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

Diff of /code/trunk/pcre_compile.c

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 438 by ph10, Sun Sep 6 20:00:47 2009 UTC revision 733 by ph10, Tue Oct 11 10:29:36 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2009 University of Cambridge             Copyright (c) 1997-2011 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57  used by pcretest. DEBUG is not defined when building a production library. */  also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60  #ifdef DEBUG  #ifdef PCRE_DEBUG
61  #include "pcre_printint.src"  #include "pcre_printint.src"
62  #endif  #endif
63    
# Line 91  is 4 there is plenty of room. */ Line 92  is 4 there is plenty of room. */
92    
93  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (4096)
94    
95    /* The overrun tests check for a slightly smaller size so that they detect the
96    overrun before it actually does run off the end of the data block. */
97    
98    #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
99    
100    
101  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
102  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
# Line 118  static const short int escapes[] = { Line 124  static const short int escapes[] = {
124       -ESC_H,                  0,       -ESC_H,                  0,
125       0,                       -ESC_K,       0,                       -ESC_K,
126       0,                       0,       0,                       0,
127       0,                       0,       -ESC_N,                  0,
128       -ESC_P,                  -ESC_Q,       -ESC_P,                  -ESC_Q,
129       -ESC_R,                  -ESC_S,       -ESC_R,                  -ESC_S,
130       0,                       0,       0,                       0,
# Line 165  static const short int escapes[] = { Line 171  static const short int escapes[] = {
171  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
172  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
173  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
175  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
177  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
# Line 182  string is built from string macros so th Line 188  string is built from string macros so th
188  platforms. */  platforms. */
189    
190  typedef struct verbitem {  typedef struct verbitem {
191    int   len;    int   len;                 /* Length of verb name */
192    int   op;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
193      int   op_arg;              /* Op when arg present, or -1 if not allowed */
194  } verbitem;  } verbitem;
195    
196  static const char verbnames[] =  static const char verbnames[] =
197      "\0"                       /* Empty name is a shorthand for MARK */
198      STRING_MARK0
199    STRING_ACCEPT0    STRING_ACCEPT0
200    STRING_COMMIT0    STRING_COMMIT0
201    STRING_F0    STRING_F0
# Line 196  static const char verbnames[] = Line 205  static const char verbnames[] =
205    STRING_THEN;    STRING_THEN;
206    
207  static const verbitem verbs[] = {  static const verbitem verbs[] = {
208    { 6, OP_ACCEPT },    { 0, -1,        OP_MARK },
209    { 6, OP_COMMIT },    { 4, -1,        OP_MARK },
210    { 1, OP_FAIL },    { 6, OP_ACCEPT, -1 },
211    { 4, OP_FAIL },    { 6, OP_COMMIT, -1 },
212    { 5, OP_PRUNE },    { 1, OP_FAIL,   -1 },
213    { 4, OP_SKIP  },    { 4, OP_FAIL,   -1 },
214    { 4, OP_THEN  }    { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217  };  };
218    
219  static const int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
# Line 250  static const int posix_class_maps[] = { Line 261  static const int posix_class_maps[] = {
261    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
262  };  };
263    
264    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
265    substitutes must be in the order of the names, defined above, and there are
266    both positive and negative cases. NULL means no substitute. */
267    
268    #ifdef SUPPORT_UCP
269    static const uschar *substitutes[] = {
270      (uschar *)"\\P{Nd}",    /* \D */
271      (uschar *)"\\p{Nd}",    /* \d */
272      (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */
273      (uschar *)"\\p{Xsp}",   /* \s */
274      (uschar *)"\\P{Xwd}",   /* \W */
275      (uschar *)"\\p{Xwd}"    /* \w */
276    };
277    
278    static const uschar *posix_substitutes[] = {
279      (uschar *)"\\p{L}",     /* alpha */
280      (uschar *)"\\p{Ll}",    /* lower */
281      (uschar *)"\\p{Lu}",    /* upper */
282      (uschar *)"\\p{Xan}",   /* alnum */
283      NULL,                   /* ascii */
284      (uschar *)"\\h",        /* blank */
285      NULL,                   /* cntrl */
286      (uschar *)"\\p{Nd}",    /* digit */
287      NULL,                   /* graph */
288      NULL,                   /* print */
289      NULL,                   /* punct */
290      (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */
291      (uschar *)"\\p{Xwd}",   /* word */
292      NULL,                   /* xdigit */
293      /* Negated cases */
294      (uschar *)"\\P{L}",     /* ^alpha */
295      (uschar *)"\\P{Ll}",    /* ^lower */
296      (uschar *)"\\P{Lu}",    /* ^upper */
297      (uschar *)"\\P{Xan}",   /* ^alnum */
298      NULL,                   /* ^ascii */
299      (uschar *)"\\H",        /* ^blank */
300      NULL,                   /* ^cntrl */
301      (uschar *)"\\P{Nd}",    /* ^digit */
302      NULL,                   /* ^graph */
303      NULL,                   /* ^print */
304      NULL,                   /* ^punct */
305      (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */
306      (uschar *)"\\P{Xwd}",   /* ^word */
307      NULL                    /* ^xdigit */
308    };
309    #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))
310    #endif
311    
312  #define STRING(a)  # a  #define STRING(a)  # a
313  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 262  the number of relocations needed when a Line 320  the number of relocations needed when a
320  it is now one long string. We cannot use a table of offsets, because the  it is now one long string. We cannot use a table of offsets, because the
321  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
322  simply count through to the one we want - this isn't a performance issue  simply count through to the one we want - this isn't a performance issue
323  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
324    
325    Each substring ends with \0 to insert a null character. This includes the final
326    substring, so that the whole string ends with \0\0, which can be detected when
327    counting through. */
328    
329  static const char error_texts[] =  static const char error_texts[] =
330    "no error\0"    "no error\0"
# Line 309  static const char error_texts[] = Line 371  static const char error_texts[] =
371    /* 35 */    /* 35 */
372    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
373    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
374    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
375    "number after (?C is > 255\0"    "number after (?C is > 255\0"
376    "closing ) for (?C expected\0"    "closing ) for (?C expected\0"
377    /* 40 */    /* 40 */
# Line 331  static const char error_texts[] = Line 393  static const char error_texts[] =
393    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
394    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
395    /* 55 */    /* 55 */
396    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
397    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
398    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
399    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
400    "(*VERB) with an argument is not supported\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
401    /* 60 */    /* 60 */
402    "(*VERB) not recognized\0"    "(*VERB) not recognized\0"
403    "number is too big\0"    "number is too big\0"
404    "subpattern name expected\0"    "subpattern name expected\0"
405    "digit expected after (?+\0"    "digit expected after (?+\0"
406    "] is an invalid data character in JavaScript compatibility mode";    "] is an invalid data character in JavaScript compatibility mode\0"
407      /* 65 */
408      "different names for subpatterns of the same number are not allowed\0"
409      "(*MARK) must have an argument\0"
410      "this version of PCRE is not compiled with PCRE_UCP support\0"
411      "\\c must be followed by an ASCII character\0"
412      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
413      ;
414    
415  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
416  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 478  static const unsigned char ebcdic_charta Line 546  static const unsigned char ebcdic_charta
546  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
547    
548  static BOOL  static BOOL
549    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,
550      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
551    
552    
# Line 500  static const char * Line 568  static const char *
568  find_error_text(int n)  find_error_text(int n)
569  {  {
570  const char *s = error_texts;  const char *s = error_texts;
571  for (; n > 0; n--) while (*s++ != 0) {};  for (; n > 0; n--)
572      {
573      while (*s++ != 0) {};
574      if (*s == 0) return "Error text not found (please report)";
575      }
576  return s;  return s;
577  }  }
578    
579    
580  /*************************************************  /*************************************************
581    *            Check for counted repeat            *
582    *************************************************/
583    
584    /* This function is called when a '{' is encountered in a place where it might
585    start a quantifier. It looks ahead to see if it really is a quantifier or not.
586    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
587    where the ddds are digits.
588    
589    Arguments:
590      p         pointer to the first char after '{'
591    
592    Returns:    TRUE or FALSE
593    */
594    
595    static BOOL
596    is_counted_repeat(const uschar *p)
597    {
598    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
599    while ((digitab[*p] & ctype_digit) != 0) p++;
600    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
601    
602    if (*p++ != CHAR_COMMA) return FALSE;
603    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
604    
605    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
606    while ((digitab[*p] & ctype_digit) != 0) p++;
607    
608    return (*p == CHAR_RIGHT_CURLY_BRACKET);
609    }
610    
611    
612    
613    /*************************************************
614  *            Handle escapes                      *  *            Handle escapes                      *
615  *************************************************/  *************************************************/
616    
# Line 571  else Line 676  else
676    
677      case CHAR_l:      case CHAR_l:
678      case CHAR_L:      case CHAR_L:
     case CHAR_N:  
679      case CHAR_u:      case CHAR_u:
680      case CHAR_U:      case CHAR_U:
681      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
682      break;      break;
683    
684      /* \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
685        class, \g must be followed by one of a number of specific things:
686    
687      (1) A number, either plain or braced. If positive, it is an absolute      (1) A number, either plain or braced. If positive, it is an absolute
688      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
# Line 594  else Line 699  else
699      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
700    
701      case CHAR_g:      case CHAR_g:
702        if (isclass) break;
703      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
704        {        {
705        c = -ESC_g;        c = -ESC_g;
# Line 772  else Line 878  else
878      break;      break;
879    
880      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
881      This coding is ASCII-specific, but then the whole concept of \cx is      An error is given if the byte following \c is not an ASCII character. This
882        coding is ASCII-specific, but then the whole concept of \cx is
883      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
884    
885      case CHAR_c:      case CHAR_c:
# Line 782  else Line 889  else
889        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
890        break;        break;
891        }        }
892    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
893  #ifndef EBCDIC  /* ASCII/UTF-8 coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
894          {
895          *errorcodeptr = ERR68;
896          break;
897          }
898      if (c >= CHAR_a && c <= CHAR_z) c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
899      c ^= 0x40;      c ^= 0x40;
900  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
901      if (c >= CHAR_a && c <= CHAR_z) c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
902      c ^= 0xC0;      c ^= 0xC0;
903  #endif  #endif
# Line 809  else Line 920  else
920      }      }
921    }    }
922    
923    /* Perl supports \N{name} for character names, as well as plain \N for "not
924    newline". PCRE does not support \N{name}. However, it does support
925    quantification such as \N{2,3}. */
926    
927    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
928         !is_counted_repeat(ptr+2))
929      *errorcodeptr = ERR37;
930    
931    /* If PCRE_UCP is set, we change the values for \d etc. */
932    
933    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
934      c -= (ESC_DU - ESC_D);
935    
936    /* Set the pointer to the final character before returning. */
937    
938  *ptrptr = ptr;  *ptrptr = ptr;
939  return c;  return c;
940  }  }
# Line 909  return -1; Line 1035  return -1;
1035    
1036    
1037  /*************************************************  /*************************************************
 *            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);  
 }  
   
   
   
