/[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 391 by ph10, Tue Mar 17 21:16:01 2009 UTC revision 535 by ph10, Thu Jun 3 19:18:24 2010 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) 1997-2010 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 100  is invalid. */ Line 106  is invalid. */
106  #ifndef EBCDIC  #ifndef EBCDIC
107    
108  /* This is the "normal" table for ASCII systems or for EBCDIC systems running  /* This is the "normal" table for ASCII systems or for EBCDIC systems running
109  in UTF-8 mode. */  in UTF-8 mode. */
110    
111  static const short int escapes[] = {  static const short int escapes[] = {
112       0,                       0,       0,                       0,
113         0,                       0,
114         0,                       0,
115       0,                       0,       0,                       0,
      0,                       0,  
116       0,                       0,       0,                       0,
      0,                       0,  
117       CHAR_COLON,              CHAR_SEMICOLON,       CHAR_COLON,              CHAR_SEMICOLON,
118       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
119       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
120       CHAR_COMMERCIAL_AT,      -ESC_A,       CHAR_COMMERCIAL_AT,      -ESC_A,
121       -ESC_B,                  -ESC_C,       -ESC_B,                  -ESC_C,
122       -ESC_D,                  -ESC_E,       -ESC_D,                  -ESC_E,
123       0,                       -ESC_G,       0,                       -ESC_G,
124       -ESC_H,                  0,       -ESC_H,                  0,
125       0,                       -ESC_K,       0,                       -ESC_K,
      0,                       0,  
126       0,                       0,       0,                       0,
127         -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,
131       -ESC_V,                  -ESC_W,       -ESC_V,                  -ESC_W,
132       -ESC_X,                  0,       -ESC_X,                  0,
133       -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,       -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
134       CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,       CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
135       CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,       CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
136       CHAR_GRAVE_ACCENT,       7,       CHAR_GRAVE_ACCENT,       7,
137       -ESC_b,                  0,       -ESC_b,                  0,
138       -ESC_d,                  ESC_e,       -ESC_d,                  ESC_e,
139       ESC_f,                   0,       ESC_f,                   0,
140       -ESC_h,                  0,       -ESC_h,                  0,
141       0,                       -ESC_k,       0,                       -ESC_k,
142       0,                       0,       0,                       0,
143       ESC_n,                   0,       ESC_n,                   0,
144       -ESC_p,                  0,       -ESC_p,                  0,
145       ESC_r,                   -ESC_s,       ESC_r,                   -ESC_s,
146       ESC_tee,                 0,       ESC_tee,                 0,
147       -ESC_v,                  -ESC_w,       -ESC_v,                  -ESC_w,
148       0,                       0,       0,                       0,
149       -ESC_z       -ESC_z
150  };  };
151    
152  #else  #else
153    
154  /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */  /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
155    
# 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 177  static const short int escapes[] = { Line 183  static const short int escapes[] = {
183    
184  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
185  searched linearly. Put all the names into a single string, in order to reduce  searched linearly. Put all the names into a single string, in order to reduce
186  the number of relocations when a shared library is dynamically linked. The  the number of relocations when a shared library is dynamically linked. The
187  string is built from string macros so that it works in UTF-8 mode on EBCDIC  string is built from string macros so that it works in UTF-8 mode on EBCDIC
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 215  length entry. The first three must be al Line 226  length entry. The first three must be al
226  for handling case independence. */  for handling case independence. */
227    
228  static const char posix_names[] =  static const char posix_names[] =
229    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
230    STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0    STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
231    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
232    STRING_word0  STRING_xdigit;    STRING_word0  STRING_xdigit;
233    
# 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 335  static const char error_texts[] = Line 397  static const char error_texts[] =
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      ;
412    
413  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
414  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 360  For convenience, we use the same bit def Line 426  For convenience, we use the same bit def
426    
427  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
428    
429  #ifndef EBCDIC  #ifndef EBCDIC
430    
431  /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in  /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
432  UTF-8 mode. */  UTF-8 mode. */
433    
434  static const unsigned char digitab[] =  static const unsigned char digitab[] =
# Line 400  static const unsigned char digitab[] = Line 466  static const unsigned char digitab[] =
466    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
467    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
468    
469  #else  #else
470    
471  /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */  /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
472    
# Line 500  static const char * Line 566  static const char *
566  find_error_text(int n)  find_error_text(int n)
567  {  {
568  const char *s = error_texts;  const char *s = error_texts;
569  for (; n > 0; n--) while (*s++ != 0) {};  for (; n > 0; n--)
570      {
571      while (*s++ != 0) {};
572      if (*s == 0) return "Error text not found (please report)";
573      }
574  return s;  return s;
575  }  }
576    
# Line 571  else Line 641  else
641    
642      case CHAR_l:      case CHAR_l:
643      case CHAR_L:      case CHAR_L:
     case CHAR_N:  
644      case CHAR_u:      case CHAR_u:
645      case CHAR_U:      case CHAR_U:
646      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
# Line 809  else Line 878  else
878      }      }
879    }    }
880    
881    /* Perl supports \N{name} for character names, as well as plain \N for "not
882    newline". PCRE does not support \N{name}. */
883    
884    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)
885      *errorcodeptr = ERR37;
886    
887    /* If PCRE_UCP is set, we change the values for \d etc. */
888    
889    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
890      c -= (ESC_DU - ESC_D);
891    
892    /* Set the pointer to the final character before returning. */
893    
894  *ptrptr = ptr;  *ptrptr = ptr;
895  return c;  return c;
896  }  }
# Line 1009  return p; Line 1091  return p;
1091    
1092    
1093  /*************************************************  /*************************************************
1094  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1095  *************************************************/  *************************************************/
1096    
1097  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1098    top-level call starts at the beginning of the pattern. All other calls must
1099    start at a parenthesis. It scans along a pattern's text looking for capturing
1100  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
1101  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
1102  returns when it reaches a given numbered subpattern. This is used for forward  returns when it reaches a given numbered subpattern. We know that if (?P< is
1103  references to subpatterns. We know that if (?P< is encountered, the name will  encountered, the name will be terminated by '>' because that is checked in the
1104  be terminated by '>' because that is checked in the first pass.  first pass. Recursion is used to keep track of subpatterns that reset the
1105    capturing group numbers - the (?| feature.
