->had_accept) reqbyte = -1;   /* Must disable after (*ACCEPT) */
  /* If not reached end of pattern on success, there's an excess bracket. */
-Line 5468 
 if (errorcode != 0)
+Line 6078 
 if (errorcode != 0)
    PCRE_EARLY_ERROR_RETURN:
    *erroroffset = ptr - (const uschar *)pattern;
    PCRE_EARLY_ERROR_RETURN2:
-   *errorptr = error_texts[errorcode];
+   *errorptr = find_error_text(errorcode);
    if (errorcodeptr != NULL) *errorcodeptr = errorcode;
    return NULL;
    }
-Line 5497 
 if ((re->options & PCRE_ANCHORED) == 0)
+Line 6107 
 if ((re->options & PCRE_ANCHORED) == 0)
        int ch = firstbyte & 255;
        re->first_byte = ((firstbyte & REQ_CASELESS) != 0 &&
           cd->fcc[ch] == ch)? ch : firstbyte;
-       re->options |= PCRE_FIRSTSET;
+       re->flags |= PCRE_FIRSTSET;
        }
      else if (is_startline(codestart, 0, cd->backref_map))
-       re->options |= PCRE_STARTLINE;
+       re->flags |= PCRE_STARTLINE;
      }
    }
-Line 5514 
 if (reqbyte >= 0 &&
+Line 6124 
 if (reqbyte >= 0 &&
    int ch = reqbyte & 255;
    re->req_byte = ((reqbyte & REQ_CASELESS) != 0 &&
      cd->fcc[ch] == ch)? (reqbyte & ~REQ_CASELESS) : reqbyte;
-   re->options |= PCRE_REQCHSET;
+   re->flags |= PCRE_REQCHSET;
    }
  /* Print out the compiled data if debugging is enabled. This is never the
-Line 5525 
 case when building a production library.
+Line 6135 
 case when building a production library.
  printf("Length = %d top_bracket = %d top_backref = %d\n",
    length, re->top_bracket, re->top_backref);
- if (re->options != 0)
+ printf("Options=%08x\n", re->options);
-   {
-   printf("%s%s%s%s%s%s%s%s%s\n",
-     ((re->options & PCRE_NOPARTIAL) != 0)? "nopartial " : "",
-     ((re->options & PCRE_ANCHORED) != 0)? "anchored " : "",
-     ((re->options & PCRE_CASELESS) != 0)? "caseless " : "",
-     ((re->options & PCRE_EXTENDED) != 0)? "extended " : "",
-     ((re->options & PCRE_MULTILINE) != 0)? "multiline " : "",
-     ((re->options & PCRE_DOTALL) != 0)? "dotall " : "",
-     ((re->options & PCRE_DOLLAR_ENDONLY) != 0)? "endonly " : "",
-     ((re->options & PCRE_EXTRA) != 0)? "extra " : "",
-     ((re->options & PCRE_UNGREEDY) != 0)? "ungreedy " : "");
-   }
- if ((re->options & PCRE_FIRSTSET) != 0)
+ if ((re->flags & PCRE_FIRSTSET) != 0)
    {
    int ch = re->first_byte & 255;
    const char *caseless = ((re->first_byte & REQ_CASELESS) == 0)?
-Line 5548 
 if ((re->options & PCRE_FIRSTSET) != 0)
+Line 6146 
 if ((re->options & PCRE_FIRSTSET) != 0)
      else printf("First char = \\x%02x%s\n", ch, caseless);
    }
- if ((re->options & PCRE_REQCHSET) != 0)
+ if ((re->flags & PCRE_REQCHSET) != 0)
    {
    int ch = re->req_byte & 255;
    const char *caseless = ((re->req_byte & REQ_CASELESS) == 0)?
-Line 5565 
 was compiled can be seen. */
+Line 6163 
 was compiled can be seen. */
  if (code - codestart > length)
    {
    (pcre_free)(re);
-   *errorptr = error_texts[ERR23];
+   *errorptr = find_error_text(ERR23);
    *erroroffset = ptr - (uschar *)pattern;
    if (errorcodeptr != NULL) *errorcodeptr = ERR23;
    return NULL;

 Legend:



Removed from v.176
 


changed lines


 
Added in v.276
 Legend:



Removed from v.176
 


changed lines


 
Added in v.276
-Removed from v.176
+Added in v.276

webmaster@exim.org	ViewVC Help
Powered by ViewVC 1.1.12