 /*************************************************  
1038  *         Read repeat counts                     *  *         Read repeat counts                     *
1039  *************************************************/  *************************************************/
1040    
# Line 1017  top-level call starts at the beginning o Line 1110  top-level call starts at the beginning o
1110  start at a parenthesis. It scans along a pattern's text looking for capturing  start at a parenthesis. It scans along a pattern's text looking for capturing
1111  subpatterns, and counting them. If it finds a named pattern that matches the  subpatterns, and counting them. If it finds a named pattern that matches the
1112  name it is given, it returns its number. Alternatively, if the name is NULL, it  name it is given, it returns its number. Alternatively, if the name is NULL, it
1113  returns when it reaches a given numbered subpattern. We know that if (?P< is  returns when it reaches a given numbered subpattern. Recursion is used to keep
1114  encountered, the name will be terminated by '>' because that is checked in the  track of subpatterns that reset the capturing group numbers - the (?| feature.
1115  first pass. Recursion is used to keep track of subpatterns that reset the  
1116  capturing group numbers - the (?| feature.  This function was originally called only from the second pass, in which we know
1117    that if (?< or (?' or (?P< is encountered, the name will be correctly
1118    terminated because that is checked in the first pass. There is now one call to
1119    this function in the first pass, to check for a recursive back reference by
1120    name (so that we can make the whole group atomic). In this case, we need check
1121    only up to the current position in the pattern, and that is still OK because
1122    and previous occurrences will have been checked. To make this work, the test
1123    for "end of pattern" is a check against cd->end_pattern in the main loop,
1124    instead of looking for a binary zero. This means that the special first-pass
1125    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1126    processing items within the loop are OK, because afterwards the main loop will
1127    terminate.)
1128    
1129  Arguments:  Arguments:
1130    ptrptr       address of the current character pointer (updated)    ptrptr       address of the current character pointer (updated)
# Line 1028  Arguments: Line 1132  Arguments:
1132    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1133    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1134    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1135      utf8         TRUE if we are in UTF-8 mode
1136    count        pointer to the current capturing subpattern number (updated)    count        pointer to the current capturing subpattern number (updated)
1137    
1138  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
# Line 1035  Returns:       the number of the named s Line 1140  Returns:       the number of the named s
1140    
1141  static int  static int
1142  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1143    BOOL xmode, int *count)    BOOL xmode, BOOL utf8, int *count)
1144  {  {
1145  uschar *ptr = *ptrptr;  uschar *ptr = *ptrptr;
1146  int start_count = *count;  int start_count = *count;
# Line 1047  dealing with. The very first call may no Line 1152  dealing with. The very first call may no
1152    
1153  if (ptr[0] == CHAR_LEFT_PARENTHESIS)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1154    {    {
1155    if (ptr[1] == CHAR_QUESTION_MARK &&    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1156        ptr[2] == CHAR_VERTICAL_LINE)  
1157      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1158    
1159      /* Handle a normal, unnamed capturing parenthesis. */
1160    
1161      else if (ptr[1] != CHAR_QUESTION_MARK)
1162        {
1163        *count += 1;
1164        if (name == NULL && *count == lorn) return *count;
1165        ptr++;
1166        }
1167    
1168      /* All cases now have (? at the start. Remember when we are in a group
1169      where the parenthesis numbers are duplicated. */
1170    
1171      else if (ptr[2] == CHAR_VERTICAL_LINE)
1172      {      {
1173      ptr += 3;      ptr += 3;
1174      dup_parens = TRUE;      dup_parens = TRUE;
1175      }      }
1176    
1177    /* Handle a normal, unnamed capturing parenthesis */    /* Handle comments; all characters are allowed until a ket is reached. */
1178    
1179    else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)    else if (ptr[2] == CHAR_NUMBER_SIGN)
1180      {      {
1181      *count += 1;      for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1182      if (name == NULL && *count == lorn) return *count;      goto FAIL_EXIT;
     ptr++;  
1183      }      }
1184    
1185    /* Handle a condition. If it is an assertion, just carry on so that it    /* Handle a condition. If it is an assertion, just carry on so that it
1186    is processed as normal. If not, skip to the closing parenthesis of the    is processed as normal. If not, skip to the closing parenthesis of the
1187    condition (there can't be any nested parens. */    condition (there can't be any nested parens). */
1188    
1189    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1190      {      {
# Line 1077  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1196  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1196        }        }
1197      }      }
1198    
1199    /* We have either (? or (* and not a condition */    /* Start with (? but not a condition. */
1200    
1201    else    else
1202      {      {
# Line 1100  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1219  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1219        if (name != NULL && lorn == ptr - thisname &&        if (name != NULL && lorn == ptr - thisname &&
1220            strncmp((const char *)name, (const char *)thisname, lorn) == 0)            strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1221          return *count;          return *count;
1222        term++;        term++;
1223        }        }
1224      }      }
1225    }    }
1226    
1227  /* Past any initial parenthesis handling, scan for parentheses or vertical  /* Past any initial parenthesis handling, scan for parentheses or vertical
1228  bars. */  bars. Stop if we get to cd->end_pattern. Note that this is important for the
1229    first-pass call when this value is temporarily adjusted to stop at the current
1230    position. So DO NOT change this to a test for binary zero. */
1231    
1232  for (; *ptr != 0; ptr++)  for (; ptr < cd->end_pattern; ptr++)
1233    {    {
1234    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1235    
# Line 1146  for (; *ptr != 0; ptr++) Line 1267  for (; *ptr != 0; ptr++)
1267            break;            break;
1268          }          }
1269        else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)        else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1270          {          {
1271          negate_class = TRUE;          negate_class = TRUE;
1272          ptr++;          ptr++;
1273          }          }
1274        else break;        else break;
1275        }        }
1276    
# Line 1182  for (; *ptr != 0; ptr++) Line 1303  for (; *ptr != 0; ptr++)
1303    
1304    if (xmode && *ptr == CHAR_NUMBER_SIGN)    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1305      {      {
1306      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};      ptr++;
1307        while (*ptr != 0)
1308          {
1309          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1310          ptr++;
1311    #ifdef SUPPORT_UTF8
1312          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1313    #endif
1314          }
1315      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1316      continue;      continue;
1317      }      }
# Line 1191  for (; *ptr != 0; ptr++) Line 1320  for (; *ptr != 0; ptr++)
1320    
1321    if (*ptr == CHAR_LEFT_PARENTHESIS)    if (*ptr == CHAR_LEFT_PARENTHESIS)
1322      {      {
1323      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1324      if (rc > 0) return rc;      if (rc > 0) return rc;
1325      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1326      }      }
# Line 1199  for (; *ptr != 0; ptr++) Line 1328  for (; *ptr != 0; ptr++)
1328    else if (*ptr == CHAR_RIGHT_PARENTHESIS)    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1329      {      {
1330      if (dup_parens && *count < hwm_count) *count = hwm_count;      if (dup_parens && *count < hwm_count) *count = hwm_count;
1331      *ptrptr = ptr;      goto FAIL_EXIT;
     return -1;  
1332      }      }
1333    
1334    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
# Line 1238  Arguments: Line 1366  Arguments:
1366    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1367    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1368    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1369      utf8         TRUE if we are in UTF-8 mode
1370    
1371  Returns:       the number of the found subpattern, or -1 if not found  Returns:       the number of the found subpattern, or -1 if not found
1372  */  */
1373    
1374  static int  static int
1375  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1376      BOOL utf8)
1377  {  {
1378  uschar *ptr = (uschar *)cd->start_pattern;  uschar *ptr = (uschar *)cd->start_pattern;
1379  int count = 0;  int count = 0;
# Line 1256  matching closing parens. That is why we Line 1386  matching closing parens. That is why we
1386    
1387  for (;;)  for (;;)
1388    {    {
1389    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1390    if (rc > 0 || *ptr++ == 0) break;    if (rc > 0 || *ptr++ == 0) break;
1391    }    }
1392    
# Line 1272  return rc; Line 1402  return rc;
1402    
1403  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1404  for a fixed first character, or an anchoring op code etc. It skips over things  for a fixed first character, or an anchoring op code etc. It skips over things
1405  that do not influence this. For some calls, a change of option is important.  that do not influence this. For some calls, it makes sense to skip negative
1406  For some calls, it makes sense to skip negative forward and all backward  forward and all backward assertions, and also the \b assertion; for others it
1407  assertions, and also the \b assertion; for others it does not.  does not.
1408    
1409  Arguments:  Arguments:
1410    code         pointer to the start of the group    code         pointer to the start of the group
   options      pointer to external options  
   optbit       the option bit whose changing is significant, or  
                  zero if none are  
1411    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1412    
1413  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1414  */  */
1415    
1416  static const uschar*  static const uschar*
1417  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1418  {  {
1419  for (;;)  for (;;)
1420    {    {
1421    switch ((int)*code)    switch ((int)*code)
1422      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1423      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1424      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1425      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 1315  for (;;) Line 1435  for (;;)
1435    
1436      case OP_CALLOUT:      case OP_CALLOUT:
1437      case OP_CREF:      case OP_CREF:
1438        case OP_NCREF:
1439      case OP_RREF:      case OP_RREF:
1440        case OP_NRREF:
1441      case OP_DEF:      case OP_DEF:
1442      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1443      break;      break;
# Line 1331  for (;;) Line 1453  for (;;)
1453    
1454    
1455  /*************************************************  /*************************************************
1456  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1457  *************************************************/  *************************************************/
1458    
1459  /* Scan a pattern and compute the fixed length of subject that will match it,  /* Scan a branch and compute the fixed length of subject that will match it,
1460  if the length is fixed. This is needed for dealing with backward assertions.  if the length is fixed. This is needed for dealing with backward assertions.
1461  In UTF8 mode, the result is in characters rather than bytes.  In UTF8 mode, the result is in characters rather than bytes. The branch is
1462    temporarily terminated with OP_END when this function is called.
1463    
1464    This function is called when a backward assertion is encountered, so that if it
1465    fails, the error message can point to the correct place in the pattern.
1466    However, we cannot do this when the assertion contains subroutine calls,
1467    because they can be forward references. We solve this by remembering this case
1468    and doing the check at the end; a flag specifies which mode we are running in.
1469    
1470  Arguments:  Arguments:
1471    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1472    options  the compiling options    utf8     TRUE in UTF-8 mode
1473      atend    TRUE if called when the pattern is complete
1474      cd       the "compile data" structure
1475    
1476  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1477                 or -1 if there is no fixed length,
1478               or -2 if \C was encountered               or -2 if \C was encountered
1479                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1480  */  */
1481    
1482  static int  static int
1483  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1484  {  {
1485  int length = -1;  int length = -1;
1486    
# Line 1360  branch, check the length against that of Line 1493  branch, check the length against that of
1493  for (;;)  for (;;)
1494    {    {
1495    int d;    int d;
1496      uschar *ce, *cs;
1497    register int op = *cc;    register int op = *cc;
1498    switch (op)    switch (op)
1499      {      {
1500        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1501        OP_BRA (normal non-capturing bracket) because the other variants of these
1502        opcodes are all concerned with unlimited repeated groups, which of course
1503        are not of fixed length. They will cause a -1 response from the default
1504        case of this switch. */
1505    
1506      case OP_CBRA:      case OP_CBRA:
1507      case OP_BRA:      case OP_BRA:
1508      case OP_ONCE:      case OP_ONCE:
1509        case OP_ONCE_NC:
1510      case OP_COND:      case OP_COND:
1511      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1512      if (d < 0) return d;      if (d < 0) return d;
1513      branchlength += d;      branchlength += d;
1514      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1376  for (;;) Line 1517  for (;;)
1517    
1518      /* 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
1519      call. If it's ALT it is an alternation in a nested call. If it is      call. If it's ALT it is an alternation in a nested call. If it is
1520      END it's the end of the outer call. All can be handled by the same code. */      END it's the end of the outer call. All can be handled by the same code.
1521        Note that we must not include the OP_KETRxxx opcodes here, because they
1522        all imply an unlimited repeat. */
1523    
1524      case OP_ALT:      case OP_ALT:
1525      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1526      case OP_END:      case OP_END:
1527      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1528        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
# Line 1390  for (;;) Line 1531  for (;;)
1531      branchlength = 0;      branchlength = 0;
1532      break;      break;
1533    
1534        /* A true recursion implies not fixed length, but a subroutine call may
1535        be OK. If the subroutine is a forward reference, we can't deal with
1536        it until the end of the pattern, so return -3. */
1537    
1538        case OP_RECURSE:
1539        if (!atend) return -3;
1540        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1541        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1542        if (cc > cs && cc < ce) return -1;                /* Recursion */
1543        d = find_fixedlength(cs + 2, utf8, atend, cd);
1544        if (d < 0) return d;
1545        branchlength += d;
1546        cc += 1 + LINK_SIZE;
1547        break;
1548    
1549      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1550    
1551      case OP_ASSERT:      case OP_ASSERT:
# Line 1403  for (;;) Line 1559  for (;;)
1559    
1560      case OP_REVERSE:      case OP_REVERSE:
1561      case OP_CREF:      case OP_CREF:
1562        case OP_NCREF:
1563      case OP_RREF:      case OP_RREF:
1564        case OP_NRREF:
1565      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
1566      case OP_CALLOUT:      case OP_CALLOUT:
1567      case OP_SOD:      case OP_SOD:
1568      case OP_SOM:      case OP_SOM:
1569        case OP_SET_SOM:
1570      case OP_EOD:      case OP_EOD:
1571      case OP_EODN:      case OP_EODN:
1572      case OP_CIRC:      case OP_CIRC:
1573        case OP_CIRCM:
1574      case OP_DOLL:      case OP_DOLL:
1575        case OP_DOLLM:
1576      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1577      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1578      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 1421  for (;;) Line 1581  for (;;)
1581      /* Handle literal characters */      /* Handle literal characters */
1582    
1583      case OP_CHAR:      case OP_CHAR:
1584      case OP_CHARNC:      case OP_CHARI:
1585      case OP_NOT:      case OP_NOT:
1586        case OP_NOTI:
1587      branchlength++;      branchlength++;
1588      cc += 2;      cc += 2;
1589  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1590      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
1591  #endif  #endif
1592      break;      break;
1593    
# Line 1438  for (;;) Line 1598  for (;;)
1598      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1599      cc += 4;      cc += 4;
1600  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1601      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
1602  #endif  #endif
1603      break;      break;
1604    
# Line 1518  for (;;) Line 1677  for (;;)
1677    
1678    
1679  /*************************************************  /*************************************************
1680  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1681  *************************************************/  *************************************************/
1682    
1683  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1684  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1685    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1686    so that it can be called from pcre_study() when finding the minimum matching
1687    length.
1688    
1689  Arguments:  Arguments:
1690    code        points to start of expression    code        points to start of expression
1691    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1692    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1693    
1694  Returns:      pointer to the opcode for the bracket, or NULL if not found  Returns:      pointer to the opcode for the bracket, or NULL if not found
1695  */  */
1696    
1697  static const uschar *  const uschar *
1698  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1699  {  {
1700  for (;;)  for (;;)
1701    {    {
1702    register int c = *code;    register int c = *code;
1703    
1704    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1705    
1706    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1546  for (;;) Line 1709  for (;;)
1709    
1710    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1711    
1712      /* Handle recursion */
1713    
1714      else if (c == OP_REVERSE)
1715        {
1716        if (number < 0) return (uschar *)code;
1717        code += _pcre_OP_lengths[c];
1718        }
1719    
1720    /* Handle capturing bracket */    /* Handle capturing bracket */
1721    
1722    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1723               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1724      {      {
1725      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1726      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1557  for (;;) Line 1729  for (;;)
1729    
1730    /* Otherwise, we can get the item's length from the table, except that for    /* Otherwise, we can get the item's length from the table, except that for