1106    
1107  Arguments:  Arguments:
1108    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1109    cd           compile background data    cd           compile background data
1110    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1111    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1112    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1113      count        pointer to the current capturing subpattern number (updated)
1114    
1115  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1116  */  */
1117    
1118  static int  static int
1119  find_parens(const uschar *ptr, compile_data *cd, const uschar *name, int lorn,  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1120    BOOL xmode)    BOOL xmode, int *count)
1121  {  {
1122  const uschar *thisname;  uschar *ptr = *ptrptr;
1123  int count = cd->bracount;  int start_count = *count;
1124    int hwm_count = start_count;
1125    BOOL dup_parens = FALSE;
1126    
1127  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1128    dealing with. The very first call may not start with a parenthesis. */
1129    
1130    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1131    {    {
1132    int term;    if (ptr[1] == CHAR_QUESTION_MARK &&
1133          ptr[2] == CHAR_VERTICAL_LINE)
1134        {
1135        ptr += 3;
1136        dup_parens = TRUE;
1137        }
1138    
1139      /* Handle a normal, unnamed capturing parenthesis */
1140    
1141      else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1142        {
1143        *count += 1;
1144        if (name == NULL && *count == lorn) return *count;
1145        ptr++;
1146        }
1147    
1148      /* Handle a condition. If it is an assertion, just carry on so that it
1149      is processed as normal. If not, skip to the closing parenthesis of the
1150      condition (there can't be any nested parens. */
1151    
1152      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1153        {
1154        ptr += 2;
1155        if (ptr[1] != CHAR_QUESTION_MARK)
1156          {
1157          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1158          if (*ptr != 0) ptr++;
1159          }
1160        }
1161    
1162      /* We have either (? or (* and not a condition */
1163    
1164      else
1165        {
1166        ptr += 2;
1167        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1168    
1169        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1170    
1171        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1172            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1173          {
1174          int term;
1175          const uschar *thisname;
1176          *count += 1;
1177          if (name == NULL && *count == lorn) return *count;
1178          term = *ptr++;
1179          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1180          thisname = ptr;
1181          while (*ptr != term) ptr++;
1182          if (name != NULL && lorn == ptr - thisname &&
1183              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1184            return *count;
1185          term++;
1186          }
1187        }
1188      }
1189    
1190    /* Past any initial parenthesis handling, scan for parentheses or vertical
1191    bars. */
1192    
1193    for (; *ptr != 0; ptr++)
1194      {
1195    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1196    
1197    if (*ptr == CHAR_BACKSLASH)    if (*ptr == CHAR_BACKSLASH)
1198      {      {
1199      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1200      if (*ptr == CHAR_Q) for (;;)      if (*ptr == CHAR_Q) for (;;)
1201        {        {
1202        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1203        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1204        if (*(++ptr) == CHAR_E) break;        if (*(++ptr) == CHAR_E) break;
1205        }        }
1206      continue;      continue;
# Line 1057  for (; *ptr != 0; ptr++) Line 1209  for (; *ptr != 0; ptr++)
1209    /* Skip over character classes; this logic must be similar to the way they    /* Skip over character classes; this logic must be similar to the way they
1210    are handled for real. If the first character is '^', skip it. Also, if the    are handled for real. If the first character is '^', skip it. Also, if the
1211    first few characters (either before or after ^) are \Q\E or \E we skip them    first few characters (either before or after ^) are \Q\E or \E we skip them
1212    too. This makes for compatibility with Perl. Note the use of STR macros to    too. This makes for compatibility with Perl. Note the use of STR macros to
1213    encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */    encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1214    
1215    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
# Line 1065  for (; *ptr != 0; ptr++) Line 1217  for (; *ptr != 0; ptr++)
1217      BOOL negate_class = FALSE;      BOOL negate_class = FALSE;
1218      for (;;)      for (;;)
1219        {        {
1220        int c = *(++ptr);        if (ptr[1] == CHAR_BACKSLASH)
       if (c == CHAR_BACKSLASH)  
1221          {          {
1222          if (ptr[1] == CHAR_E)          if (ptr[2] == CHAR_E)
1223            ptr++;            ptr+= 2;
1224          else if (strncmp((const char *)ptr+1,          else if (strncmp((const char *)ptr+2,
1225                   STR_Q STR_BACKSLASH STR_E, 3) == 0)                   STR_Q STR_BACKSLASH STR_E, 3) == 0)
1226            ptr += 3;            ptr += 4;
1227          else          else
1228            break;            break;
1229          }          }
1230        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)        else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1231            {
1232          negate_class = TRUE;          negate_class = TRUE;
1233            ptr++;
1234            }
1235        else break;        else break;
1236        }        }
1237    
1238      /* If the next character is ']', it is a data character that must be      /* If the next character is ']', it is a data character that must be
1239      skipped, except in JavaScript compatibility mode. */      skipped, except in JavaScript compatibility mode. */
1240    
1241      if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&      if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1242          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1243        ptr++;        ptr++;
1244    
# Line 1093  for (; *ptr != 0; ptr++) Line 1247  for (; *ptr != 0; ptr++)
1247        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1248        if (*ptr == CHAR_BACKSLASH)        if (*ptr == CHAR_BACKSLASH)
1249          {          {
1250          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1251          if (*ptr == CHAR_Q) for (;;)          if (*ptr == CHAR_Q) for (;;)
1252            {            {
1253            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1254            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1255            if (*(++ptr) == CHAR_E) break;            if (*(++ptr) == CHAR_E) break;
1256            }            }
1257          continue;          continue;
# Line 1111  for (; *ptr != 0; ptr++) Line 1265  for (; *ptr != 0; ptr++)
1265    if (xmode && *ptr == CHAR_NUMBER_SIGN)    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1266      {      {
1267      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1268      if (*ptr == 0) return -1;      if (*ptr == 0) goto FAIL_EXIT;
1269      continue;      continue;
1270      }      }
1271    
1272    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1273    
1274    if (*ptr != CHAR_LEFT_PARENTHESIS) continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)  
1275      {      {
1276      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1277      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1278      continue;      if (*ptr == 0) goto FAIL_EXIT;
1279      }      }
1280    
1281    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1282    if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */      {
1283        if (dup_parens && *count < hwm_count) *count = hwm_count;
1284        *ptrptr = ptr;
1285        return -1;
1286        }
1287    
1288    /* We have to disambiguate (?<! and (?<= from (?<name> */    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1289        {
1290        if (*count > hwm_count) hwm_count = *count;
1291        *count = start_count;
1292        }
1293      }
1294    
1295    FAIL_EXIT:
1296    *ptrptr = ptr;
1297    return -1;
1298    }
1299    
   if ((*ptr != CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_EXCLAMATION_MARK ||  
       ptr[1] == CHAR_EQUALS_SIGN) && *ptr != CHAR_APOSTROPHE)  
     continue;  
1300    
   count++;  
1301    
1302    if (name == NULL && count == lorn) return count;  
1303    term = *ptr++;  /*************************************************
1304    if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;  *       Find forward referenced subpattern       *
1305    thisname = ptr;  *************************************************/
1306    while (*ptr != term) ptr++;  
1307    if (name != NULL && lorn == ptr - thisname &&  /* This function scans along a pattern's text looking for capturing
1308        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  subpatterns, and counting them. If it finds a named pattern that matches the
1309      return count;  name it is given, it returns its number. Alternatively, if the name is NULL, it
1310    returns when it reaches a given numbered subpattern. This is used for forward
1311    references to subpatterns. We used to be able to start this scan from the
1312    current compiling point, using the current count value from cd->bracount, and
1313    do it all in a single loop, but the addition of the possibility of duplicate
1314    subpattern numbers means that we have to scan from the very start, in order to
1315    take account of such duplicates, and to use a recursive function to keep track
1316    of the different types of group.
1317    
1318    Arguments:
1319      cd           compile background data
1320      name         name to seek, or NULL if seeking a numbered subpattern
1321      lorn         name length, or subpattern number if name is NULL
1322      xmode        TRUE if we are in /x mode
1323    
1324    Returns:       the number of the found subpattern, or -1 if not found
1325    */
1326    
1327    static int
1328    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1329    {
1330    uschar *ptr = (uschar *)cd->start_pattern;
1331    int count = 0;
1332    int rc;
1333    
1334    /* If the pattern does not start with an opening parenthesis, the first call
1335    to find_parens_sub() will scan right to the end (if necessary). However, if it
1336    does start with a parenthesis, find_parens_sub() will return when it hits the
1337    matching closing parens. That is why we have to have a loop. */
1338    
1339    for (;;)
1340      {
1341      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1342      if (rc > 0 || *ptr++ == 0) break;
1343    }    }
1344    
1345  return -1;  return rc;
1346  }  }
1347    
1348    
1349    
1350    
1351  /*************************************************  /*************************************************
1352  *      Find first significant op code            *  *      Find first significant op code            *
1353  *************************************************/  *************************************************/
# Line 1200  for (;;) Line 1397  for (;;)
1397    
1398      case OP_CALLOUT:      case OP_CALLOUT:
1399      case OP_CREF:      case OP_CREF:
1400        case OP_NCREF:
1401      case OP_RREF:      case OP_RREF:
1402        case OP_NRREF:
1403      case OP_DEF:      case OP_DEF:
1404      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1405      break;      break;
# Line 1216  for (;;) Line 1415  for (;;)
1415    
1416    
1417  /*************************************************  /*************************************************
1418  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1419  *************************************************/  *************************************************/
1420    
1421  /* 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,
1422  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.