1731    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
1732    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1733      must add in its length. */
1734    
1735    else    else
1736      {      {
# Line 1581  for (;;) Line 1754  for (;;)
1754        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1755        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1756        break;        break;
1757    
1758          case OP_MARK:
1759          case OP_PRUNE_ARG:
1760          case OP_SKIP_ARG:
1761          code += code[1];
1762          break;
1763    
1764          case OP_THEN_ARG:
1765          code += code[1];
1766          break;
1767        }        }
1768    
1769      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1595  for (;;) Line 1778  for (;;)
1778      if (utf8) switch(c)      if (utf8) switch(c)
1779        {        {
1780        case OP_CHAR:        case OP_CHAR:
1781        case OP_CHARNC:        case OP_CHARI:
1782        case OP_EXACT:        case OP_EXACT:
1783          case OP_EXACTI:
1784        case OP_UPTO:        case OP_UPTO:
1785          case OP_UPTOI:
1786        case OP_MINUPTO:        case OP_MINUPTO:
1787          case OP_MINUPTOI:
1788        case OP_POSUPTO:        case OP_POSUPTO:
1789          case OP_POSUPTOI:
1790        case OP_STAR:        case OP_STAR:
1791          case OP_STARI:
1792        case OP_MINSTAR:        case OP_MINSTAR:
1793          case OP_MINSTARI:
1794        case OP_POSSTAR:        case OP_POSSTAR:
1795          case OP_POSSTARI:
1796        case OP_PLUS:        case OP_PLUS:
1797          case OP_PLUSI:
1798        case OP_MINPLUS:        case OP_MINPLUS:
1799          case OP_MINPLUSI:
1800        case OP_POSPLUS:        case OP_POSPLUS:
1801          case OP_POSPLUSI:
1802        case OP_QUERY:        case OP_QUERY:
1803          case OP_QUERYI:
1804        case OP_MINQUERY:        case OP_MINQUERY:
1805          case OP_MINQUERYI:
1806        case OP_POSQUERY:        case OP_POSQUERY:
1807          case OP_POSQUERYI:
1808        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1809        break;        break;
1810        }        }
# Line 1652  for (;;) Line 1848  for (;;)
1848    
1849    /* Otherwise, we can get the item's length from the table, except that for    /* Otherwise, we can get the item's length from the table, except that for
1850    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
1851    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1852      must add in its length. */
1853    
1854    else    else
1855      {      {
# Line 1676  for (;;) Line 1873  for (;;)
1873        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1874        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1875        break;        break;
1876    
1877          case OP_MARK:
1878          case OP_PRUNE_ARG:
1879          case OP_SKIP_ARG:
1880          code += code[1];
1881          break;
1882    
1883          case OP_THEN_ARG:
1884          code += code[1];
1885          break;
1886        }        }
1887    
1888      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1690  for (;;) Line 1897  for (;;)
1897      if (utf8) switch(c)      if (utf8) switch(c)
1898        {        {
1899        case OP_CHAR:        case OP_CHAR:
1900        case OP_CHARNC:        case OP_CHARI:
1901        case OP_EXACT:        case OP_EXACT:
1902          case OP_EXACTI:
1903        case OP_UPTO:        case OP_UPTO:
1904          case OP_UPTOI:
1905        case OP_MINUPTO:        case OP_MINUPTO:
1906          case OP_MINUPTOI:
1907        case OP_POSUPTO:        case OP_POSUPTO:
1908          case OP_POSUPTOI:
1909        case OP_STAR:        case OP_STAR:
1910          case OP_STARI:
1911        case OP_MINSTAR:        case OP_MINSTAR:
1912          case OP_MINSTARI:
1913        case OP_POSSTAR:        case OP_POSSTAR:
1914          case OP_POSSTARI:
1915        case OP_PLUS:        case OP_PLUS:
1916          case OP_PLUSI:
1917        case OP_MINPLUS:        case OP_MINPLUS:
1918          case OP_MINPLUSI:
1919        case OP_POSPLUS:        case OP_POSPLUS:
1920          case OP_POSPLUSI:
1921        case OP_QUERY:        case OP_QUERY:
1922          case OP_QUERYI:
1923        case OP_MINQUERY:        case OP_MINQUERY:
1924          case OP_MINQUERYI:
1925        case OP_POSQUERY:        case OP_POSQUERY:
1926          case OP_POSQUERYI:
1927        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1928        break;        break;
1929        }        }
# Line 1732  Arguments: Line 1952  Arguments:
1952    code        points to start of search    code        points to start of search
1953    endcode     points to where to stop    endcode     points to where to stop
1954    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1955      cd          contains pointers to tables etc.
1956    
1957  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1958  */  */
1959    
1960  static BOOL  static BOOL
1961  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1962      compile_data *cd)
1963  {  {
1964  register int c;  register int c;
1965  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
1966       code < endcode;       code < endcode;
1967       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
1968    {    {
1969    const uschar *ccode;    const uschar *ccode;
1970    
# Line 1758  for (code = first_significant_code(code Line 1980  for (code = first_significant_code(code
1980      continue;      continue;
1981      }      }
1982    
1983      /* For a recursion/subroutine call, if its end has been reached, which
1984      implies a backward reference subroutine call, we can scan it. If it's a
1985      forward reference subroutine call, we can't. To detect forward reference
1986      we have to scan up the list that is kept in the workspace. This function is
1987      called only when doing the real compile, not during the pre-compile that
1988      measures the size of the compiled pattern. */
1989    
1990      if (c == OP_RECURSE)
1991        {
1992        const uschar *scode;
1993        BOOL empty_branch;
1994    
1995        /* Test for forward reference */
1996    
1997        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
1998          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
1999    
2000        /* Not a forward reference, test for completed backward reference */
2001    
2002        empty_branch = FALSE;
2003        scode = cd->start_code + GET(code, 1);
2004        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2005    
2006        /* Completed backwards reference */
2007    
2008        do
2009          {
2010          if (could_be_empty_branch(scode, endcode, utf8, cd))
2011            {
2012            empty_branch = TRUE;
2013            break;
2014            }
2015          scode += GET(scode, 1);
2016          }
2017        while (*scode == OP_ALT);
2018    
2019        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2020        continue;
2021        }
2022    
2023    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
2024    
2025    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2026          c == OP_BRAPOSZERO)
2027      {      {
2028      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
2029      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1768  for (code = first_significant_code(code Line 2031  for (code = first_significant_code(code
2031      continue;      continue;
2032      }      }
2033    
2034      /* A nested group that is already marked as "could be empty" can just be
2035      skipped. */
2036    
2037      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2038          c == OP_SCBRA || c == OP_SCBRAPOS)
2039        {
2040        do code += GET(code, 1); while (*code == OP_ALT);
2041        c = *code;
2042        continue;
2043        }
2044    
2045    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2046    
2047    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2048          c == OP_CBRA || c == OP_CBRAPOS ||
2049          c == OP_ONCE || c == OP_ONCE_NC ||
2050          c == OP_COND)
2051      {      {
2052      BOOL empty_branch;      BOOL empty_branch;
2053      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1786  for (code = first_significant_code(code Line 2063  for (code = first_significant_code(code
2063        empty_branch = FALSE;        empty_branch = FALSE;
2064        do        do
2065          {          {
2066          if (!empty_branch && could_be_empty_branch(code, endcode, utf8))          if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2067            empty_branch = TRUE;            empty_branch = TRUE;
2068          code += GET(code, 1);          code += GET(code, 1);
2069          }          }
# Line 1857  for (code = first_significant_code(code Line 2134  for (code = first_significant_code(code
2134      case OP_ALLANY:      case OP_ALLANY:
2135      case OP_ANYBYTE:      case OP_ANYBYTE:
2136      case OP_CHAR:      case OP_CHAR:
2137      case OP_CHARNC:      case OP_CHARI:
2138      case OP_NOT:      case OP_NOT:
2139        case OP_NOTI:
2140      case OP_PLUS:      case OP_PLUS:
2141      case OP_MINPLUS:      case OP_MINPLUS:
2142      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1898  for (code = first_significant_code(code Line 2176  for (code = first_significant_code(code
2176      case OP_KET:      case OP_KET:
2177      case OP_KETRMAX:      case OP_KETRMAX:
2178      case OP_KETRMIN:      case OP_KETRMIN:
2179        case OP_KETRPOS:
2180      case OP_ALT:      case OP_ALT:
2181      return TRUE;      return TRUE;
2182    
# Line 1906  for (code = first_significant_code(code Line 2185  for (code = first_significant_code(code
2185    
2186  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2187      case OP_STAR:      case OP_STAR:
2188        case OP_STARI:
2189      case OP_MINSTAR:      case OP_MINSTAR:
2190        case OP_MINSTARI:
2191      case OP_POSSTAR:      case OP_POSSTAR:
2192        case OP_POSSTARI:
2193      case OP_QUERY:      case OP_QUERY:
2194        case OP_QUERYI:
2195      case OP_MINQUERY:      case OP_MINQUERY:
2196        case OP_MINQUERYI:
2197      case OP_POSQUERY:      case OP_POSQUERY:
2198        case OP_POSQUERYI:
2199      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2200      break;      break;
2201    
2202      case OP_UPTO:      case OP_UPTO:
2203        case OP_UPTOI:
2204      case OP_MINUPTO:      case OP_MINUPTO:
2205        case OP_MINUPTOI:
2206      case OP_POSUPTO:      case OP_POSUPTO:
2207        case OP_POSUPTOI:
2208      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2209      break;      break;
2210  #endif  #endif
2211    
2212        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2213        string. */
2214    
2215        case OP_MARK:
2216        case OP_PRUNE_ARG:
2217        case OP_SKIP_ARG:
2218        code += code[1];
2219        break;
2220    
2221        case OP_THEN_ARG:
2222        code += code[1];
2223        break;
2224    
2225        /* None of the remaining opcodes are required to match a character. */
2226    
2227        default:
2228        break;
2229      }      }
2230    }    }
2231    
# Line 1936  return TRUE; Line 2242  return TRUE;
2242  the current branch of the current pattern to see if it could match the empty  the current branch of the current pattern to see if it could match the empty
2243  string. If it could, we must look outwards for branches at other levels,  string. If it could, we must look outwards for branches at other levels,
2244  stopping when we pass beyond the bracket which is the subject of the recursion.  stopping when we pass beyond the bracket which is the subject of the recursion.
2245    This function is called only during the real compile, not during the
2246    pre-compile.
2247    
2248  Arguments:  Arguments:
2249    code        points to start of the recursion    code        points to start of the recursion
2250    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2251    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2252    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2253      cd          pointers to tables etc
2254    
2255  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2256  */  */
2257    
2258  static BOOL  static BOOL
2259  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2260    BOOL utf8)    BOOL utf8, compile_data *cd)
2261  {  {
2262  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2263    {    {
2264    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2265        return FALSE;
2266    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2267    }    }
2268  return TRUE;  return TRUE;
# Line 1984  where Perl recognizes it as the POSIX cl Line 2294  where Perl recognizes it as the POSIX cl
2294  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2295  I think.  I think.
2296    
2297    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2298    It seems that the appearance of a nested POSIX class supersedes an apparent
2299    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2300    a digit.
2301    
2302    In Perl, unescaped square brackets may also appear as part of class names. For
2303    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2304    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2305    seem right at all. PCRE does not allow closing square brackets in POSIX class
2306    names.
2307    
2308  Arguments:  Arguments:
2309    ptr      pointer to the initial [    ptr      pointer to the initial [
2310    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 1998  int terminator;          /* Don't combin Line 2319  int terminator;          /* Don't combin
2319  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2320  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2321    {    {
2322    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2323        ptr++;
2324      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2325      else
2326      {      {
     if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
2327      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2328        {        {
2329        *endptr = ptr;        *endptr = ptr;
2330        return TRUE;        return TRUE;
2331        }        }
2332        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2333             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2334              ptr[1] == CHAR_EQUALS_SIGN) &&
2335            check_posix_syntax(ptr, endptr))
2336          return FALSE;
2337      }      }
2338    }    }
2339  return FALSE;  return FALSE;
# Line 2134  auto_callout(uschar *code, const uschar Line 2462  auto_callout(uschar *code, const uschar
2462  {  {
2463  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2464  *code++ = 255;  *code++ = 255;
2465  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2466  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2467  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2468  }  }
2469    
# Line 2160  Returns:             nothing Line 2488  Returns:             nothing
2488  static void  static void
2489  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2490  {  {
2491  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2492  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2493  }  }
2494    
# Line 2210  for (++c; c <= d; c++) Line 2538  for (++c; c <= d; c++)
2538    
2539  return TRUE;  return TRUE;
2540  }  }
2541    
2542    
2543    
2544    /*************************************************
2545    *        Check a character and a property        *
2546    *************************************************/
2547    
2548    /* This function is called by check_auto_possessive() when a property item
2549    is adjacent to a fixed character.
2550    
2551    Arguments:
2552      c            the character
2553      ptype        the property type
2554      pdata        the data for the type
2555      negated      TRUE if it's a negated property (\P or \p{^)
2556    
2557    Returns:       TRUE if auto-possessifying is OK
2558    */
2559    
2560    static BOOL
2561    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2562    {
2563    const ucd_record *prop = GET_UCD(c);
2564    switch(ptype)
2565      {
2566      case PT_LAMP:
2567      return (prop->chartype == ucp_Lu ||
2568              prop->chartype == ucp_Ll ||
2569              prop->chartype == ucp_Lt) == negated;
2570    
2571      case PT_GC:
2572      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2573    
2574      case PT_PC:
2575      return (pdata == prop->chartype) == negated;
2576    
2577      case PT_SC:
2578      return (pdata == prop->script) == negated;
2579    
2580      /* These are specials */
2581    
2582      case PT_ALNUM:
2583      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2584              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2585    
2586      case PT_SPACE:    /* Perl space */
2587      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2588              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2589              == negated;
2590    
2591      case PT_PXSPACE:  /* POSIX space */
2592      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2593              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2594              c == CHAR_FF || c == CHAR_CR)
2595              == negated;
2596    
2597      case PT_WORD:
2598      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2599              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2600              c == CHAR_UNDERSCORE) == negated;
2601      }
2602    return FALSE;
2603    }
2604  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2605    
2606    
# Line 2223  whether the next thing could possibly ma Line 2614  whether the next thing could possibly ma
2614  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2615    
2616  Arguments:  Arguments:
2617    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2618    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2619    ptr           next character in pattern    ptr           next character in pattern
2620    options       options bits    options       options bits
2621    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2235  Returns:        TRUE if possessifying is Line 2624  Returns:        TRUE if possessifying is
2624  */  */
2625    
2626  static BOOL  static BOOL
2627  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2628    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2629  {  {
2630  int next;  int c, next;
2631    int op_code = *previous++;
2632    
2633  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2634    
# Line 2249  if ((options & PCRE_EXTENDED) != 0) Line 2639  if ((options & PCRE_EXTENDED) != 0)
2639      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2640      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2641        {        {
2642        while (*(++ptr) != 0)        ptr++;
2643          while (*ptr != 0)
2644            {
2645          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2646            ptr++;
2647    #ifdef SUPPORT_UTF8
2648            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2649    #endif
2650            }
2651        }        }
2652      else break;      else break;
2653      }      }
# Line 2286  if ((options & PCRE_EXTENDED) != 0) Line 2683  if ((options & PCRE_EXTENDED) != 0)
2683      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2684      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2685        {        {
2686        while (*(++ptr) != 0)        ptr++;
2687          while (*ptr != 0)
2688            {
2689          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2690            ptr++;
2691    #ifdef SUPPORT_UTF8
2692            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2693    #endif
2694            }
2695        }        }
2696      else break;      else break;
2697      }      }
# Line 2299  if (*ptr == CHAR_ASTERISK || *ptr == CHA Line 2703  if (*ptr == CHAR_ASTERISK || *ptr == CHA
2703    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2704      return FALSE;      return FALSE;
2705    
2706  /* Now compare the next item with the previous opcode. If the previous is a  /* Now compare the next item with the previous opcode. First, handle cases when
2707  positive single character match, "item" either contains the character or, if  the next item is a character. */
 "item" is greater than 127 in utf8 mode, the character's bytes are in  
 utf8_char. */  
   