1423  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
1424    temporarily terminated with OP_END when this function is called.
1425    
1426    This function is called when a backward assertion is encountered, so that if it
1427    fails, the error message can point to the correct place in the pattern.
1428    However, we cannot do this when the assertion contains subroutine calls,
1429    because they can be forward references. We solve this by remembering this case
1430    and doing the check at the end; a flag specifies which mode we are running in.
1431    
1432  Arguments:  Arguments:
1433    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1434    options  the compiling options    options  the compiling options
1435      atend    TRUE if called when the pattern is complete
1436      cd       the "compile data" structure
1437    
1438  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1439                 or -1 if there is no fixed length,
1440               or -2 if \C was encountered               or -2 if \C was encountered
1441                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1442  */  */
1443    
1444  static int  static int
1445  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1446  {  {
1447  int length = -1;  int length = -1;
1448    
# Line 1245  branch, check the length against that of Line 1455  branch, check the length against that of
1455  for (;;)  for (;;)
1456    {    {
1457    int d;    int d;
1458      uschar *ce, *cs;
1459    register int op = *cc;    register int op = *cc;
1460    switch (op)    switch (op)
1461      {      {
# Line 1252  for (;;) Line 1463  for (;;)
1463      case OP_BRA:      case OP_BRA:
1464      case OP_ONCE:      case OP_ONCE:
1465      case OP_COND:      case OP_COND:
1466      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1467      if (d < 0) return d;      if (d < 0) return d;
1468      branchlength += d;      branchlength += d;
1469      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1275  for (;;) Line 1486  for (;;)
1486      branchlength = 0;      branchlength = 0;
1487      break;      break;
1488    
1489        /* A true recursion implies not fixed length, but a subroutine call may
1490        be OK. If the subroutine is a forward reference, we can't deal with
1491        it until the end of the pattern, so return -3. */
1492    
1493        case OP_RECURSE:
1494        if (!atend) return -3;
1495        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1496        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1497        if (cc > cs && cc < ce) return -1;                /* Recursion */
1498        d = find_fixedlength(cs + 2, options, atend, cd);
1499        if (d < 0) return d;
1500        branchlength += d;
1501        cc += 1 + LINK_SIZE;
1502        break;
1503    
1504      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1505    
1506      case OP_ASSERT:      case OP_ASSERT:
# Line 1288  for (;;) Line 1514  for (;;)
1514    
1515      case OP_REVERSE:      case OP_REVERSE:
1516      case OP_CREF:      case OP_CREF:
1517        case OP_NCREF:
1518      case OP_RREF:      case OP_RREF:
1519        case OP_NRREF:
1520      case OP_DEF:      case OP_DEF:
1521      case OP_OPT:      case OP_OPT:
1522      case OP_CALLOUT:      case OP_CALLOUT:
1523      case OP_SOD:      case OP_SOD:
1524      case OP_SOM:      case OP_SOM:
1525        case OP_SET_SOM:
1526      case OP_EOD:      case OP_EOD:
1527      case OP_EODN:      case OP_EODN:
1528      case OP_CIRC:      case OP_CIRC:
# Line 1311  for (;;) Line 1540  for (;;)
1540      branchlength++;      branchlength++;
1541      cc += 2;      cc += 2;
1542  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1543      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1544        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1545  #endif  #endif
1546      break;      break;
1547    
# Line 1325  for (;;) Line 1552  for (;;)
1552      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1553      cc += 4;      cc += 4;
1554  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1555      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1556        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1557  #endif  #endif
1558      break;      break;
1559    
# Line 1407  for (;;) Line 1632  for (;;)
1632    
1633    
1634  /*************************************************  /*************************************************
1635  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1636  *************************************************/  *************************************************/
1637    
1638  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1639  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1640    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1641    so that it can be called from pcre_study() when finding the minimum matching
1642    length.
1643    
1644  Arguments:  Arguments:
1645    code        points to start of expression    code        points to start of expression
1646    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1647    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1648    
1649  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
1650  */  */
1651    
1652  static const uschar *  const uschar *
1653  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1654  {  {
1655  for (;;)  for (;;)
1656    {    {
# Line 1435  for (;;) Line 1663  for (;;)
1663    
1664    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1665    
1666      /* Handle recursion */
1667    
1668      else if (c == OP_REVERSE)
1669        {
1670        if (number < 0) return (uschar *)code;
1671        code += _pcre_OP_lengths[c];
1672        }
1673    
1674    /* Handle capturing bracket */    /* Handle capturing bracket */
1675    
1676    else if (c == OP_CBRA)    else if (c == OP_CBRA)
# Line 1446  for (;;) Line 1682  for (;;)
1682    
1683    /* 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
1684    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
1685    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1686      must add in its length. */
1687    
1688    else    else
1689      {      {
# Line 1470  for (;;) Line 1707  for (;;)
1707        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1708        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1709        break;        break;
1710    
1711          case OP_MARK:
1712          case OP_PRUNE_ARG:
1713          case OP_SKIP_ARG:
1714          case OP_THEN_ARG:
1715          code += code[1];
1716          break;
1717        }        }
1718    
1719      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1541  for (;;) Line 1785  for (;;)
1785    
1786    /* 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
1787    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
1788    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1789      must add in its length. */
1790    
1791    else    else
1792      {      {
# Line 1565  for (;;) Line 1810  for (;;)
1810        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1811        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1812        break;        break;
1813    
1814          case OP_MARK:
1815          case OP_PRUNE_ARG:
1816          case OP_SKIP_ARG:
1817          case OP_THEN_ARG:
1818          code += code[1];
1819          break;
1820        }        }
1821    
1822      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1621  Arguments: Line 1873  Arguments:
1873    code        points to start of search    code        points to start of search
1874    endcode     points to where to stop    endcode     points to where to stop
1875    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1876      cd          contains pointers to tables etc.
1877    
1878  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1879  */  */
1880    
1881  static BOOL  static BOOL
1882  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1883      compile_data *cd)
1884  {  {
1885  register int c;  register int c;
1886  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);
# Line 1657  for (code = first_significant_code(code Line 1911  for (code = first_significant_code(code
1911      continue;      continue;
1912      }      }
1913    
1914      /* For a recursion/subroutine call, if its end has been reached, which
1915      implies a subroutine call, we can scan it. */
1916    
1917      if (c == OP_RECURSE)
1918        {
1919        BOOL empty_branch = FALSE;
1920        const uschar *scode = cd->start_code + GET(code, 1);
1921        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1922        do
1923          {
1924          if (could_be_empty_branch(scode, endcode, utf8, cd))
1925            {
1926            empty_branch = TRUE;
1927            break;
1928            }
1929          scode += GET(scode, 1);
1930          }
1931        while (*scode == OP_ALT);
1932        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1933        continue;
1934        }
1935    
1936    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
1937    
1938    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
# Line 1664  for (code = first_significant_code(code Line 1940  for (code = first_significant_code(code
1940      BOOL empty_branch;      BOOL empty_branch;
1941      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1942    
1943      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1944        empty branch, so just skip over the conditional, because it could be empty.