   
 /* Handle cases when the next item is a character. */  
2708    
2709  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2710    {    {
2711    case OP_CHAR:    case OP_CHAR:
2712  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2713    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2714  #else  #else
2715    (void)(utf8_char);  /* Keep compiler happy by referencing function argument */    c = *previous;
2716  #endif  #endif
2717    return item != next;    return c != next;
2718    
2719    /* For CHARNC (caseless character) we must check the other case. If we have    /* For CHARI (caseless character) we must check the other case. If we have
2720    Unicode property support, we can use it to test the other case of    Unicode property support, we can use it to test the other case of
2721    high-valued characters. */    high-valued characters. */
2722    
2723    case OP_CHARNC:    case OP_CHARI:
2724  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2725    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2726    #else
2727      c = *previous;
2728  #endif  #endif
2729    if (item == next) return FALSE;    if (c == next) return FALSE;
2730  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2731    if (utf8)    if (utf8)
2732      {      {
# Line 2336  if (next >= 0) switch(op_code) Line 2737  if (next >= 0) switch(op_code)
2737  #else  #else
2738      othercase = NOTACHAR;      othercase = NOTACHAR;
2739  #endif  #endif
2740      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2741      }      }
2742    else    else
2743  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2744    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2745    
2746    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These
2747      opcodes are not used for multi-byte characters, because they are coded using
2748      an XCLASS instead. */
2749    
2750    case OP_NOT:    case OP_NOT:
2751    if (item == next) return TRUE;    return (c = *previous) == next;
2752    if ((options & PCRE_CASELESS) == 0) return FALSE;  
2753      case OP_NOTI:
2754      if ((c = *previous) == next) return TRUE;
2755  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2756    if (utf8)    if (utf8)
2757      {      {
# Line 2357  if (next >= 0) switch(op_code) Line 2762  if (next >= 0) switch(op_code)
2762  #else  #else
2763      othercase = NOTACHAR;      othercase = NOTACHAR;
2764  #endif  #endif
2765      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2766      }      }
2767    else    else
2768  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2769    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2770    
2771      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2772      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2773    
2774    case OP_DIGIT:    case OP_DIGIT:
2775    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2404  if (next >= 0) switch(op_code) Line 2812  if (next >= 0) switch(op_code)
2812      case 0x202f:      case 0x202f:
2813      case 0x205f:      case 0x205f:
2814      case 0x3000:      case 0x3000:
2815      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2816      default:      default:
2817      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2818      }      }
2819    
2820      case OP_ANYNL:
2821    case OP_VSPACE:    case OP_VSPACE:
2822    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2823    switch(next)    switch(next)
# Line 2420  if (next >= 0) switch(op_code) Line 2829  if (next >= 0) switch(op_code)
2829      case 0x85:      case 0x85:
2830      case 0x2028:      case 0x2028:
2831      case 0x2029:      case 0x2029:
2832      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2833      default:      default:
2834      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2835      }      }
2836    
2837    #ifdef SUPPORT_UCP
2838      case OP_PROP:
2839      return check_char_prop(next, previous[0], previous[1], FALSE);
2840    
2841      case OP_NOTPROP:
2842      return check_char_prop(next, previous[0], previous[1], TRUE);
2843    #endif
2844    
2845    default:    default:
2846    return FALSE;    return FALSE;
2847    }    }
2848    
2849    
2850  /* Handle the case when the next item is \d, \s, etc. */  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
2851    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2852    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2853    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2854    replaced by OP_PROP codes when PCRE_UCP is set. */
2855    
2856  switch(op_code)  switch(op_code)
2857    {    {
2858    case OP_CHAR:    case OP_CHAR:
2859    case OP_CHARNC:    case OP_CHARI:
2860  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2861    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2862    #else
2863      c = *previous;
2864  #endif  #endif
2865    switch(-next)    switch(-next)
2866      {      {
2867      case ESC_d:      case ESC_d:
2868      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2869    
2870      case ESC_D:      case ESC_D:
2871      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2872    
2873      case ESC_s:      case ESC_s:
2874      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2875    
2876      case ESC_S:      case ESC_S:
2877      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2878    
2879      case ESC_w:      case ESC_w:
2880      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2881    
2882      case ESC_W:      case ESC_W:
2883      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2884    
2885      case ESC_h:      case ESC_h:
2886      case ESC_H:      case ESC_H:
2887      switch(item)      switch(c)
2888        {        {
2889        case 0x09:        case 0x09:
2890        case 0x20:        case 0x20:
# Line 2489  switch(op_code) Line 2912  switch(op_code)
2912    
2913      case ESC_v:      case ESC_v:
2914      case ESC_V:      case ESC_V:
2915      switch(item)      switch(c)
2916        {        {
2917        case 0x0a:        case 0x0a:
2918        case 0x0b:        case 0x0b:
# Line 2503  switch(op_code) Line 2926  switch(op_code)
2926        return -next == ESC_v;        return -next == ESC_v;
2927        }        }
2928    
2929        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2930        their substitutions and process them. The result will always be either
2931        -ESC_p or -ESC_P. Then fall through to process those values. */
2932    
2933    #ifdef SUPPORT_UCP
2934        case ESC_du:
2935        case ESC_DU:
2936        case ESC_wu:
2937        case ESC_WU:
2938        case ESC_su:
2939        case ESC_SU:
2940          {
2941          int temperrorcode = 0;
2942          ptr = substitutes[-next - ESC_DU];
2943          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2944          if (temperrorcode != 0) return FALSE;
2945          ptr++;    /* For compatibility */
2946          }
2947        /* Fall through */
2948    
2949        case ESC_p:
2950        case ESC_P:
2951          {
2952          int ptype, pdata, errorcodeptr;
2953          BOOL negated;
2954    
2955          ptr--;      /* Make ptr point at the p or P */
2956          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2957          if (ptype < 0) return FALSE;
2958          ptr++;      /* Point past the final curly ket */
2959    
2960          /* If the property item is optional, we have to give up. (When generated
2961          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2962          to the original \d etc. At this point, ptr will point to a zero byte. */
2963    
2964          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2965            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2966              return FALSE;
2967    
2968          /* Do the property check. */
2969    
2970          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2971          }
2972    #endif
2973    
2974      default:      default:
2975      return FALSE;      return FALSE;
2976      }      }
2977    
2978      /* In principle, support for Unicode properties should be integrated here as
2979      well. It means re-organizing the above code so as to get hold of the property
2980      values before switching on the op-code. However, I wonder how many patterns
2981      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2982      these op-codes are never generated.) */
2983    
2984    case OP_DIGIT:    case OP_DIGIT:
2985    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2986           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
2987    
2988    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2989    return next == -ESC_d;    return next == -ESC_d;
2990    
2991    case OP_WHITESPACE:    case OP_WHITESPACE:
2992    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2993    
2994    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2995    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2996    
2997    case OP_HSPACE:    case OP_HSPACE:
2998    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2999             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3000    
3001    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3002    return next == -ESC_h;    return next == -ESC_h;
3003    
3004    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3005      case OP_ANYNL:
3006    case OP_VSPACE:    case OP_VSPACE:
3007    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3008    
3009    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3010    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
3011    
3012    case OP_WORDCHAR:    case OP_WORDCHAR:
3013    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3014             next == -ESC_v || next == -ESC_R;
3015    
3016    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3017    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2566  Arguments: Line 3043  Arguments:
3043    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3044    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3045    bcptr          points to current branch chain    bcptr          points to current branch chain
3046      cond_depth     conditional nesting depth
3047    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3048    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3049                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2577  Returns:         TRUE on success Line 3055  Returns:         TRUE on success
3055  static BOOL  static BOOL
3056  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3057    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3058    compile_data *cd, int *lengthptr)    int cond_depth, compile_data *cd, int *lengthptr)
3059  {  {
3060  int repeat_type, op_type;  int repeat_type, op_type;
3061  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 2586  int greedy_default, greedy_non_default; Line 3064  int greedy_default, greedy_non_default;
3064  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3065  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3066  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3067  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3068  int after_manual_callout = 0;  int after_manual_callout = 0;
3069  int length_prevgroup = 0;  int length_prevgroup = 0;
3070  register int c;  register int c;
# Line 2598  BOOL inescq = FALSE; Line 3076  BOOL inescq = FALSE;
3076  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3077  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3078  const uschar *tempptr;  const uschar *tempptr;
3079    const uschar *nestptr = NULL;
3080  uschar *previous = NULL;  uschar *previous = NULL;
3081  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3082  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3083  uschar classbits[32];  uschar classbits[32];
3084    
3085    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3086    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3087    dynamically as we process the pattern. */
3088    
3089  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3090  BOOL class_utf8;  BOOL class_utf8;
3091  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
# Line 2611  uschar *class_utf8data_base; Line 3094  uschar *class_utf8data_base;
3094  uschar utf8_char[6];  uschar utf8_char[6];
3095  #else  #else
3096  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3097  #endif  #endif
3098    
3099  #ifdef DEBUG  #ifdef PCRE_DEBUG
3100  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3101  #endif  #endif
3102    
# Line 2662  for (;; ptr++) Line 3144  for (;; ptr++)
3144    int subfirstbyte;    int subfirstbyte;
3145    int terminator;    int terminator;
3146    int mclength;    int mclength;
3147      int tempbracount;
3148    uschar mcbuffer[8];    uschar mcbuffer[8];
3149    
3150    /* Get next byte in the pattern */    /* Get next byte in the pattern */
3151    
3152    c = *ptr;    c = *ptr;
3153    
3154      /* If we are at the end of a nested substitution, revert to the outer level
3155      string. Nesting only happens one level deep. */
3156    
3157      if (c == 0 && nestptr != NULL)
3158        {
3159        ptr = nestptr;
3160        nestptr = NULL;
3161        c = *ptr;
3162        }
3163    
3164    /* If we are in the pre-compile phase, accumulate the length used for the    /* If we are in the pre-compile phase, accumulate the length used for the
3165    previous cycle of this loop. */    previous cycle of this loop. */
3166    
3167    if (lengthptr != NULL)    if (lengthptr != NULL)
3168      {      {
3169  #ifdef DEBUG  #ifdef PCRE_DEBUG
3170      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3171  #endif  #endif
3172      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3173        {        {
3174        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3175        goto FAILED;        goto FAILED;
# Line 2698  for (;; ptr++) Line 3191  for (;; ptr++)
3191        goto FAILED;        goto FAILED;
3192        }        }
3193    
3194      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3195      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
3196    
3197      /* If "previous" is set and it is not at the start of the work space, move      /* If "previous" is set and it is not at the start of the work space, move
# Line 2725  for (;; ptr++) Line 3218  for (;; ptr++)
3218    /* In the real compile phase, just check the workspace used by the forward    /* In the real compile phase, just check the workspace used by the forward
3219    reference list. */    reference list. */
3220    
3221    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3222      {      {
3223      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3224      goto FAILED;      goto FAILED;
# Line 2773  for (;; ptr++) Line 3266  for (;; ptr++)
3266      previous_callout = NULL;      previous_callout = NULL;
3267      }      }
3268    
3269    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3270    
3271    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3272      {      {
3273      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3274      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
3275        {        {
3276        while (*(++ptr) != 0)        ptr++;
3277          while (*ptr != 0)
3278          {          {
3279          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3280            ptr++;
3281    #ifdef SUPPORT_UTF8
3282            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3283    #endif
3284          }          }
3285        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3286    
# Line 2816  for (;; ptr++) Line 3314  for (;; ptr++)
3314          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3315          goto FAILED;          goto FAILED;
3316          }          }
3317        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3318        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3319        }        }
3320      return TRUE;      return TRUE;
# Line 2827  for (;; ptr++) Line 3325  for (;; ptr++)
3325      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3326    
3327      case CHAR_CIRCUMFLEX_ACCENT:      case CHAR_CIRCUMFLEX_ACCENT:
3328        previous = NULL;
3329      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3330        {        {
3331        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3332          *code++ = OP_CIRCM;
3333        }        }
3334      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3335      break;      break;
3336    
3337      case CHAR_DOLLAR_SIGN:      case CHAR_DOLLAR_SIGN:
3338      previous = NULL;      previous = NULL;
3339      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3340      break;      break;
3341    
3342      /* There can never be a first char if '.' is first, whatever happens about      /* There can never be a first char if '.' is first, whatever happens about
# Line 3021  for (;; ptr++) Line 3520  for (;; ptr++)
3520            ptr++;            ptr++;
3521            }            }
3522    
3523          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3524          if (posix_class < 0)          if (posix_class < 0)
3525            {            {
3526            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 3035  for (;; ptr++) Line 3534  for (;; ptr++)
3534          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3535            posix_class = 0;            posix_class = 0;
3536    
3537          /* We build the bit map for the POSIX class in a chunk of local store          /* When PCRE_UCP is set, some of the POSIX classes are converted to
3538          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3539          subtract bits that may be in the main map already. At the end we or the  
3540          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3541            if ((options & PCRE_UCP) != 0)
3542              {
3543              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3544              if (posix_substitutes[pc] != NULL)
3545                {
3546                nestptr = tempptr + 1;
3547                ptr = posix_substitutes[pc] - 1;
3548                continue;
3549                }
3550              }
3551    #endif
3552            /* In the non-UCP case, we build the bit map for the POSIX class in a
3553            chunk of local store because we may be adding and subtracting from it,
3554            and we don't want to subtract bits that may be in the main map already.
3555            At the end we or the result into the bit map that is being built. */
3556    
3557          posix_class *= 3;          posix_class *= 3;
3558    
# Line 3082  for (;; ptr++) Line 3596  for (;; ptr++)
3596    
3597        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3598        of the specials, which just set a flag. The sequence \b is a special        of the specials, which just set a flag. The sequence \b is a special
3599        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace. We
3600        Elsewhere it marks a word boundary. Other escapes have preset maps ready        assume that other escapes have more than one character in them, so set
3601        to 'or' into the one we are building. We assume they have more than one        class_charcount bigger than one. Unrecognized escapes fall through and
3602        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3603          PCRE_EXTRA is set. */
3604    
3605        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
3606          {          {
3607          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3608          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3609    
3610          if (-c == ESC_b) c = CHAR_BS;       /* \b is backspace in a class */          if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */
         else if (-c == ESC_X) c = CHAR_X;   /* \X is literal X in a class */  
         else if (-c == ESC_R) c = CHAR_R;   /* \R is literal R in a class */  
3611          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3612            {            {
3613            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
# Line 3111  for (;; ptr++) Line 3624  for (;; ptr++)
3624            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3625            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3626    
3627            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3628              {              {
3629    #ifdef SUPPORT_UCP
3630                case ESC_du:     /* These are the values given for \d etc */
3631                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3632                case ESC_wu:     /* escape sequence with an appropriate \p */
3633                case ESC_WU:     /* or \P to test Unicode properties instead */
3634                case ESC_su:     /* of the default ASCII testing. */
3635                case ESC_SU:
3636                nestptr = ptr;
3637                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3638                class_charcount -= 2;                /* Undo! */
3639                continue;
3640    #endif
3641              case ESC_d:              case ESC_d:
3642              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3643              continue;              continue;
# Line 3133  for (;; ptr++) Line 3656  for (;; ptr++)
3656              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3657              continue;              continue;
3658    
3659                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3660                if it was previously set by something earlier in the character
3661                class. */
3662    
3663              case ESC_s:              case ESC_s:
3664              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3665              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3666                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3667              continue;              continue;
3668    
3669              case ESC_S:              case ESC_S:
# Line 3144  for (;; ptr++) Line 3672  for (;; ptr++)
3672              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3673              continue;              continue;
3674    
3675              default:    /* Not recognized; fall through */              case ESC_h:
             break;      /* Need "default" setting to stop compiler warning. */  
             }  
   
           /* In the pre-compile phase, just do the recognition. */  
   
           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)  
             {  
3676              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3677              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3678              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 3181  for (;; ptr++) Line 3696  for (;; ptr++)
3696                }                }
3697  #endif  #endif
3698              continue;              continue;
             }  
3699    
3700            if (-c == ESC_H)              case ESC_H:
             {  
3701              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3702                {                {
3703                int x = 0xff;                int x = 0xff;
# Line 3226  for (;; ptr++) Line 3739  for (;; ptr++)
3739                }                }
3740  #endif  #endif
3741              continue;              continue;
             }  
3742    
3743            if (-c == ESC_v)              case ESC_v:
             {  
3744              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3745              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3746              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 3245  for (;; ptr++) Line 3756  for (;; ptr++)
3756                }                }
3757  #endif  #endif
3758              continue;              continue;
             }  
3759    
3760            if (-c == ESC_V)              case ESC_V:
             {  
3761              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3762                {                {
3763                int x = 0xff;                int x = 0xff;
# Line 3278  for (;; ptr++) Line 3787  for (;; ptr++)
3787                }                }
3788  #endif  #endif
3789              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3790    
3791  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3792            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3793              {              case ESC_P:
3794              BOOL negated;                {
3795              int pdata;                BOOL negated;
3796              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3797              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3798              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3799              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3800                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3801              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3802              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3803              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3804              continue;                class_charcount -= 2;   /* Not a < 256 character */
3805              }                continue;
3806                  }
3807  #endif  #endif
3808            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3809            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3810            treated as literals. */              treated as literals. */
3811    
3812            if ((options & PCRE_EXTRA) != 0)              default:
3813              {              if ((options & PCRE_EXTRA) != 0)
3814              *errorcodeptr = ERR7;                {
3815              goto FAILED;                *errorcodeptr = ERR7;
3816                  goto FAILED;
3817                  }
3818                class_charcount -= 2;  /* Undo the default count from above */
3819                c = *ptr;              /* Get the final character and fall through */
3820                break;
3821              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3822            }            }
3823    
3824          /* Fall through if we have a single character (c >= 0). This may be          /* Fall through if we have a single character (c >= 0). This may be
# Line 3379  for (;; ptr++) Line 3888  for (;; ptr++)
3888            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3889            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3890    
3891            /* \b is backspace; \X is literal X; \R is literal R; any other            /* \b is backspace; any other special means the '-' was literal */
           special means the '-' was literal */  
3892    
3893            if (d < 0)            if (d < 0)
3894              {              {
3895              if (d == -ESC_b) d = CHAR_BS;              if (d == -ESC_b) d = CHAR_BS; else
             else if (d == -ESC_X) d = CHAR_X;  
             else if (d == -ESC_R) d = CHAR_R; else  
3896                {                {
3897                ptr = oldptr;                ptr = oldptr;
3898                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3552  for (;; ptr++) Line 4058  for (;; ptr++)
4058          }          }
4059        }        }
4060    
4061      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" far above.
4062        If we are at the end of an internal nested string, revert to the outer
4063        string. */
4064    
4065        while (((c = *(++ptr)) != 0 ||
4066               (nestptr != NULL &&
4067                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4068               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4069    
4070      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));      /* Check for missing terminating ']' */
4071    
4072      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4073        {        {
4074        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4075        goto FAILED;        goto FAILED;
4076        }        }
4077    
   