1945        Otherwise, scan the individual branches of the group. */
1946    
1947      empty_branch = FALSE;      if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
     do  
       {  
       if (!empty_branch && could_be_empty_branch(code, endcode, utf8))  
         empty_branch = TRUE;  
1948        code += GET(code, 1);        code += GET(code, 1);
1949        else
1950          {
1951          empty_branch = FALSE;
1952          do
1953            {
1954            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1955              empty_branch = TRUE;
1956            code += GET(code, 1);
1957            }
1958          while (*code == OP_ALT);
1959          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1960        }        }
1961      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
1962      c = *code;      c = *code;
1963      continue;      continue;
1964      }      }
# Line 1792  for (code = first_significant_code(code Line 2076  for (code = first_significant_code(code
2076      case OP_QUERY:      case OP_QUERY:
2077      case OP_MINQUERY:      case OP_MINQUERY:
2078      case OP_POSQUERY:      case OP_POSQUERY:
2079        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2080        break;
2081    
2082      case OP_UPTO:      case OP_UPTO:
2083      case OP_MINUPTO:      case OP_MINUPTO:
2084      case OP_POSUPTO:      case OP_POSUPTO:
2085      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2086      break;      break;
2087  #endif  #endif
2088    
2089        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2090        string. */
2091    
2092        case OP_MARK:
2093        case OP_PRUNE_ARG:
2094        case OP_SKIP_ARG:
2095        case OP_THEN_ARG:
2096        code += code[1];
2097        break;
2098    
2099        /* None of the remaining opcodes are required to match a character. */
2100    
2101        default:
2102        break;
2103      }      }
2104    }    }
2105    
# Line 1820  Arguments: Line 2122  Arguments:
2122    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2123    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2124    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2125      cd          pointers to tables etc
2126    
2127  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2128  */  */
2129    
2130  static BOOL  static BOOL
2131  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2132    BOOL utf8)    BOOL utf8, compile_data *cd)
2133  {  {
2134  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2135    {    {
2136    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2137        return FALSE;
2138    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2139    }    }
2140  return TRUE;  return TRUE;
# Line 2012  auto_callout(uschar *code, const uschar Line 2316  auto_callout(uschar *code, const uschar
2316  {  {
2317  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2318  *code++ = 255;  *code++ = 255;
2319  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2320  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2321  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2322  }  }
2323    
# Line 2038  Returns:             nothing Line 2342  Returns:             nothing
2342  static void  static void
2343  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2344  {  {
2345  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2346  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2347  }  }
2348    
# Line 2088  for (++c; c <= d; c++) Line 2392  for (++c; c <= d; c++)
2392    
2393  return TRUE;  return TRUE;
2394  }  }
2395    
2396    
2397    
2398    /*************************************************
2399    *        Check a character and a property        *
2400    *************************************************/
2401    
2402    /* This function is called by check_auto_possessive() when a property item
2403    is adjacent to a fixed character.
2404    
2405    Arguments:
2406      c            the character
2407      ptype        the property type
2408      pdata        the data for the type
2409      negated      TRUE if it's a negated property (\P or \p{^)
2410    
2411    Returns:       TRUE if auto-possessifying is OK
2412    */
2413    
2414    static BOOL
2415    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2416    {
2417    const ucd_record *prop = GET_UCD(c);
2418    switch(ptype)
2419      {
2420      case PT_LAMP:
2421      return (prop->chartype == ucp_Lu ||
2422              prop->chartype == ucp_Ll ||
2423              prop->chartype == ucp_Lt) == negated;
2424    
2425      case PT_GC:
2426      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2427    
2428      case PT_PC:
2429      return (pdata == prop->chartype) == negated;
2430    
2431      case PT_SC:
2432      return (pdata == prop->script) == negated;
2433    
2434      /* These are specials */
2435    
2436      case PT_ALNUM:
2437      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2438              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2439    
2440      case PT_SPACE:    /* Perl space */
2441      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2442              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2443              == negated;
2444    
2445      case PT_PXSPACE:  /* POSIX space */
2446      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2447              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2448              c == CHAR_FF || c == CHAR_CR)
2449              == negated;
2450    
2451      case PT_WORD:
2452      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2453              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2454              c == CHAR_UNDERSCORE) == negated;
2455      }
2456    return FALSE;
2457    }
2458  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2459    
2460    
# Line 2101  whether the next thing could possibly ma Line 2468  whether the next thing could possibly ma
2468  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2469    
2470  Arguments:  Arguments:
2471    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2472    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2473    ptr           next character in pattern    ptr           next character in pattern
2474    options       options bits    options       options bits
2475    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2113  Returns:        TRUE if possessifying is Line 2478  Returns:        TRUE if possessifying is
2478  */  */
2479    
2480  static BOOL  static BOOL
2481  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2482    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2483  {  {
2484  int next;  int c, next;
2485    int op_code = *previous++;
2486    
2487  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2488    
# Line 2173  if ((options & PCRE_EXTENDED) != 0) Line 2539  if ((options & PCRE_EXTENDED) != 0)
2539    
2540  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2541    
2542  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2543    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)
2544      return FALSE;      return FALSE;
2545    
2546  /* 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
2547  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. */  
2548    
2549  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2550    {    {
2551    case OP_CHAR:    case OP_CHAR:
2552  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2553    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2554  #else  #else
2555    (void)(utf8_char);  /* Keep compiler happy by referencing function argument */    c = *previous;
2556  #endif  #endif
2557    return item != next;    return c != next;
2558    
2559    /* For CHARNC (caseless character) we must check the other case. If we have    /* For CHARNC (caseless character) we must check the other case. If we have
2560    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
# Line 2201  if (next >= 0) switch(op_code) Line 2562  if (next >= 0) switch(op_code)
2562    
2563    case OP_CHARNC:    case OP_CHARNC:
2564  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2565    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2566    #else
2567      c = *previous;
2568  #endif  #endif
2569    if (item == next) return FALSE;    if (c == next) return FALSE;
2570  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2571    if (utf8)    if (utf8)
2572      {      {
# Line 2214  if (next >= 0) switch(op_code) Line 2577  if (next >= 0) switch(op_code)
2577  #else  #else
2578      othercase = NOTACHAR;      othercase = NOTACHAR;
2579  #endif  #endif
2580      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2581      }      }
2582    else    else
2583  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2584    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2585    
2586    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, its data is always a single-byte character. */
2587    
2588    case OP_NOT:    case OP_NOT:
2589    if (item == next) return TRUE;    if ((c = *previous) == next) return TRUE;
2590    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2591  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2592    if (utf8)    if (utf8)
# Line 2235  if (next >= 0) switch(op_code) Line 2598  if (next >= 0) switch(op_code)
2598  #else  #else
2599      othercase = NOTACHAR;      othercase = NOTACHAR;
2600  #endif  #endif
2601      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2602      }      }
2603    else    else
2604  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2605    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2606    
2607      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2608      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2609    
2610    case OP_DIGIT:    case OP_DIGIT:
2611    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2282  if (next >= 0) switch(op_code) Line 2648  if (next >= 0) switch(op_code)
2648      case 0x202f:      case 0x202f:
2649      case 0x205f:      case 0x205f:
2650      case 0x3000:      case 0x3000:
2651      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2652      default:      default:
2653      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2654      }      }
2655    
2656      case OP_ANYNL:
2657    case OP_VSPACE:    case OP_VSPACE:
2658    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2659    switch(next)    switch(next)
# Line 2298  if (next >= 0) switch(op_code) Line 2665  if (next >= 0) switch(op_code)
2665      case 0x85:      case 0x85:
2666      case 0x2028:      case 0x2028:
2667      case 0x2029:      case 0x2029:
2668      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2669      default:      default:
2670      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2671      }      }
2672    
2673    #ifdef SUPPORT_UCP
2674      case OP_PROP:
2675      return check_char_prop(next, previous[0], previous[1], FALSE);
2676    
2677      case OP_NOTPROP:
2678      return check_char_prop(next, previous[0], previous[1], TRUE);
2679    #endif
2680    
2681    default:    default:
2682    return FALSE;    return FALSE;
2683    }    }
2684    
2685    
2686  /* 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
2687    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2688    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2689    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2690    replaced by OP_PROP codes when PCRE_UCP is set. */
2691    
2692  switch(op_code)  switch(op_code)
2693    {    {
2694    case OP_CHAR:    case OP_CHAR:
2695    case OP_CHARNC:    case OP_CHARNC:
2696  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2697    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2698    #else
2699      c = *previous;
2700  #endif  #endif
2701    switch(-next)    switch(-next)
2702      {      {
2703      case ESC_d:      case ESC_d:
2704      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2705    
2706      case ESC_D:      case ESC_D:
2707      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2708    
2709      case ESC_s:      case ESC_s:
2710      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2711    
2712      case ESC_S:      case ESC_S:
2713      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2714    
2715      case ESC_w:      case ESC_w:
2716      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2717    
2718      case ESC_W:      case ESC_W:
2719      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2720    
2721      case ESC_h:      case ESC_h:
2722      case ESC_H:      case ESC_H:
2723      switch(item)      switch(c)
2724        {        {
2725        case 0x09:        case 0x09:
2726        case 0x20:        case 0x20:
# Line 2367  switch(op_code) Line 2748  switch(op_code)
2748    
2749      case ESC_v:      case ESC_v:
2750      case ESC_V:      case ESC_V:
2751      switch(item)      switch(c)
2752        {        {
2753        case 0x0a:        case 0x0a:
2754        case 0x0b:        case 0x0b:
# Line 2381  switch(op_code) Line 2762  switch(op_code)
2762        return -next == ESC_v;        return -next == ESC_v;
2763        }        }
2764    
2765        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2766        their substitutions and process them. The result will always be either
2767        -ESC_p or -ESC_P. Then fall through to process those values. */
2768    
2769    #ifdef SUPPORT_UCP
2770        case ESC_du:
2771        case ESC_DU:
2772        case ESC_wu:
2773        case ESC_WU:
2774        case ESC_su:
2775        case ESC_SU:
2776          {
2777          int temperrorcode = 0;
2778          ptr = substitutes[-next - ESC_DU];
2779          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2780          if (temperrorcode != 0) return FALSE;
2781          ptr++;    /* For compatibility */
2782          }
2783        /* Fall through */
2784    
2785        case ESC_p:
2786        case ESC_P:
2787          {
2788          int ptype, pdata, errorcodeptr;
2789          BOOL negated;
2790    
2791          ptr--;      /* Make ptr point at the p or P */
2792          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2793          if (ptype < 0) return FALSE;
2794          ptr++;      /* Point past the final curly ket */
2795    
2796          /* If the property item is optional, we have to give up. (When generated
2797          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2798          to the original \d etc. At this point, ptr will point to a zero byte. */
2799    
2800          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2801            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2802              return FALSE;
2803    
2804          /* Do the property check. */
2805    
2806          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2807          }
2808    #endif
2809    
2810      default:      default:
2811      return FALSE;      return FALSE;
2812      }      }
2813    
2814      /* In principle, support for Unicode properties should be integrated here as
2815      well. It means re-organizing the above code so as to get hold of the property
2816      values before switching on the op-code. However, I wonder how many patterns
2817      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2818      these op-codes are never generated.) */
2819    
2820    case OP_DIGIT:    case OP_DIGIT:
2821    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2822           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
2823    
2824    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2825    return next == -ESC_d;    return next == -ESC_d;
2826    
2827    case OP_WHITESPACE:    case OP_WHITESPACE:
2828    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2829    
2830    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2831    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2832    
2833    case OP_HSPACE:    case OP_HSPACE:
2834    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2835             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2836    
2837    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
2838    return next == -ESC_h;    return next == -ESC_h;
2839    
2840    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
2841      case OP_ANYNL:
2842    case OP_VSPACE:    case OP_VSPACE:
2843    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2844    
2845    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2846    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
2847    
2848    case OP_WORDCHAR:    case OP_WORDCHAR:
2849    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2850             next == -ESC_v || next == -ESC_R;
2851    
2852    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2853    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2476  BOOL inescq = FALSE; Line 2911  BOOL inescq = FALSE;
2911  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
2912  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
2913  const uschar *tempptr;  const uschar *tempptr;
2914    const uschar *nestptr = NULL;
2915  uschar *previous = NULL;  uschar *previous = NULL;
2916  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2917  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
# Line 2492  BOOL utf8 = FALSE; Line 2928  BOOL utf8 = FALSE;
2928  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2929  #endif  #endif
2930    
2931  #ifdef DEBUG  #ifdef PCRE_DEBUG
2932  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2933  #endif  #endif
2934    
# Line 2546  for (;; ptr++) Line 2982  for (;; ptr++)
2982    
2983    c = *ptr;    c = *ptr;
2984    
2985      /* If we are at the end of a nested substitution, revert to the outer level
2986      string. Nesting only happens one level deep. */
2987    
2988      if (c == 0 && nestptr != NULL)
2989        {
2990        ptr = nestptr;
2991        nestptr = NULL;
2992        c = *ptr;
2993        }
2994    
2995    /* 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
2996    previous cycle of this loop. */    previous cycle of this loop. */
2997    
2998    if (lengthptr != NULL)    if (lengthptr != NULL)
2999      {      {
3000  #ifdef DEBUG  #ifdef PCRE_DEBUG
3001      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3002  #endif  #endif
3003      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3004        {        {
3005        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3006        goto FAILED;        goto FAILED;
# Line 2576  for (;; ptr++) Line 3022  for (;; ptr++)
3022        goto FAILED;        goto FAILED;
3023        }        }
3024    
3025      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3026      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));
3027    
3028      /* 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 2603  for (;; ptr++) Line 3049  for (;; ptr++)
3049    /* 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
3050    reference list. */    reference list. */
3051    
3052    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3053      {      {
3054      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3055      goto FAILED;      goto FAILED;
# Line 2639  for (;; ptr++) Line 3085  for (;; ptr++)
3085    /* Fill in length of a previous callout, except when the next thing is    /* Fill in length of a previous callout, except when the next thing is
3086    a quantifier. */    a quantifier. */
3087    
3088    is_quantifier =    is_quantifier =
3089      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3090      (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));      (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3091    
# Line 2694  for (;; ptr++) Line 3140  for (;; ptr++)
3140          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3141          goto FAILED;          goto FAILED;
3142          }          }
3143        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3144        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3145        }        }
3146      return TRUE;      return TRUE;
# Line 2759  for (;; ptr++) Line 3205  for (;; ptr++)
3205      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
3206      they are encountered at the top level, so we'll do that too. */      they are encountered at the top level, so we'll do that too. */
3207    
3208      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3209           ptr[1] == CHAR_EQUALS_SIGN) &&           ptr[1] == CHAR_EQUALS_SIGN) &&
3210          check_posix_syntax(ptr, &tempptr))          check_posix_syntax(ptr, &tempptr))
3211        {        {
# Line 2777  for (;; ptr++) Line 3223  for (;; ptr++)
3223        c = *(++ptr);        c = *(++ptr);
3224        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
3225          {          {
3226          if (ptr[1] == CHAR_E)          if (ptr[1] == CHAR_E)
3227            ptr++;            ptr++;