 /* 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  
   
   
4078      /* If class_charcount is 1, we saw precisely one character whose value is      /* If class_charcount is 1, we saw precisely one character whose value is
4079      less than 256. As long as there were no characters >= 128 and there was no      less than 256. As long as there were no characters >= 128 and there was no
4080      use of \p or \P, in other words, no use of any XCLASS features, we can      use of \p or \P, in other words, no use of any XCLASS features, we can
# Line 3588  we set the flag only if there is a liter Line 4082  we set the flag only if there is a liter
4082    
4083      In UTF-8 mode, we can optimize the negative case only if there were no      In UTF-8 mode, we can optimize the negative case only if there were no
4084      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4085      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4086      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4087    
4088      The optimization throws away the bit map. We turn the item into a      The optimization throws away the bit map. We turn the item into a
4089      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[I] if it's positive, or OP_NOT[I] if it's negative.
4090      that OP_NOT does not support multibyte characters. In the positive case, it      Note that OP_NOT[I] does not support multibyte characters. In the positive
4091      can cause firstbyte to be set. Otherwise, there can be no first char if      case, it can cause firstbyte to be set. Otherwise, there can be no first
4092      this item is first, whatever repeat count may follow. In the case of      char if this item is first, whatever repeat count may follow. In the case
4093      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4094    
4095  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4096      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
# Line 3607  we set the flag only if there is a liter Line 4101  we set the flag only if there is a liter
4101        {        {
4102        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4103    
4104        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4105    
4106        if (negate_class)        if (negate_class)
4107          {          {
4108          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4109          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4110          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4111          *code++ = class_lastchar;          *code++ = class_lastchar;
4112          break;          break;
4113          }          }
# Line 3644  we set the flag only if there is a liter Line 4138  we set the flag only if there is a liter
4138    
4139      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4140      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
4141      such as \S in the class, because in that case all characters > 255 are in      such as \S in the class, and PCRE_UCP is not set, because in that case all
4142      the class, so any that were explicitly given as well can be ignored. If      characters > 255 are in the class, so any that were explicitly given as
4143      (when there are explicit characters > 255 that must be listed) there are no      well can be ignored. If (when there are explicit characters > 255 that must
4144      characters < 256, we can omit the bitmap in the actual compiled code. */      be listed) there are no characters < 256, we can omit the bitmap in the
4145        actual compiled code. */
4146    
4147  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4148      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4149        {        {
4150        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4151        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3676  we set the flag only if there is a liter Line 4171  we set the flag only if there is a liter
4171        }        }
4172  #endif  #endif
4173    
4174      /* If there are no characters > 255, set the opcode to OP_CLASS or      /* If there are no characters > 255, or they are all to be included or
4175      OP_NCLASS, depending on whether the whole class was negated and whether      excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
4176      there were negative specials such as \S in the class. Then copy the 32-byte      whole class was negated and whether there were negative specials such as \S
4177      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4178        negating it if necessary. */
4179    
4180      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4181      if (negate_class)      if (negate_class)
# Line 3739  we set the flag only if there is a liter Line 4235  we set the flag only if there is a liter
4235      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4236      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4237    
4238      /* 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
4239      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4240    
4241      tempcode = previous;      tempcode = previous;
4242    
# Line 3763  we set the flag only if there is a liter Line 4259  we set the flag only if there is a liter
4259        }        }
4260      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4261    
4262        /* If previous was a recursion call, wrap it in atomic brackets so that
4263        previous becomes the atomic group. All recursions were so wrapped in the
4264        past, but it no longer happens for non-repeated recursions. In fact, the
4265        repeated ones could be re-implemented independently so as not to need this,
4266        but for the moment we rely on the code for repeating groups. */
4267    
4268        if (*previous == OP_RECURSE)
4269          {
4270          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4271          *previous = OP_ONCE;
4272          PUT(previous, 1, 2 + 2*LINK_SIZE);
4273          previous[2 + 2*LINK_SIZE] = OP_KET;
4274          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4275          code += 2 + 2 * LINK_SIZE;
4276          length_prevgroup = 3 + 3*LINK_SIZE;
4277    
4278          /* When actually compiling, we need to check whether this was a forward
4279          reference, and if so, adjust the offset. */
4280    
4281          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4282            {
4283            int offset = GET(cd->hwm, -LINK_SIZE);
4284            if (offset == previous + 1 - cd->start_code)
4285              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4286            }
4287          }
4288    
4289        /* Now handle repetition for the different types of item. */
4290    
4291      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4292      repeat item instead. If a char item has a minumum of more than one, ensure      repeat item instead. If a char item has a minumum of more than one, ensure
4293      that it is set in reqbyte - it might not be if a sequence such as x{3} is      that it is set in reqbyte - it might not be if a sequence such as x{3} is
4294      the first thing in a branch because the x will have gone into firstbyte      the first thing in a branch because the x will have gone into firstbyte
4295      instead.  */      instead.  */
4296    
4297      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4298        {        {
4299          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4300    
4301        /* Deal with UTF-8 characters that take up more than one byte. It's        /* Deal with UTF-8 characters that take up more than one byte. It's
4302        easier to write this out separately than try to macrify it. Use c to        easier to write this out separately than try to macrify it. Use c to
4303        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 0x80 to flag that it's a
# Line 3803  we set the flag only if there is a liter Line 4330  we set the flag only if there is a liter
4330    
4331        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4332            repeat_max < 0 &&            repeat_max < 0 &&
4333            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4334          {          {
4335          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4336          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3816  we set the flag only if there is a liter Line 4342  we set the flag only if there is a liter
4342      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4343      one of the special opcodes, replacing it. The code is shared with single-      one of the special opcodes, replacing it. The code is shared with single-
4344      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4345      repeat_type. We can also test for auto-possessification. OP_NOT is      repeat_type. We can also test for auto-possessification. OP_NOT and OP_NOTI
4346      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4347    
4348      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4349        {        {
4350        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4351        c = previous[1];        c = previous[1];
4352        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4353            repeat_max < 0 &&            repeat_max < 0 &&
4354            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4355          {          {
4356          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4357          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3849  we set the flag only if there is a liter Line 4375  we set the flag only if there is a liter
4375    
4376        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4377            repeat_max < 0 &&            repeat_max < 0 &&
4378            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4379          {          {
4380          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4381          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3871  we set the flag only if there is a liter Line 4397  we set the flag only if there is a liter
4397    
4398        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4399    
4400        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
4401        /* This code is obsolete from release 8.00; the restriction was finally        /* This code is obsolete from release 8.00; the restriction was finally
4402        removed: */        removed: */
4403    
4404        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4405        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4406    
4407        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4408        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
4409    
4410        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4411    
# Line 4018  we set the flag only if there is a liter Line 4544  we set the flag only if there is a liter
4544  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4545               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4546  #endif  #endif
4547               *previous == OP_REF)               *previous == OP_REF ||
4548                 *previous == OP_REFI)
4549        {        {
4550        if (repeat_max == 0)        if (repeat_max == 0)
4551          {          {
# Line 4026  we set the flag only if there is a liter Line 4553  we set the flag only if there is a liter
4553          goto END_REPEAT;          goto END_REPEAT;
4554          }          }
4555    
4556        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
4557        /* This code is obsolete from release 8.00; the restriction was finally        /* This code is obsolete from release 8.00; the restriction was finally
4558        removed: */        removed: */
4559    
# Line 4034  we set the flag only if there is a liter Line 4561  we set the flag only if there is a liter
4561        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4562    
4563        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4564        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
4565    
4566        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4567          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 4052  we set the flag only if there is a liter Line 4579  we set the flag only if there is a liter
4579        }        }
4580    
4581      /* If previous was a bracket group, we may have to replicate it in certain      /* If previous was a bracket group, we may have to replicate it in certain
4582      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
4583        opcodes such as BRA and CBRA, as this is the place where they get converted
4584        into the more special varieties such as BRAPOS and SBRA. A test for >=
4585        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4586        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4587        repetition of assertions, but now it does, for Perl compatibility. */
4588    
4589      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4590        {        {
4591        register int i;        register int i;
4592        int ketoffset = 0;        int len = (int)(code - previous);
       int len = code - previous;  
4593        uschar *bralink = NULL;        uschar *bralink = NULL;
4594          uschar *brazeroptr = NULL;
4595    
4596        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4597          we just ignore the repeat. */
4598    
4599        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4600          {          goto END_REPEAT;
         *errorcodeptr = ERR55;  
         goto FAILED;  
         }  
4601    
4602        /* If the maximum repeat count is unlimited, find the end of the bracket        /* There is no sense in actually repeating assertions. The only potential
4603        by scanning through from the start, and compute the offset back to it        use of repetition is in cases when the assertion is optional. Therefore,
4604        from the current code pointer. There may be an OP_OPT setting following        if the minimum is greater than zero, just ignore the repeat. If the
4605        the final KET, so we can't find the end just by going back from the code        maximum is not not zero or one, set it to 1. */
4606        pointer. */  
4607          if (*previous < OP_ONCE)    /* Assertion */
4608        if (repeat_max == -1)          {
4609          {          if (repeat_min > 0) goto END_REPEAT;
4610          register uschar *ket = previous;          if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
         do ket += GET(ket, 1); while (*ket != OP_KET);  
         ketoffset = code - ket;  
4611          }          }
4612    
4613        /* The case of a zero minimum is special because of the need to stick        /* The case of a zero minimum is special because of the need to stick
# Line 4101  we set the flag only if there is a liter Line 4628  we set the flag only if there is a liter
4628          **   goto END_REPEAT;          **   goto END_REPEAT;
4629          **   }          **   }
4630    
4631          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
4632          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
4633          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
4634          groups are referenced, we cannot do this selectively.          don't have a list of which groups are referenced, we cannot do this
4635            selectively.
4636    
4637          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4638          and do no more at this point. However, we do need to adjust any          and do no more at this point. However, we do need to adjust any
# Line 4124  we set the flag only if there is a liter Line 4652  we set the flag only if there is a liter
4652              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
4653              goto END_REPEAT;              goto END_REPEAT;
4654              }              }
4655              brazeroptr = previous;    /* Save for possessive optimizing */
4656            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4657            }            }
4658    
# Line 4148  we set the flag only if there is a liter Line 4677  we set the flag only if there is a liter
4677            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4678            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4679    
4680            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4681            bralink = previous;            bralink = previous;
4682            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4683            }            }
# Line 4169  we set the flag only if there is a liter Line 4698  we set the flag only if there is a liter
4698            {            {
4699            /* In the pre-compile phase, we don't actually do the replication. We            /* In the pre-compile phase, we don't actually do the replication. We
4700            just adjust the length as if we had. Do some paranoid checks for            just adjust the length as if we had. Do some paranoid checks for
4701            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4702              integer type when available, otherwise double. */
4703    
4704            if (lengthptr != NULL)            if (lengthptr != NULL)
4705              {              {
4706              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4707              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4708                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4709                        (INT64_OR_DOUBLE)INT_MAX ||
4710                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4711                {                {
4712                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 4221  we set the flag only if there is a liter Line 4752  we set the flag only if there is a liter
4752          just adjust the length as if we had. For each repetition we must add 1          just adjust the length as if we had. For each repetition we must add 1
4753          to the length for BRAZERO and for all but the last repetition we must          to the length for BRAZERO and for all but the last repetition we must
4754          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4755          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4756            a 64-bit integer type when available, otherwise double. */
4757    
4758          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4759            {            {
4760            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4761                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4762            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4763                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4764                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4765                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4766              {              {
4767              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 4254  we set the flag only if there is a liter Line 4786  we set the flag only if there is a liter
4786              {              {
4787              int offset;              int offset;
4788              *code++ = OP_BRA;              *code++ = OP_BRA;
4789              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4790              bralink = code;              bralink = code;
4791              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4792              }              }
# Line 4275  we set the flag only if there is a liter Line 4807  we set the flag only if there is a liter
4807          while (bralink != NULL)          while (bralink != NULL)
4808            {            {
4809            int oldlinkoffset;            int oldlinkoffset;
4810            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4811            uschar *bra = code - offset;            uschar *bra = code - offset;
4812            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4813            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 4285  we set the flag only if there is a liter Line 4817  we set the flag only if there is a liter
4817            }            }
4818          }          }
4819    
4820        /* If the maximum is unlimited, set a repeater in the final copy. We        /* If the maximum is unlimited, set a repeater in the final copy. For
4821        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
4822        don't know if there's been an options resetting after the ket. The        ONCE brackets can be converted into non-capturing brackets, as the
4823        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
4824          deal with possessive ONCEs specially.
4825    
4826          Otherwise, if the quantifier was possessive, we convert the BRA code to
4827          the POS form, and the KET code to KETRPOS. (It turns out to be convenient
4828          at runtime to detect this kind of subpattern at both the start and at the
4829          end.) The use of special opcodes makes it possible to reduce greatly the
4830          stack usage in pcre_exec(). If the group is preceded by OP_BRAZERO,
4831          convert this to OP_BRAPOSZERO. Then cancel the possessive flag so that
4832          the default action below, of wrapping everything inside atomic brackets,
4833          does not happen.
4834    
4835        Then, when we are doing the actual compile phase, check to see whether        Then, when we are doing the actual compile phase, check to see whether
4836        this group is a non-atomic one that could match an empty string. If so,        this group is one that could match an empty string. If so, convert the
4837        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so        initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so that runtime
4838        that runtime checking can be done. [This check is also applied to        checking can be done. [This check is also applied to ONCE groups at
4839        atomic groups at runtime, but in a different way.] */        runtime, but in a different way.] */
4840    
4841        else        else
4842          {          {
4843          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
4844          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
4845          *ketcode = OP_KETRMAX + repeat_type;  
4846          if (lengthptr == NULL && *bracode != OP_ONCE)          if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
4847                possessive_quantifier) *bracode = OP_BRA;
4848    
4849            if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
4850              *ketcode = OP_KETRMAX + repeat_type;
4851            else
4852            {            {
4853            uschar *scode = bracode;            if (possessive_quantifier)
           do  
4854              {              {
4855              if (could_be_empty_branch(scode, ketcode, utf8))              *bracode += 1;                   /* Switch to xxxPOS opcodes */
4856                *ketcode = OP_KETRPOS;
4857                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
4858                possessive_quantifier = FALSE;
4859                }
4860              else *ketcode = OP_KETRMAX + repeat_type;
4861    
4862              if (lengthptr == NULL)
4863                {
4864                uschar *scode = bracode;
4865                do
4866                {                {
4867                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
4868                break;                  {
4869                    *bracode += OP_SBRA - OP_BRA;
4870                    break;
4871                    }
4872                  scode += GET(scode, 1);
4873                }                }
4874              scode += GET(scode, 1);              while (*scode == OP_ALT);
4875              }              }
           while (*scode == OP_ALT);  
4876            }            }
4877          }          }
4878        }        }
# Line 4334  we set the flag only if there is a liter Line 4893  we set the flag only if there is a liter
4893        }        }
4894    
4895      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
4896      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
4897      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
4898      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
4899      The '+' notation is just syntactic sugar, taken from Sun's Java package,      notation is just syntactic sugar, taken from Sun's Java package, but the
4900      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
4901      tempcode, not at previous, which might be the first part of a string whose  
4902      (former) last char we repeated.      Possessively repeated subpatterns have already been handled in the code
4903        just above, so possessive_quantifier is always FALSE for them at this
4904        stage.
4905    
4906        Note that the repeated item starts at tempcode, not at previous, which
4907        might be the first part of a string whose (former) last char we repeated.
4908    
4909      Possessifying an 'exact' quantifier has no effect, so we can ignore it. But      Possessifying an 'exact' quantifier has no effect, so we can ignore it. But
4910      an 'upto' may follow. We skip over an 'exact' item, and then test the      an 'upto' may follow. We skip over an 'exact' item, and then test the
# Line 4349  we set the flag only if there is a liter Line 4913  we set the flag only if there is a liter
4913      if (possessive_quantifier)      if (possessive_quantifier)
4914        {        {
4915        int len;        int len;
4916    
4917        if (*tempcode == OP_TYPEEXACT)        if (*tempcode == OP_TYPEEXACT)
4918          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += _pcre_OP_lengths[*tempcode] +
4919            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4920    
4921        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4922          {          {
4923          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
# Line 4361  we set the flag only if there is a liter Line 4925  we set the flag only if there is a liter
4925          if (utf8 && tempcode[-1] >= 0xc0)          if (utf8 && tempcode[-1] >= 0xc0)
4926            tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];            tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4927  #endif  #endif
4928          }          }
4929    
4930        len = code - tempcode;        len = (int)(code - tempcode);
4931        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4932          {          {
4933          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 4371  we set the flag only if there is a liter Line 4935  we set the flag only if there is a liter
4935          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
4936          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4937    
4938          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
4939          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
4940          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
4941          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
4942    
4943          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4944          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4945          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4946          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4947    
4948            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
4949            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
4950            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
4951            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
4952    
4953            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
4954            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
4955            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
4956            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
4957    
4958            /* Because we are moving code along, we must ensure that any
4959            pending recursive references are updated. */
4960    
4961          default:          default:
4962            *code = OP_END;
4963            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4964          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4965          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4966          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4417  we set the flag only if there is a liter Line 4996  we set the flag only if there is a liter
4996    
4997      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4998    
4999      if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
5000             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
5001        {        {
5002        int i, namelen;        int i, namelen;
5003          int arglen = 0;
5004        const char *vn = verbnames;        const char *vn = verbnames;
5005        const uschar *name = ++ptr;        const uschar *name = ptr + 1;
5006          const uschar *arg = NULL;
5007        previous = NULL;        previous = NULL;
5008        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
5009          namelen = (int)(ptr - name);
5010    
5011          /* It appears that Perl allows any characters whatsoever, other than
5012          a closing parenthesis, to appear in arguments, so we no longer insist on
5013          letters, digits, and underscores. */
5014    
5015        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
5016          {          {
5017          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
5018          goto FAILED;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
5019            arglen = (int)(ptr - arg);
5020          }          }
5021    
5022        if (*ptr != CHAR_RIGHT_PARENTHESIS)        if (*ptr != CHAR_RIGHT_PARENTHESIS)
5023          {          {
5024          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
5025          goto FAILED;          goto FAILED;
5026          }          }
5027        namelen = ptr - name;  
5028          /* Scan the table of verb names */
5029    
5030        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
5031          {          {
5032          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
5033              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
5034            {            {
5035            *code = verbs[i].op;            /* Check for open captures before ACCEPT and convert it to
5036            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;            ASSERT_ACCEPT if in an assertion. */
5037            break;  
5038              if (verbs[i].op == OP_ACCEPT)
5039                {
5040                open_capitem *oc;
5041                if (arglen != 0)
5042                  {
5043                  *errorcodeptr = ERR59;
5044                  goto FAILED;
5045                  }
5046                cd->had_accept = TRUE;
5047                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5048                  {
5049                  *code++ = OP_CLOSE;
5050                  PUT2INC(code, 0, oc->number);
5051                  }
5052                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5053    
5054                /* Do not set firstbyte after *ACCEPT */
5055                if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
5056                }
5057    
5058              /* Handle other cases with/without an argument */
5059    
5060              else if (arglen == 0)
5061                {
5062                if (verbs[i].op < 0)   /* Argument is mandatory */
5063                  {
5064                  *errorcodeptr = ERR66;
5065                  goto FAILED;
5066                  }
5067                *code = verbs[i].op;
5068                if (*code++ == OP_THEN) cd->external_flags |= PCRE_HASTHEN;
5069                }
5070    
5071              else
5072                {
5073                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
5074                  {
5075                  *errorcodeptr = ERR59;
5076                  goto FAILED;
5077                  }
5078                *code = verbs[i].op_arg;
5079                if (*code++ == OP_THEN_ARG) cd->external_flags |= PCRE_HASTHEN;
5080                *code++ = arglen;
5081                memcpy(code, arg, arglen);
5082                code += arglen;
5083                *code++ = 0;
5084                }
5085    
5086              break;  /* Found verb, exit loop */
5087            }            }
5088    
5089          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
5090          }          }
5091        if (i < verbcount) continue;  
5092        *errorcodeptr = ERR60;        if (i < verbcount) continue;    /* Successfully handled a verb */
5093          *errorcodeptr = ERR60;          /* Verb not recognized */
5094        goto FAILED;        goto FAILED;
5095        }        }
5096    
# Line 4566  we set the flag only if there is a liter Line 5209  we set the flag only if there is a liter
5209                recno * 10 + *ptr - CHAR_0 : -1;                recno * 10 + *ptr - CHAR_0 : -1;
5210            ptr++;            ptr++;
5211            }            }
5212          namelen = ptr - name;          namelen = (int)(ptr - name);
5213    
5214          if ((terminator > 0 && *ptr++ != terminator) ||          if ((terminator > 0 && *ptr++ != terminator) ||
5215              *ptr++ != CHAR_RIGHT_PARENTHESIS)              *ptr++ != CHAR_RIGHT_PARENTHESIS)
# Line 4603  we set the flag only if there is a liter Line 5246  we set the flag only if there is a liter
5246            }            }
5247    
5248          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
5249          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
5250            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
5251            except they record that the reference was originally to a name. The
5252            information is used to check duplicate names. */
5253    
5254          slot = cd->name_table;          slot = cd->name_table;
5255          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 4618  we set the flag only if there is a liter Line 5264  we set the flag only if there is a liter
5264            {            {
5265            recno = GET2(slot, 0);            recno = GET2(slot, 0);
5266            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
5267              code[1+LINK_SIZE]++;
5268            }            }
5269    
5270          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5271    
5272          else if ((i = find_parens(cd, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5273                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0, utf8)) > 0)
5274            {            {
5275            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5276              code[1+LINK_SIZE]++;
5277            }            }
5278    
5279          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 4690  we set the flag only if there is a liter Line 5338  we set the flag only if there is a liter
5338          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5339          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5340          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5341            cd->assert_depth += 1;
5342          ptr++;          ptr++;
5343          break;          break;
5344    
# Line 4704  we set the flag only if there is a liter Line 5353  we set the flag only if there is a liter
5353            continue;            continue;
5354            }            }
5355          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5356            cd->assert_depth += 1;
5357          break;          break;
5358    
5359    
# Line 4713  we set the flag only if there is a liter Line 5363  we set the flag only if there is a liter
5363            {            {
5364            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5365            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5366              cd->assert_depth += 1;
5367            ptr += 2;            ptr += 2;
5368            break;            break;
5369    
5370            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5371            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5372              cd->assert_depth += 1;
5373            ptr += 2;            ptr += 2;
5374            break;            break;
5375    
# Line 4757  we set the flag only if there is a liter Line 5409  we set the flag only if there is a liter
5409              goto FAILED;              goto FAILED;
5410              }              }
5411            *code++ = n;            *code++ = n;
5412            PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */            PUT(code, 0, (int)(ptr - cd->start_pattern + 1)); /* Pattern offset */
5413            PUT(code, LINK_SIZE, 0);                    /* Default length */            PUT(code, LINK_SIZE, 0);                          /* Default length */
5414            code += 2 * LINK_SIZE;            code += 2 * LINK_SIZE;
5415            }            }
5416          previous = NULL;          previous = NULL;
# Line 4791  we set the flag only if there is a liter Line 5443  we set the flag only if there is a liter
5443            name = ++ptr;            name = ++ptr;
5444    
5445            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5446            namelen = ptr - name;            namelen = (int)(ptr - name);
5447    
5448            /* In the pre-compile phase, just do a syntax check. */            /* In the pre-compile phase, just do a syntax check. */
5449    
# Line 4818  we set the flag only if there is a liter Line 5470  we set the flag only if there is a liter
5470                }                }
5471              }              }
5472    
5473            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
5474              alphabetical order. Duplicate names for different numbers are
5475              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
5476              number are always OK. (An existing number can be re-used if (?|
5477              appears in the pattern.) In either event, a duplicate name results in
5478              a duplicate entry in the table, even if the number is the same. This
5479              is because the number of names, and hence the table size, is computed
5480              in the pre-compile, and it affects various numbers and pointers which
5481              would all have to be modified, and the compiled code moved down, if
5482              duplicates with the same number were omitted from the table. This
5483              doesn't seem worth the hassle. However, *different* names for the
5484              same number are not permitted. */
5485    
5486            else            else
5487              {              {
5488                BOOL dupname = FALSE;
5489              slot = cd->name_table;              slot = cd->name_table;
5490    
5491              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
5492                {                {
5493                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 4830  we set the flag only if there is a liter Line 5495  we set the flag only if there is a liter
5495                  {                  {
5496                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
5497                    {                    {
5498                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
5499                          (options & PCRE_DUPNAMES) == 0)
5500                      {                      {
5501                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
5502                      goto FAILED;                      goto FAILED;
5503                      }                      }
5504                      else dupname = TRUE;
5505                    }                    }
5506                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
5507                  }                  }
5508    
5509                  /* Make space in the table and break the loop for an earlier
5510                  name. For a duplicate or later name, carry on. We do this for
5511                  duplicates so that in the simple case (when ?(| is not used) they
5512                  are in order of their numbers. */
5513    
5514                if (crc < 0)                if (crc < 0)
5515                  {                  {
5516                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
5517                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
5518                  break;                  break;
5519                  }                  }
5520    
5521                  /* Continue the loop for a later or duplicate name */
5522    
5523                slot += cd->name_entry_size;                slot += cd->name_entry_size;
5524                }                }
5525    
5526                /* For non-duplicate names, check for a duplicate number before
5527                adding the new name. */
5528    
5529                if (!dupname)
5530                  {
5531                  uschar *cslot = cd->name_table;
5532                  for (i = 0; i < cd->names_found; i++)
5533                    {
5534                    if (cslot != slot)
5535                      {
5536                      if (GET2(cslot, 0) == cd->bracount + 1)
5537                        {
5538                        *errorcodeptr = ERR65;
5539                        goto FAILED;
5540                        }
5541                      }
5542                    else i--;
5543                    cslot += cd->name_entry_size;
5544                    }
5545                  }
5546    
5547              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
5548              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
5549              slot[2+namelen] = 0;              slot[2+namelen] = 0;
5550              }              }
5551            }            }
5552    
5553          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
5554            encountered. */
5555    
         ptr++;                    /* Move past > or ' */  
5556          cd->names_found++;          cd->names_found++;
5557            ptr++;                    /* Move past > or ' */
5558          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
5559    
5560    
# Line 4875  we set the flag only if there is a liter Line 5573  we set the flag only if there is a liter
5573          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
5574          name = ++ptr;          name = ++ptr;
5575          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5576          namelen = ptr - name;          namelen = (int)(ptr - name);
5577    
5578          /* In the pre-compile phase, do a syntax check and set a dummy          /* In the pre-compile phase, do a syntax check. We used to just set
5579          reference number. */          a dummy reference number, because it was not used in the first pass.
5580            However, with the change of recursive back references to be atomic,
5581            we have to look for the number so that this state can be identified, as
5582            otherwise the incorrect length is computed. If it's not a backwards
5583            reference, the dummy number will do. */
5584    
5585          if (lengthptr != NULL)          if (lengthptr != NULL)
5586            {            {
5587              const uschar *temp;
5588    
5589            if (namelen == 0)            if (namelen == 0)
5590              {              {
5591              *errorcodeptr = ERR62;              *errorcodeptr = ERR62;
# Line 4897  we set the flag only if there is a liter Line 5601  we set the flag only if there is a liter
5601              *errorcodeptr = ERR48;              *errorcodeptr = ERR48;
5602              goto FAILED;              goto FAILED;
5603              }              }
5604            recno = 0;  
5605              /* The name table does not exist in the first pass, so we cannot
5606              do a simple search as in the code below. Instead, we have to scan the
5607              pattern to find the number. It is important that we scan it only as
5608              far as we have got because the syntax of named subpatterns has not
5609              been checked for the rest of the pattern, and find_parens() assumes
5610              correct syntax. In any case, it's a waste of resources to scan
5611              further. We stop the scan at the current point by temporarily
5612              adjusting the value of cd->endpattern. */
5613    
5614              temp = cd->end_pattern;
5615              cd->end_pattern = ptr;
5616              recno = find_parens(cd, name, namelen,
5617                (options & PCRE_EXTENDED) != 0, utf8);
5618              cd->end_pattern = temp;
5619              if (recno < 0) recno = 0;    /* Forward ref; set dummy number */
5620            }            }
5621    
5622          /* In the real compile, seek the name in the table. We check the name          /* In the real compile, seek the name in the table. We check the name
# Line 4922  we set the flag only if there is a liter Line 5641  we set the flag only if there is a liter
5641              }              }
5642            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5643                      find_parens(cd, name, namelen,                      find_parens(cd, name, namelen,
5644                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0, utf8)) <= 0)
5645              {              {
5646              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
5647              goto FAILED;              goto FAILED;
# Line 5025  we set the flag only if there is a liter Line 5744  we set the flag only if there is a liter
5744            if (lengthptr == NULL)            if (lengthptr == NULL)
5745              {              {
5746              *code = OP_END;              *code = OP_END;
5747              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);              if (recno != 0)
5748                  called = _pcre_find_bracket(cd->start_code, utf8, recno);
5749    
5750              /* Forward reference */              /* Forward reference */
5751    
5752              if (called == NULL)              if (called == NULL)
5753                {                {
5754                if (find_parens(cd, NULL, recno,                if (find_parens(cd, NULL, recno,
5755                      (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0, utf8) < 0)
5756                  {                  {
5757                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5758                  goto FAILED;                  goto FAILED;
5759                  }                  }
5760    
5761                  /* Fudge the value of "called" so that when it is inserted as an
5762                  offset below, what it actually inserted is the reference number
5763                  of the group. Then remember the forward reference. */
5764    
5765                called = cd->start_code + recno;                called = cd->start_code + recno;
5766                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
5767                }                }
5768    
5769              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
5770              this is a recursive call. We check to see if this is a left              this is a recursive call. We check to see if this is a left
5771              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. We
5772                must not, however, do this check if we are in a conditional
5773                subpattern because the condition might be testing for recursion in
5774                a pattern such as /(?(R)a+|(?R)b)/, which is perfectly valid.
5775                Forever loops are also detected at runtime, so those that occur in
5776                conditional subpatterns will be picked up then. */
5777    
5778              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 && cond_depth <= 0 &&
5779                       could_be_empty(called, code, bcptr, utf8))                       could_be_empty(called, code, bcptr, utf8, cd))
5780                {                {
5781                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
5782                goto FAILED;                goto FAILED;
5783                }                }
5784              }              }
5785    
5786            /* Insert the recursion/subroutine item, automatically wrapped inside            /* Insert the recursion/subroutine item. */
           "once" brackets. Set up a "previous group" length so that a  
           subsequent quantifier will work. */  
   