3228          else if (strncmp((const char *)ptr+1,          else if (strncmp((const char *)ptr+1,
3229                            STR_Q STR_BACKSLASH STR_E, 3) == 0)                            STR_Q STR_BACKSLASH STR_E, 3) == 0)
3230            ptr += 3;            ptr += 3;
3231          else          else
3232            break;            break;
3233          }          }
3234        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
# Line 2795  for (;; ptr++) Line 3241  for (;; ptr++)
3241      that. In JS mode, [] must always fail, so generate OP_FAIL, whereas      that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3242      [^] must match any character, so generate OP_ALLANY. */      [^] must match any character, so generate OP_ALLANY. */
3243    
3244      if (c == CHAR_RIGHT_SQUARE_BRACKET &&      if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3245          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3246        {        {
3247        *code++ = negate_class? OP_ALLANY : OP_FAIL;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
# Line 2877  for (;; ptr++) Line 3323  for (;; ptr++)
3323        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3324    
3325        if (c == CHAR_LEFT_SQUARE_BRACKET &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3326            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3327             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3328          {          {
3329          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
# Line 2899  for (;; ptr++) Line 3345  for (;; ptr++)
3345            ptr++;            ptr++;
3346            }            }
3347    
3348          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3349          if (posix_class < 0)          if (posix_class < 0)
3350            {            {
3351            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2913  for (;; ptr++) Line 3359  for (;; ptr++)
3359          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3360            posix_class = 0;            posix_class = 0;
3361    
3362          /* 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
3363          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3364          subtract bits that may be in the main map already. At the end we or the  
3365          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3366            if ((options & PCRE_UCP) != 0)
3367              {
3368              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3369              if (posix_substitutes[pc] != NULL)
3370                {
3371                nestptr = tempptr + 1;
3372                ptr = posix_substitutes[pc] - 1;
3373                continue;
3374                }
3375              }
3376    #endif
3377            /* In the non-UCP case, we build the bit map for the POSIX class in a
3378            chunk of local store because we may be adding and subtracting from it,
3379            and we don't want to subtract bits that may be in the main map already.
3380            At the end we or the result into the bit map that is being built. */
3381    
3382          posix_class *= 3;          posix_class *= 3;
3383    
# Line 2960  for (;; ptr++) Line 3421  for (;; ptr++)
3421    
3422        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3423        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
3424        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
3425        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
3426        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
3427        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3428          PCRE_EXTRA is set. */
3429    
3430        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
3431          {          {
3432          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3433          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3434    
3435          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 */  
3436          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3437            {            {
3438            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
# Line 2989  for (;; ptr++) Line 3449  for (;; ptr++)
3449            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3450            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3451    
3452            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3453              {              {
3454    #ifdef SUPPORT_UCP
3455                case ESC_du:     /* These are the values given for \d etc */
3456                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3457                case ESC_wu:     /* escape sequence with an appropriate \p */
3458                case ESC_WU:     /* or \P to test Unicode properties instead */
3459                case ESC_su:     /* of the default ASCII testing. */
3460                case ESC_SU:
3461                nestptr = ptr;
3462                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3463                class_charcount -= 2;                /* Undo! */
3464                continue;
3465    #endif
3466              case ESC_d:              case ESC_d:
3467              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3468              continue;              continue;
# Line 3022  for (;; ptr++) Line 3492  for (;; ptr++)
3492              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3493              continue;              continue;
3494    
3495              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)  
             {  
3496              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3497              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3498              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 3059  for (;; ptr++) Line 3516  for (;; ptr++)
3516                }                }
3517  #endif  #endif
3518              continue;              continue;
             }  
3519    
3520            if (-c == ESC_H)              case ESC_H:
             {  
3521              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3522                {                {
3523                int x = 0xff;                int x = 0xff;
# Line 3104  for (;; ptr++) Line 3559  for (;; ptr++)
3559                }                }
3560  #endif  #endif
3561              continue;              continue;
             }  
3562    
3563            if (-c == ESC_v)              case ESC_v:
             {  
3564              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3565              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3566              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 3123  for (;; ptr++) Line 3576  for (;; ptr++)
3576                }                }
3577  #endif  #endif
3578              continue;              continue;
             }  
3579    
3580            if (-c == ESC_V)              case ESC_V:
             {  
3581              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3582                {                {
3583                int x = 0xff;                int x = 0xff;
# Line 3156  for (;; ptr++) Line 3607  for (;; ptr++)
3607                }                }
3608  #endif  #endif
3609              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3610    
3611  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3612            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3613              {              case ESC_P:
3614              BOOL negated;                {
3615              int pdata;                BOOL negated;
3616              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3617              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3618              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3619              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3620                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3621              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3622              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3623              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3624              continue;                class_charcount -= 2;   /* Not a < 256 character */
3625              }                continue;
3626                  }
3627  #endif  #endif
3628            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3629            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3630            treated as literals. */              treated as literals. */
3631    
3632            if ((options & PCRE_EXTRA) != 0)              default:
3633              {              if ((options & PCRE_EXTRA) != 0)
3634              *errorcodeptr = ERR7;                {
3635              goto FAILED;                *errorcodeptr = ERR7;
3636                  goto FAILED;
3637                  }
3638                class_charcount -= 2;  /* Undo the default count from above */
3639                c = *ptr;              /* Get the final character and fall through */
3640                break;
3641              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3642            }            }
3643    
3644          /* 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 3227  for (;; ptr++) Line 3678  for (;; ptr++)
3678          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3679            {            {
3680            ptr += 2;            ptr += 2;
3681            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3682              { ptr += 2; continue; }              { ptr += 2; continue; }
3683            inescq = TRUE;            inescq = TRUE;
3684            break;            break;
# Line 3257  for (;; ptr++) Line 3708  for (;; ptr++)
3708            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3709            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3710    
3711            /* \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 */  
3712    
3713            if (d < 0)            if (d < 0)
3714              {              {
3715              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  
3716                {                {
3717                ptr = oldptr;                ptr = oldptr;
3718                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3430  for (;; ptr++) Line 3878  for (;; ptr++)
3878          }          }
3879        }        }
3880    
3881      /* 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.