           *code = OP_ONCE;  
           PUT(code, 1, 2 + 2*LINK_SIZE);  
           code += 1 + LINK_SIZE;  
5787    
5788            *code = OP_RECURSE;            *code = OP_RECURSE;
5789            PUT(code, 1, called - cd->start_code);            PUT(code, 1, (int)(called - cd->start_code));
5790            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
   
           *code = OP_KET;  
           PUT(code, 1, 2 + 2*LINK_SIZE);  
           code += 1 + LINK_SIZE;  
   
           length_prevgroup = 3 + 3*LINK_SIZE;  
5791            }            }
5792    
5793          /* Can't determine a first byte now */          /* Can't determine a first byte now */
# Line 5130  we set the flag only if there is a liter Line 5848  we set the flag only if there is a liter
5848          is necessary to ensure we correctly detect the start of the pattern in          is necessary to ensure we correctly detect the start of the pattern in
5849          both phases.          both phases.
5850    
5851          If we are not at the pattern start, compile code to change the ims          If we are not at the pattern start, reset the greedy defaults and the
5852          options if this setting actually changes any of them, and reset the          case value for firstbyte and reqbyte. */
         greedy defaults and the case value for firstbyte and reqbyte. */  
5853    
5854          if (*ptr == CHAR_RIGHT_PARENTHESIS)          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5855            {            {
# Line 5141  we set the flag only if there is a liter Line 5858  we set the flag only if there is a liter
5858              {              {
5859              cd->external_options = newoptions;              cd->external_options = newoptions;
5860              }              }
5861           else            else
5862              {              {
             if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))  
               {  
               *code++ = OP_OPT;  
               *code++ = newoptions & PCRE_IMS;  
               }  
5863              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5864              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5865              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5866              }              }
5867    
5868            /* Change options at this level, and pass them back for use            /* Change options at this level, and pass them back for use
5869            in subsequent branches. When not at the start of the pattern, this            in subsequent branches. */
           information is also necessary so that a resetting item can be  
           compiled at the end of a group (if we are in a group). */  
5870    
5871            *optionsptr = options = newoptions;            *optionsptr = options = newoptions;
5872            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
# Line 5173  we set the flag only if there is a liter Line 5883  we set the flag only if there is a liter
5883          }     /* End of switch for character following (? */          }     /* End of switch for character following (? */
5884        }       /* End of (? handling */        }       /* End of (? handling */
5885    
5886      /* Opening parenthesis not followed by '?'. If PCRE_NO_AUTO_CAPTURE is set,      /* Opening parenthesis not followed by '*' or '?'. If PCRE_NO_AUTO_CAPTURE
5887      all unadorned brackets become non-capturing and behave like (?:...)      is set, all unadorned brackets become non-capturing and behave like (?:...)
5888      brackets. */      brackets. */
5889    
5890      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)
# Line 5192  we set the flag only if there is a liter Line 5902  we set the flag only if there is a liter
5902        skipbytes = 2;        skipbytes = 2;
5903        }        }
5904    
5905      /* Process nested bracketed regex. Assertions may not be repeated, but      /* Process nested bracketed regex. Assertions used not to be repeatable,
5906      other kinds can be. All their opcodes are >= OP_ONCE. We copy code into a      but this was changed for Perl compatibility, so all kinds can now be
5907      non-register variable in order to be able to pass its address because some      repeated. We copy code into a non-register variable (tempcode) in order to
5908      compilers complain otherwise. Pass in a new setting for the ims options if      be able to pass its address because some compilers complain otherwise. */
     they have changed. */  
5909    
5910      previous = (bravalue >= OP_ONCE)? code : NULL;      previous = code;                      /* For handling repetition */
5911      *code = bravalue;      *code = bravalue;
5912      tempcode = code;      tempcode = code;
5913      tempreqvary = cd->req_varyopt;     /* Save value before bracket */      tempreqvary = cd->req_varyopt;        /* Save value before bracket */
5914      length_prevgroup = 0;              /* Initialize for pre-compile phase */      tempbracount = cd->bracount;          /* Save value before bracket */
5915        length_prevgroup = 0;                 /* Initialize for pre-compile phase */
5916    
5917      if (!compile_regex(      if (!compile_regex(
5918           newoptions,                   /* The complete new option state */           newoptions,                      /* The complete new option state */
5919           options & PCRE_IMS,           /* The previous ims option state */           &tempcode,                       /* Where to put code (updated) */
5920           &tempcode,                    /* Where to put code (updated) */           &ptr,                            /* Input pointer (updated) */
5921           &ptr,                         /* Input pointer (updated) */           errorcodeptr,                    /* Where to put an error message */
          errorcodeptr,                 /* Where to put an error message */  
5922           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
5923            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
5924           reset_bracount,               /* True if (?| group */           reset_bracount,                  /* True if (?| group */
5925           skipbytes,                    /* Skip over bracket number */           skipbytes,                       /* Skip over bracket number */
5926           &subfirstbyte,                /* For possible first char */           cond_depth +
5927           &subreqbyte,                  /* For possible last char */             ((bravalue == OP_COND)?1:0),   /* Depth of condition subpatterns */
5928           bcptr,                        /* Current branch chain */           &subfirstbyte,                   /* For possible first char */
5929           cd,                           /* Tables block */           &subreqbyte,                     /* For possible last char */
5930           (lengthptr == NULL)? NULL :   /* Actual compile phase */           bcptr,                           /* Current branch chain */
5931             &length_prevgroup           /* Pre-compile phase */           cd,                              /* Tables block */
5932             (lengthptr == NULL)? NULL :      /* Actual compile phase */
5933               &length_prevgroup              /* Pre-compile phase */
5934           ))           ))
5935        goto FAILED;        goto FAILED;
5936    
5937        /* If this was an atomic group and there are no capturing groups within it,
5938        generate OP_ONCE_NC instead of OP_ONCE. */
5939    
5940        if (bravalue == OP_ONCE && cd->bracount <= tempbracount)
5941          *code = OP_ONCE_NC;
5942    
5943        if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
5944          cd->assert_depth -= 1;
5945    
5946      /* At the end of compiling, code is still pointing to the start of the      /* At the end of compiling, code is still pointing to the start of the
5947      group, while tempcode has been updated to point past the end of the group      group, while tempcode has been updated to point past the end of the group.
5948      and any option resetting that may follow it. The pattern pointer (ptr)      The pattern pointer (ptr) is on the bracket.
     is on the bracket. */  
5949    
5950      /* If this is a conditional bracket, check that there are no more than      If this is a conditional bracket, check that there are no more than
5951      two branches in the group, or just one if it's a DEFINE group. We do this      two branches in the group, or just one if it's a DEFINE group. We do this
5952      in the real compile phase, not in the pre-pass, where the whole group may      in the real compile phase, not in the pre-pass, where the whole group may
5953      not be available. */      not be available. */
# Line 5293  we set the flag only if there is a liter Line 6012  we set the flag only if there is a liter
6012          goto FAILED;          goto FAILED;
6013          }          }
6014        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
6015        *code++ = OP_BRA;        code++;   /* This already contains bravalue */
6016        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
6017        *code++ = OP_KET;        *code++ = OP_KET;
6018        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
# Line 5366  we set the flag only if there is a liter Line 6085  we set the flag only if there is a liter
6085    
6086      /* ===================================================================*/      /* ===================================================================*/
6087      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values
6088      are arranged to be the negation of the corresponding OP_values. For the      are arranged to be the negation of the corresponding OP_values in the
6089      back references, the values are ESC_REF plus the reference number. Only      default case when PCRE_UCP is not set. For the back references, the values
6090      back references and those types that consume a character may be repeated.      are ESC_REF plus the reference number. Only back references and those types
6091      We can test for values between ESC_b and ESC_Z for the latter; this may      that consume a character may be repeated. We can test for values between
6092      have to change if any new ones are ever created. */      ESC_b and ESC_Z for the latter; this may have to change if any new ones are
6093        ever created. */
6094    
6095      case CHAR_BACKSLASH:      case CHAR_BACKSLASH:
6096      tempptr = ptr;      tempptr = ptr;
# Line 5460  we set the flag only if there is a liter Line 6180  we set the flag only if there is a liter
6180          }          }
6181    
6182        /* \k<name> or \k'name' is a back reference by name (Perl syntax).        /* \k<name> or \k'name' is a back reference by name (Perl syntax).
6183        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax).  */
6184    
6185        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||        if (-c == ESC_k)
           ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))  
6186          {          {
6187            if ((ptr[1] != CHAR_LESS_THAN_SIGN &&
6188              ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))
6189              {
6190              *errorcodeptr = ERR69;
6191              break;
6192              }
6193          is_recurse = FALSE;          is_recurse = FALSE;
6194          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
6195            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
# Line 5478  we set the flag only if there is a liter Line 6203  we set the flag only if there is a liter
6203    
6204        if (-c >= ESC_REF)        if (-c >= ESC_REF)
6205          {          {
6206            open_capitem *oc;
6207          recno = -c - ESC_REF;          recno = -c - ESC_REF;
6208    
6209          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
6210          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
6211          previous = code;          previous = code;
6212          *code++ = OP_REF;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_REFI : OP_REF;
6213          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
6214          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
6215          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
6216    
6217            /* Check to see if this back reference is recursive, that it, it
6218            is inside the group that it references. A flag is set so that the
6219            group can be made atomic. */
6220    
6221            for (oc = cd->open_caps; oc != NULL; oc = oc->next)
6222              {
6223              if (oc->number == recno)
6224                {
6225                oc->flag = TRUE;
6226                break;
6227                }
6228              }
6229          }          }
6230    
6231        /* So are Unicode property matches, if supported. */        /* So are Unicode property matches, if supported. */
# Line 5516  we set the flag only if there is a liter Line 6255  we set the flag only if there is a liter
6255  #endif  #endif
6256    
6257        /* For the rest (including \X when Unicode properties are supported), we        /* For the rest (including \X when Unicode properties are supported), we
6258        can obtain the OP value by negating the escape value. */        can obtain the OP value by negating the escape value in the default
6259          situation when PCRE_UCP is not set. When it *is* set, we substitute
6260          Unicode property tests. */
6261    
6262        else        else
6263          {          {
6264          previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;  #ifdef SUPPORT_UCP
6265          *code++ = -c;          if (-c >= ESC_DU && -c <= ESC_wu)
6266              {
6267              nestptr = ptr + 1;                   /* Where to resume */
6268              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
6269              }
6270            else
6271    #endif
6272              {
6273              previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;
6274              *code++ = -c;
6275              }
6276          }          }
6277        continue;        continue;
6278        }        }
# Line 5566  we set the flag only if there is a liter Line 6317  we set the flag only if there is a liter
6317    
6318      ONE_CHAR:      ONE_CHAR:
6319      previous = code;      previous = code;
6320      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARI : OP_CHAR;
6321      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
6322    
6323      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
# Line 5595  we set the flag only if there is a liter Line 6346  we set the flag only if there is a liter
6346        else firstbyte = reqbyte = REQ_NONE;        else firstbyte = reqbyte = REQ_NONE;
6347        }        }
6348    
6349      /* firstbyte was previously set; we can set reqbyte only the length is      /* firstbyte was previously set; we can set reqbyte only if the length is
6350      1 or the matching is caseful. */      1 or the matching is caseful. */
6351    
6352      else      else
# Line 5630  return FALSE; Line 6381  return FALSE;
6381  /* On entry, ptr is pointing past the bracket character, but on return it  /* On entry, ptr is pointing past the bracket character, but on return it
6382  points to the closing bracket, or vertical bar, or end of string. The code  points to the closing bracket, or vertical bar, or end of string. The code
6383  variable is pointing at the byte into which the BRA operator has been stored.  variable is pointing at the byte into which the BRA operator has been stored.
 If the ims options are changed at the start (for a (?ims: group) or during any  
 branch, we need to insert an OP_OPT item at the start of every following branch  
 to ensure they get set correctly at run time, and also pass the new options  
 into every subsequent branch compile.  
   
6384  This function is used during the pre-compile phase when we are trying to find  This function is used during the pre-compile phase when we are trying to find
6385  out the amount of memory needed, as well as during the real compile phase. The  out the amount of memory needed, as well as during the real compile phase. The
6386  value of lengthptr distinguishes the two phases.  value of lengthptr distinguishes the two phases.
6387    
6388  Arguments:  Arguments:
6389    options        option bits, including any changes for this subpattern    options        option bits, including any changes for this subpattern
   oldims         previous settings of ims option bits  
6390    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
6391    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
6392    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
6393    lookbehind     TRUE if this is a lookbehind assertion    lookbehind     TRUE if this is a lookbehind assertion
6394    reset_bracount TRUE to reset the count for each branch    reset_bracount TRUE to reset the count for each branch
6395    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes      skip this many bytes at start (for brackets and OP_COND)
6396      cond_depth     depth of nesting for conditional subpatterns
6397    firstbyteptr   place to put the first required character, or a negative number    firstbyteptr   place to put the first required character, or a negative number
6398    reqbyteptr     place to put the last required character, or a negative number    reqbyteptr     place to put the last required character, or a negative number
6399    bcptr          pointer to the chain of currently open branches    bcptr          pointer to the chain of currently open branches
# Line 5659  Returns:         TRUE on success Line 6405  Returns:         TRUE on success
6405  */  */
6406    
6407  static BOOL  static BOOL
6408  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,
6409    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
6410    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,    int cond_depth, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
6411    int *lengthptr)    compile_data *cd, int *lengthptr)
6412  {  {
6413  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
6414  uschar *code = *codeptr;  uschar *code = *codeptr;
6415  uschar *last_branch = code;  uschar *last_branch = code;
6416  uschar *start_bracket = code;  uschar *start_bracket = code;
6417  uschar *reverse_count = NULL;  uschar *reverse_count = NULL;
6418    open_capitem capitem;
6419    int capnumber = 0;
6420  int firstbyte, reqbyte;  int firstbyte, reqbyte;
6421  int branchfirstbyte, branchreqbyte;  int branchfirstbyte, branchreqbyte;
6422  int length;  int length;
# Line 5677  int max_bracount; Line 6425  int max_bracount;
6425  branch_chain bc;  branch_chain bc;
6426    
6427  bc.outer = bcptr;  bc.outer = bcptr;
6428  bc.current = code;  bc.current_branch = code;
6429    
6430  firstbyte = reqbyte = REQ_UNSET;  firstbyte = reqbyte = REQ_UNSET;
6431    
# Line 5695  the code that abstracts option settings Line 6443  the code that abstracts option settings
6443  them global. It tests the value of length for (2 + 2*LINK_SIZE) in the  them global. It tests the value of length for (2 + 2*LINK_SIZE) in the
6444  pre-compile phase to find out whether anything has yet been compiled or not. */  pre-compile phase to find out whether anything has yet been compiled or not. */
6445    
6446    /* If this is a capturing subpattern, add to the chain of open capturing items
6447    so that we can detect them if (*ACCEPT) is encountered. This is also used to
6448    detect groups that contain recursive back references to themselves. Note that
6449    only OP_CBRA need be tested here; changing this opcode to one of its variants,
6450    e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
6451    
6452    if (*code == OP_CBRA)
6453      {
6454      capnumber = GET2(code, 1 + LINK_SIZE);
6455      capitem.number = capnumber;
6456      capitem.next = cd->open_caps;
6457      capitem.flag = FALSE;
6458      cd->open_caps = &capitem;
6459      }
6460    
6461  /* Offset is set zero to mark that this bracket is still open */  /* Offset is set zero to mark that this bracket is still open */
6462    
6463  PUT(code, 1, 0);  PUT(code, 1, 0);
# Line 5710  for (;;) Line 6473  for (;;)
6473    
6474    if (reset_bracount) cd->bracount = orig_bracount;    if (reset_bracount) cd->bracount = orig_bracount;
6475    
   /* Handle a change of ims options at the start of the branch */  
   
   if ((options & PCRE_IMS) != oldims)  
     {  
     *code++ = OP_OPT;  
     *code++ = options & PCRE_IMS;  
     length += 2;  
     }  
   