3882        If we are at the end of an internal nested string, revert to the outer
3883        string. */
3884    
3885        while (((c = *(++ptr)) != 0 ||
3886               (nestptr != NULL &&
3887                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
3888               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3889    
3890      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));      /* Check for missing terminating ']' */
3891    
3892      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
3893        {        {
3894        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
3895        goto FAILED;        goto FAILED;
3896        }        }
3897    
   
 /* 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  
   
   
3898      /* 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
3899      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
3900      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 3522  we set the flag only if there is a liter Line 3958  we set the flag only if there is a liter
3958    
3959      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3960      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
3961      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
3962      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
3963      (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
3964      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
3965        actual compiled code. */
3966    
3967  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3968      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
3969        {        {
3970        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3971        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3554  we set the flag only if there is a liter Line 3991  we set the flag only if there is a liter
3991        }        }
3992  #endif  #endif
3993    
3994      /* 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
3995      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
3996      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
3997      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
3998        negating it if necessary. */
3999    
4000      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4001      if (negate_class)      if (negate_class)
# Line 3681  we set the flag only if there is a liter Line 4119  we set the flag only if there is a liter
4119    
4120        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4121            repeat_max < 0 &&            repeat_max < 0 &&
4122            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4123          {          {
4124          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4125          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3703  we set the flag only if there is a liter Line 4140  we set the flag only if there is a liter
4140        c = previous[1];        c = previous[1];
4141        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4142            repeat_max < 0 &&            repeat_max < 0 &&
4143            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4144          {          {
4145          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4146          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3727  we set the flag only if there is a liter Line 4164  we set the flag only if there is a liter
4164    
4165        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4166            repeat_max < 0 &&            repeat_max < 0 &&
4167            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4168          {          {
4169          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4170          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3749  we set the flag only if there is a liter Line 4186  we set the flag only if there is a liter
4186    
4187        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4188    
4189          /*--------------------------------------------------------------------*/
4190          /* This code is obsolete from release 8.00; the restriction was finally
4191          removed: */
4192    
4193        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4194        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4195    
4196        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4197          /*--------------------------------------------------------------------*/
4198    
4199        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4200    
# Line 3899  we set the flag only if there is a liter Line 4341  we set the flag only if there is a liter
4341          goto END_REPEAT;          goto END_REPEAT;
4342          }          }
4343    
4344          /*--------------------------------------------------------------------*/
4345          /* This code is obsolete from release 8.00; the restriction was finally
4346          removed: */
4347    
4348        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4349        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4350    
4351        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4352          /*--------------------------------------------------------------------*/
4353    
4354        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4355          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3927  we set the flag only if there is a liter Line 4374  we set the flag only if there is a liter
4374        {        {
4375        register int i;        register int i;
4376        int ketoffset = 0;        int ketoffset = 0;
4377        int len = code - previous;        int len = (int)(code - previous);
4378        uschar *bralink = NULL;        uschar *bralink = NULL;
4379    
4380        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless */
# Line 3948  we set the flag only if there is a liter Line 4395  we set the flag only if there is a liter
4395          {          {
4396          register uschar *ket = previous;          register uschar *ket = previous;
4397          do ket += GET(ket, 1); while (*ket != OP_KET);          do ket += GET(ket, 1); while (*ket != OP_KET);
4398          ketoffset = code - ket;          ketoffset = (int)(code - ket);
4399          }          }
4400    
4401        /* 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 4016  we set the flag only if there is a liter Line 4463  we set the flag only if there is a liter
4463            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4464            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4465    
4466            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4467            bralink = previous;            bralink = previous;
4468            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4469            }            }
# Line 4037  we set the flag only if there is a liter Line 4484  we set the flag only if there is a liter
4484            {            {
4485            /* 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
4486            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
4487            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4488              integer type when available, otherwise double. */
4489    
4490            if (lengthptr != NULL)            if (lengthptr != NULL)
4491              {              {
4492              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4493              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4494                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4495                        (INT64_OR_DOUBLE)INT_MAX ||
4496                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4497                {                {
4498                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 4089  we set the flag only if there is a liter Line 4538  we set the flag only if there is a liter
4538          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
4539          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
4540          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
4541          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4542            a 64-bit integer type when available, otherwise double. */
4543    
4544          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4545            {            {
4546            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4547                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4548            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4549                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4550                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4551                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4552              {              {
4553              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 4122  we set the flag only if there is a liter Line 4572  we set the flag only if there is a liter
4572              {              {
4573              int offset;              int offset;
4574              *code++ = OP_BRA;              *code++ = OP_BRA;
4575              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4576              bralink = code;              bralink = code;
4577              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4578              }              }
# Line 4143  we set the flag only if there is a liter Line 4593  we set the flag only if there is a liter
4593          while (bralink != NULL)          while (bralink != NULL)
4594            {            {
4595            int oldlinkoffset;            int oldlinkoffset;
4596            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4597            uschar *bra = code - offset;            uschar *bra = code - offset;
4598            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4599            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 4174  we set the flag only if there is a liter Line 4624  we set the flag only if there is a liter
4624            uschar *scode = bracode;            uschar *scode = bracode;
4625            do            do
4626              {              {
4627              if (could_be_empty_branch(scode, ketcode, utf8))              if (could_be_empty_branch(scode, ketcode, utf8, cd))
4628                {                {
4629                *bracode += OP_SBRA - OP_BRA;                *bracode += OP_SBRA - OP_BRA;
4630                break;                break;
# Line 4217  we set the flag only if there is a liter Line 4667  we set the flag only if there is a liter
4667      if (possessive_quantifier)      if (possessive_quantifier)
4668        {        {
4669        int len;        int len;
4670        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4671            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4672          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += _pcre_OP_lengths[*tempcode] +
4673            ((*tempcode == OP_TYPEEXACT &&            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4674               (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);  
4675        len = code - tempcode;        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4676            {
4677            tempcode += _pcre_OP_lengths[*tempcode];
4678    #ifdef SUPPORT_UTF8
4679            if (utf8 && tempcode[-1] >= 0xc0)
4680              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4681    #endif
4682            }
4683    
4684          len = (int)(code - tempcode);
4685        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4686          {          {
4687          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 4240  we set the flag only if there is a liter Line 4699  we set the flag only if there is a liter
4699          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4700          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4701    
4702            /* Because we are moving code along, we must ensure that any
4703            pending recursive references are updated. */
4704    
4705          default:          default:
4706            *code = OP_END;
4707            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4708          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4709          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4710          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4276  we set the flag only if there is a liter Line 4740  we set the flag only if there is a liter
4740    
4741      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4742    
4743      if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
4744             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
4745        {        {
4746        int i, namelen;        int i, namelen;
4747          int arglen = 0;
4748        const char *vn = verbnames;        const char *vn = verbnames;
4749        const uschar *name = ++ptr;        const uschar *name = ptr + 1;
4750          const uschar *arg = NULL;
4751        previous = NULL;        previous = NULL;
4752        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4753          namelen = (int)(ptr - name);
4754    
4755        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
4756          {          {