6476    /* Set up dummy OP_REVERSE if lookbehind assertion */    /* Set up dummy OP_REVERSE if lookbehind assertion */
6477    
6478    if (lookbehind)    if (lookbehind)
# Line 5733  for (;;) Line 6487  for (;;)
6487    into the length. */    into the length. */
6488    
6489    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,
6490          &branchreqbyte, &bc, cd, (lengthptr == NULL)? NULL : &length))          &branchreqbyte, &bc, cond_depth, cd,
6491            (lengthptr == NULL)? NULL : &length))
6492      {      {
6493      *ptrptr = ptr;      *ptrptr = ptr;
6494      return FALSE;      return FALSE;
# Line 5789  for (;;) Line 6544  for (;;)
6544    
6545      /* If lookbehind, check that this branch matches a fixed-length string, and      /* If lookbehind, check that this branch matches a fixed-length string, and
6546      put the length into the OP_REVERSE item. Temporarily mark the end of the      put the length into the OP_REVERSE item. Temporarily mark the end of the
6547      branch with OP_END. */      branch with OP_END. If the branch contains OP_RECURSE, the result is -3
6548        because there may be forward references that we can't check here. Set a
6549        flag to cause another lookbehind check at the end. Why not do it all at the
6550        end? Because common, erroneous checks are picked up here and the offset of
6551        the problem can be shown. */
6552    
6553      if (lookbehind)      if (lookbehind)
6554        {        {
6555        int fixed_length;        int fixed_length;
6556        *code = OP_END;        *code = OP_END;
6557        fixed_length = find_fixedlength(last_branch, options);        fixed_length = find_fixedlength(last_branch,  (options & PCRE_UTF8) != 0,
6558            FALSE, cd);
6559        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6560        if (fixed_length < 0)        if (fixed_length == -3)
6561            {
6562            cd->check_lookbehind = TRUE;
6563            }
6564          else if (fixed_length < 0)
6565          {          {
6566          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;
6567          *ptrptr = ptr;          *ptrptr = ptr;
6568          return FALSE;          return FALSE;
6569          }          }
6570        PUT(reverse_count, 0, fixed_length);        else { PUT(reverse_count, 0, fixed_length); }
6571        }        }
6572      }      }
6573    
# Line 5812  for (;;) Line 6576  for (;;)
6576    of offsets, with the field in the BRA item now becoming an offset to the    of offsets, with the field in the BRA item now becoming an offset to the
6577    first alternative. If there are no alternatives, it points to the end of the    first alternative. If there are no alternatives, it points to the end of the
6578    group. The length in the terminating ket is always the length of the whole    group. The length in the terminating ket is always the length of the whole
6579    bracketed item. If any of the ims options were changed inside the group,    bracketed item. Return leaving the pointer at the terminating char. */
   compile a resetting op-code following, except at the very end of the pattern.  
   Return leaving the pointer at the terminating char. */  
6580    
6581    if (*ptr != CHAR_VERTICAL_LINE)    if (*ptr != CHAR_VERTICAL_LINE)
6582      {      {
6583      if (lengthptr == NULL)      if (lengthptr == NULL)
6584        {        {
6585        int branch_length = code - last_branch;        int branch_length = (int)(code - last_branch);
6586        do        do
6587          {          {
6588          int prev_length = GET(last_branch, 1);          int prev_length = GET(last_branch, 1);
# Line 5834  for (;;) Line 6596  for (;;)
6596      /* Fill in the ket */      /* Fill in the ket */
6597    
6598      *code = OP_KET;      *code = OP_KET;
6599      PUT(code, 1, code - start_bracket);      PUT(code, 1, (int)(code - start_bracket));
6600      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6601    
6602      /* Resetting option if needed */      /* If it was a capturing subpattern, check to see if it contained any
6603        recursive back references. If so, we must wrap it in atomic brackets.
6604        In any event, remove the block from the chain. */
6605    
6606      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)      if (capnumber > 0)
6607        {        {
6608        *code++ = OP_OPT;        if (cd->open_caps->flag)
6609        *code++ = oldims;          {
6610        length += 2;          memmove(start_bracket + 1 + LINK_SIZE, start_bracket,
6611              code - start_bracket);
6612            *start_bracket = OP_ONCE;
6613            code += 1 + LINK_SIZE;
6614            PUT(start_bracket, 1, (int)(code - start_bracket));
6615            *code = OP_KET;
6616            PUT(code, 1, (int)(code - start_bracket));
6617            code += 1 + LINK_SIZE;
6618            length += 2 + 2*LINK_SIZE;
6619            }
6620          cd->open_caps = cd->open_caps->next;
6621        }        }
6622    
6623      /* Retain the highest bracket number, in case resetting was used. */      /* Retain the highest bracket number, in case resetting was used. */
# Line 5885  for (;;) Line 6659  for (;;)
6659    else    else
6660      {      {
6661      *code = OP_ALT;      *code = OP_ALT;
6662      PUT(code, 1, code - last_branch);      PUT(code, 1, (int)(code - last_branch));
6663      bc.current = last_branch = code;      bc.current_branch = last_branch = code;
6664      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6665      }      }
6666    
# Line 5905  for (;;) Line 6679  for (;;)
6679  /* Try to find out if this is an anchored regular expression. Consider each  /* Try to find out if this is an anchored regular expression. Consider each
6680  alternative branch. If they all start with OP_SOD or OP_CIRC, or with a bracket  alternative branch. If they all start with OP_SOD or OP_CIRC, or with a bracket
6681  all of whose alternatives start with OP_SOD or OP_CIRC (recurse ad lib), then  all of whose alternatives start with OP_SOD or OP_CIRC (recurse ad lib), then
6682  it's anchored. However, if this is a multiline pattern, then only OP_SOD  it's anchored. However, if this is a multiline pattern, then only OP_SOD will
6683  counts, since OP_CIRC can match in the middle.  be found, because ^ generates OP_CIRCM in that mode.
6684    
6685  We can also consider a regex to be anchored if OP_SOM starts all its branches.  We can also consider a regex to be anchored if OP_SOM starts all its branches.
6686  This is the code for \G, which means "match at start of match position, taking  This is the code for \G, which means "match at start of match position, taking
# Line 5927  of the more common cases more precisely. Line 6701  of the more common cases more precisely.
6701    
6702  Arguments:  Arguments:
6703    code           points to start of expression (the bracket)    code           points to start of expression (the bracket)
   options        points to the options setting  
6704    bracket_map    a bitmap of which brackets we are inside while testing; this    bracket_map    a bitmap of which brackets we are inside while testing; this
6705                    handles up to substring 31; after that we just have to take                    handles up to substring 31; after that we just have to take
6706                    the less precise approach                    the less precise approach
# Line 5937  Returns:     TRUE or FALSE Line 6710  Returns:     TRUE or FALSE
6710  */  */
6711    
6712  static BOOL  static BOOL
6713  is_anchored(register const uschar *code, int *options, unsigned int bracket_map,  is_anchored(register const uschar *code, unsigned int bracket_map,
6714    unsigned int backref_map)    unsigned int backref_map)
6715  {  {
6716  do {  do {
6717     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6718       options, PCRE_MULTILINE, FALSE);       FALSE);
6719     register int op = *scode;     register int op = *scode;
6720    
6721     /* Non-capturing brackets */     /* Non-capturing brackets */
6722    
6723     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6724           op == OP_SBRA || op == OP_SBRAPOS)
6725       {       {
6726       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6727       }       }
6728    
6729     /* Capturing brackets */     /* Capturing brackets */
6730    
6731     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6732                op == OP_SCBRA || op == OP_SCBRAPOS)
6733       {       {
6734       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6735       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
6736       if (!is_anchored(scode, options, new_map, backref_map)) return FALSE;       if (!is_anchored(scode, new_map, backref_map)) return FALSE;
6737       }       }
6738    
6739     /* Other brackets */     /* Other brackets */
6740    
6741     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC ||
6742                op == OP_COND)
6743       {       {
6744       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6745       }       }
6746    
6747     /* .* is not anchored unless DOTALL is set (which generates OP_ALLANY) and     /* .* is not anchored unless DOTALL is set (which generates OP_ALLANY) and
# Line 5980  do { Line 6756  do {
6756    
6757     /* Check for explicit anchoring */     /* Check for explicit anchoring */
6758    
6759     else if (op != OP_SOD && op != OP_SOM &&     else if (op != OP_SOD && op != OP_SOM && op != OP_CIRC) return FALSE;
            ((*options & PCRE_MULTILINE) != 0 || op != OP_CIRC))  
      return FALSE;  
6760     code += GET(code, 1);     code += GET(code, 1);
6761     }     }
6762  while (*code == OP_ALT);   /* Loop for each alternative */  while (*code == OP_ALT);   /* Loop for each alternative */
# Line 6018  is_startline(const uschar *code, unsigne Line 6792  is_startline(const uschar *code, unsigne
6792  {  {
6793  do {  do {
6794     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6795       NULL, 0, FALSE);       FALSE);
6796     register int op = *scode;     register int op = *scode;
6797    
6798     /* If we are at the start of a conditional assertion group, *both* the     /* If we are at the start of a conditional assertion group, *both* the
# Line 6033  do { Line 6807  do {
6807       switch (*scode)       switch (*scode)
6808         {         {
6809         case OP_CREF:         case OP_CREF:
6810           case OP_NCREF:
6811         case OP_RREF:         case OP_RREF:
6812           case OP_NRREF:
6813         case OP_DEF:         case OP_DEF:
6814         return FALSE;         return FALSE;
6815    
# Line 6043  do { Line 6819  do {
6819         scode += 1 + LINK_SIZE;         scode += 1 + LINK_SIZE;
6820         break;         break;
6821         }         }
6822       scode = first_significant_code(scode, NULL, 0, FALSE);       scode = first_significant_code(scode, FALSE);
6823       op = *scode;       op = *scode;
6824       }       }
6825    
6826     /* Non-capturing brackets */     /* Non-capturing brackets */
6827    
6828     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6829           op == OP_SBRA || op == OP_SBRAPOS)
6830       {       {
6831       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6832       }       }
6833    
6834     /* Capturing brackets */     /* Capturing brackets */
6835    
6836     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6837                op == OP_SCBRA || op == OP_SCBRAPOS)
6838       {       {
6839       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6840       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
# Line 6065  do { Line 6843  do {
6843    
6844     /* Other brackets */     /* Other brackets */
6845    
6846     else if (op == OP_ASSERT || op == OP_ONCE)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC)
6847       {       {
6848       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6849       }       }
# Line 6080  do { Line 6858  do {
6858    
6859     /* Check for explicit circumflex */     /* Check for explicit circumflex */
6860    
6861     else if (op != OP_CIRC) return FALSE;     else if (op != OP_CIRC && op != OP_CIRCM) return FALSE;
6862    
6863     /* Move on to the next alternative */     /* Move on to the next alternative */
6864    
# Line 6106  we return that char, otherwise -1. Line 6884  we return that char, otherwise -1.
6884    
6885  Arguments:  Arguments:
6886    code       points to start of expression (the bracket)    code       points to start of expression (the bracket)
   options    pointer to the options (used to check casing changes)  
6887    inassert   TRUE if in an assertion    inassert   TRUE if in an assertion
6888    
6889  Returns:     -1 or the fixed first char  Returns:     -1 or the fixed first char
6890  */  */
6891    
6892  static int  static int
6893  find_firstassertedchar(const uschar *code, int *options, BOOL inassert)  find_firstassertedchar(const uschar *code, BOOL inassert)
6894  {  {
6895  register int c = -1;  register int c = -1;
6896  do {  do {
6897     int d;     int d;
6898     const uschar *scode =     int xl = (*code == OP_CBRA || *code == OP_SCBRA ||
6899       first_significant_code(code + 1+LINK_SIZE, options, PCRE_CASELESS, TRUE);               *code == OP_CBRAPOS || *code == OP_SCBRAPOS)? 2:0;
6900       const uschar *scode = first_significant_code(code + 1+LINK_SIZE + xl, TRUE);
6901     register int op = *scode;     register int op = *scode;
6902    
6903     switch(op)     switch(op)
# Line 6128  do { Line 6906  do {
6906       return -1;       return -1;
6907    
6908       case OP_BRA:       case OP_BRA:
6909         case OP_BRAPOS:
6910       case OP_CBRA:       case OP_CBRA:
6911         case OP_SCBRA:
6912         case OP_CBRAPOS:
6913         case OP_SCBRAPOS:
6914       case OP_ASSERT:       case OP_ASSERT:
6915       case OP_ONCE:       case OP_ONCE:
6916         case OP_ONCE_NC:
6917       case OP_COND:       case OP_COND:
6918       if ((d = find_firstassertedchar(scode, options, op == OP_ASSERT)) < 0)       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)
6919         return -1;         return -1;
6920       if (c < 0) c = d; else if (c != d) return -1;       if (c < 0) c = d; else if (c != d) return -1;
6921       break;       break;
6922    
6923       case OP_EXACT:       /* Fall through */       case OP_EXACT:
6924       scode += 2;       scode += 2;
6925         /* Fall through */
6926    
6927       case OP_CHAR:       case OP_CHAR:
      case OP_CHARNC:  
6928       case OP_PLUS:       case OP_PLUS:
6929       case OP_MINPLUS:       case OP_MINPLUS:
6930       case OP_POSPLUS:       case OP_POSPLUS:
6931       if (!inassert) return -1;       if (!inassert) return -1;
6932       if (c < 0)       if (c < 0) c = scode[1];
6933         {         else if (c != scode[1]) return -1;
6934         c = scode[1];       break;
6935         if ((*options & PCRE_CASELESS) != 0) c |= REQ_CASELESS;  
6936         }       case OP_EXACTI:
6937       else if (c != scode[1]) return -1;       scode += 2;
6938         /* Fall through */
6939    
6940         case OP_CHARI:
6941         case OP_PLUSI:
6942         case OP_MINPLUSI:
6943         case OP_POSPLUSI:
6944         if (!inassert) return -1;
6945         if (c < 0) c = scode[1] | REQ_CASELESS;
6946           else if (c != scode[1]) return -1;
6947       break;       break;
6948       }       }
6949    
# Line 6202  int length = 1;  /* For final END opcode Line 6994  int length = 1;  /* For final END opcode
6994  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
6995  int errorcode = 0;  int errorcode = 0;
6996  int skipatstart = 0;  int skipatstart = 0;
 #ifdef SUPPORT_UTF8  
6997  BOOL utf8;  BOOL utf8;
 #endif  
6998  size_t size;  size_t size;
6999  uschar *code;  uschar *code;
7000  const uschar *codestart;  const uschar *codestart;
# Line 6274  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7064  while (ptr[skipatstart] == CHAR_LEFT_PAR
7064    
7065    if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)
7066      { skipatstart += 7; options |= PCRE_UTF8; continue; }      { skipatstart += 7; options |= PCRE_UTF8; continue; }
7067      else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)
7068        { skipatstart += 6; options |= PCRE_UCP; continue; }
7069      else if (strncmp((char *)(ptr+skipatstart+2), STRING_NO_START_OPT_RIGHTPAR, 13) == 0)
7070        { skipatstart += 15; options |= PCRE_NO_START_OPTIMIZE; continue; }
7071    
7072    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
7073      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
# Line 6298  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7092  while (ptr[skipatstart] == CHAR_LEFT_PAR
7092    else break;    else break;
7093    }    }
7094    
7095  /* Can't support UTF8 unless PCRE has been compiled to include the code. */  utf8 = (options & PCRE_UTF8) != 0;
7096    
7097    /* Can't support UTF8 unless PCRE has been compiled to include the code. The
7098    return of an error code from _pcre_valid_utf8() is a new feature, introduced in
7099    release 8.13. It is passed back from pcre_[dfa_]exec(), but at the moment is
7100    not used here. */
7101    
7102  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
 utf8 = (options & PCRE_UTF8) != 0;  
7103  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
7104       (*erroroffset = _pcre_valid_utf8((uschar *)pattern, -1)) >= 0)       (errorcode = _pcre_valid_utf8((USPTR)pattern, -1, erroroffset)) != 0)
7105    {    {
7106    errorcode = ERR44;    errorcode = ERR44;
7107    goto PCRE_EARLY_ERROR_RETURN2;    goto PCRE_EARLY_ERROR_RETURN2;
7108    }    }
7109  #else  #else
7110  if ((options & PCRE_UTF8) != 0)  if (utf8)
7111    {    {
7112    errorcode = ERR32;    errorcode = ERR32;
7113    goto PCRE_EARLY_ERROR_RETURN;    goto PCRE_EARLY_ERROR_RETURN;
7114    }    }
7115  #endif  #endif
7116    
7117    /* Can't support UCP unless PCRE has been compiled to include the code. */
7118    
7119    #ifndef SUPPORT_UCP
7120    if ((options & PCRE_UCP) != 0)
7121      {
7122      errorcode = ERR67;
7123      goto PCRE_EARLY_ERROR_RETURN;
7124      }
7125    #endif
7126    
7127  /* Check validity of \R options. */  /* Check validity of \R options. */
7128    
7129  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  if ((options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) ==
7130         (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
7131    {    {
7132    case 0:    errorcode = ERR56;
7133    case PCRE_BSR_ANYCRLF:    goto PCRE_EARLY_ERROR_RETURN;
   case PCRE_BSR_UNICODE:  
   break;  
   default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;  
7134    }    }
7135    
7136  /* Handle different types of newline. The three bits give seven cases. The  /* Handle different types of newline. The three bits give seven cases. The
# Line 6398  cd->end_pattern = (const uschar *)(patte Line 7204  cd->end_pattern = (const uschar *)(patte
7204  cd->req_varyopt = 0;  cd->req_varyopt = 0;
7205  cd->external_options = options;  cd->external_options = options;
7206  cd->external_flags = 0;  cd->external_flags = 0;
7207    cd->open_caps = NULL;
7208    
7209  /* Now do the pre-compile. On error, errorcode will be set non-zero, so we  /* Now do the pre-compile. On error, errorcode will be set non-zero, so we
7210  don't need to look at the result of the function here. The initial options have  don't need to look at the result of the function here. The initial options have
# Line 6408  outside can help speed up starting point Line 7215  outside can help speed up starting point
7215  ptr += skipatstart;  ptr += skipatstart;
7216  code = cworkspace;  code = cworkspace;
7217  *code = OP_BRA;  *code = OP_BRA;
7218  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
7219    &code, &ptr, &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd,    FALSE, 0, 0, &firstbyte, &reqbyte, NULL, cd, &length);
   &length);  
7220  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
7221    
7222  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,
# Line 6443  regex compiled on a system with 4-byte p Line 7249  regex compiled on a system with 4-byte p