4757          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
4758          goto FAILED;          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
4759              || *ptr == '_') ptr++;
4760            arglen = (int)(ptr - arg);
4761          }          }
4762    
4763        if (*ptr != CHAR_RIGHT_PARENTHESIS)        if (*ptr != CHAR_RIGHT_PARENTHESIS)
4764          {          {
4765          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4766          goto FAILED;          goto FAILED;
4767          }          }
4768        namelen = ptr - name;  
4769          /* Scan the table of verb names */
4770    
4771        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
4772          {          {
4773          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4774              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4775            {            {
4776            *code = verbs[i].op;            /* Check for open captures before ACCEPT */
4777            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;  
4778            break;            if (verbs[i].op == OP_ACCEPT)
4779                {
4780                open_capitem *oc;
4781                cd->had_accept = TRUE;
4782                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4783                  {
4784                  *code++ = OP_CLOSE;
4785                  PUT2INC(code, 0, oc->number);
4786                  }
4787                }
4788    
4789              /* Handle the cases with/without an argument */
4790    
4791              if (arglen == 0)
4792                {
4793                if (verbs[i].op < 0)   /* Argument is mandatory */
4794                  {
4795                  *errorcodeptr = ERR66;
4796                  goto FAILED;
4797                  }
4798                *code++ = verbs[i].op;
4799                }
4800    
4801              else
4802                {
4803                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
4804                  {
4805                  *errorcodeptr = ERR59;
4806                  goto FAILED;
4807                  }
4808                *code++ = verbs[i].op_arg;
4809                *code++ = arglen;
4810                memcpy(code, arg, arglen);
4811                code += arglen;
4812                *code++ = 0;
4813                }
4814    
4815              break;  /* Found verb, exit loop */
4816            }            }
4817    
4818          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
4819          }          }
4820        if (i < verbcount) continue;  
4821        *errorcodeptr = ERR60;        if (i < verbcount) continue;    /* Successfully handled a verb */
4822          *errorcodeptr = ERR60;          /* Verb not recognized */
4823        goto FAILED;        goto FAILED;
4824        }        }
4825    
# Line 4425  we set the flag only if there is a liter Line 4938  we set the flag only if there is a liter
4938                recno * 10 + *ptr - CHAR_0 : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4939            ptr++;            ptr++;
4940            }            }
4941          namelen = ptr - name;          namelen = (int)(ptr - name);
4942    
4943          if ((terminator > 0 && *ptr++ != terminator) ||          if ((terminator > 0 && *ptr++ != terminator) ||
4944              *ptr++ != CHAR_RIGHT_PARENTHESIS)              *ptr++ != CHAR_RIGHT_PARENTHESIS)
4945            {            {
4946            ptr--;      /* Error offset */            ptr--;      /* Error offset */
# Line 4462  we set the flag only if there is a liter Line 4975  we set the flag only if there is a liter
4975            }            }
4976    
4977          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
4978          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
4979            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
4980            except they record that the reference was originally to a name. The
4981            information is used to check duplicate names. */
4982    
4983          slot = cd->name_table;          slot = cd->name_table;
4984          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 4477  we set the flag only if there is a liter Line 4993  we set the flag only if there is a liter
4993            {            {
4994            recno = GET2(slot, 0);            recno = GET2(slot, 0);
4995            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4996              code[1+LINK_SIZE]++;
4997            }            }
4998    
4999          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5000    
5001          else if ((i = find_parens(ptr, cd, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5002                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
5003            {            {
5004            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5005              code[1+LINK_SIZE]++;
5006            }            }
5007    
5008          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 4616  we set the flag only if there is a liter Line 5134  we set the flag only if there is a liter
5134              goto FAILED;              goto FAILED;
5135              }              }
5136            *code++ = n;            *code++ = n;
5137            PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */            PUT(code, 0, (int)(ptr - cd->start_pattern + 1)); /* Pattern offset */
5138            PUT(code, LINK_SIZE, 0);                    /* Default length */            PUT(code, LINK_SIZE, 0);                          /* Default length */
5139            code += 2 * LINK_SIZE;            code += 2 * LINK_SIZE;
5140            }            }
5141          previous = NULL;          previous = NULL;
# Line 4626  we set the flag only if there is a liter Line 5144  we set the flag only if there is a liter
5144    
5145          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5146          case CHAR_P:              /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
5147          if (*(++ptr) == CHAR_EQUALS_SIGN ||          if (*(++ptr) == CHAR_EQUALS_SIGN ||
5148              *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */              *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
5149            {            {
5150            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
# Line 4645  we set the flag only if there is a liter Line 5163  we set the flag only if there is a liter
5163          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
5164          case CHAR_APOSTROPHE:          case CHAR_APOSTROPHE:
5165            {            {
5166            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
5167              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5168            name = ++ptr;            name = ++ptr;
5169    
5170            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5171            namelen = ptr - name;            namelen = (int)(ptr - name);
5172    
5173            /* In the pre-compile phase, just do a syntax check. */            /* In the pre-compile phase, just do a syntax check. */
5174    
# Line 4677  we set the flag only if there is a liter Line 5195  we set the flag only if there is a liter
5195                }                }
5196              }              }
5197    
5198            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
5199              alphabetical order. Duplicate names for different numbers are
5200              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
5201              number are always OK. (An existing number can be re-used if (?|
5202              appears in the pattern.) In either event, a duplicate name results in
5203              a duplicate entry in the table, even if the number is the same. This
5204              is because the number of names, and hence the table size, is computed
5205              in the pre-compile, and it affects various numbers and pointers which
5206              would all have to be modified, and the compiled code moved down, if
5207              duplicates with the same number were omitted from the table. This
5208              doesn't seem worth the hassle. However, *different* names for the
5209              same number are not permitted. */
5210    
5211            else            else
5212              {              {
5213                BOOL dupname = FALSE;
5214              slot = cd->name_table;              slot = cd->name_table;
5215    
5216              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
5217                {                {
5218                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 4689  we set the flag only if there is a liter Line 5220  we set the flag only if there is a liter
5220                  {                  {
5221                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
5222                    {                    {
5223                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
5224                          (options & PCRE_DUPNAMES) == 0)
5225                      {                      {
5226                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
5227                      goto FAILED;                      goto FAILED;
5228                      }                      }
5229                      else dupname = TRUE;
5230                    }                    }
5231                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
5232                  }                  }
5233    
5234                  /* Make space in the table and break the loop for an earlier
5235                  name. For a duplicate or later name, carry on. We do this for
5236                  duplicates so that in the simple case (when ?(| is not used) they
5237                  are in order of their numbers. */
5238    
5239                if (crc < 0)                if (crc < 0)
5240                  {                  {
5241                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
5242                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
5243                  break;                  break;
5244                  }                  }
5245    
5246                  /* Continue the loop for a later or duplicate name */
5247    
5248                slot += cd->name_entry_size;                slot += cd->name_entry_size;
5249                }                }
5250    
5251                /* For non-duplicate names, check for a duplicate number before
5252                adding the new name. */
5253    
5254                if (!dupname)
5255                  {
5256                  uschar *cslot = cd->name_table;
5257                  for (i = 0; i < cd->names_found; i++)
5258                    {
5259                    if (cslot != slot)
5260                      {
5261                      if (GET2(cslot, 0) == cd->bracount + 1)
5262                        {
5263                        *errorcodeptr = ERR65;
5264                        goto FAILED;
5265                        }
5266                      }
5267                    else i--;
5268                    cslot += cd->name_entry_size;
5269                    }
5270                  }
5271    
5272              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
5273              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
5274              slot[2+namelen] = 0;              slot[2+namelen] = 0;
5275              }              }
5276            }            }
5277    
5278          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
5279            encountered. */
5280    
         ptr++;                    /* Move past > or ' */  
5281          cd->names_found++;          cd->names_found++;
5282            ptr++;                    /* Move past > or ' */
5283          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
5284    
5285    
# Line 4734  we set the flag only if there is a liter Line 5298  we set the flag only if there is a liter
5298          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
5299          name = ++ptr;          name = ++ptr;
5300          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5301          namelen = ptr - name;          namelen = (int)(ptr - name);
5302    
5303          /* In the pre-compile phase, do a syntax check and set a dummy          /* In the pre-compile phase, do a syntax check and set a dummy
5304          reference number. */          reference number. */
# Line 4780  we set the flag only if there is a liter Line 5344  we set the flag only if there is a liter
5344              recno = GET2(slot, 0);              recno = GET2(slot, 0);
5345              }              }
5346            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5347                      find_parens(ptr, cd, name, namelen,                      find_parens(cd, name, namelen,
5348                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
5349              {              {
5350              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 4884  we set the flag only if there is a liter Line 5448  we set the flag only if there is a liter
5448            if (lengthptr == NULL)            if (lengthptr == NULL)
5449              {              {
5450              *code = OP_END;              *code = OP_END;
5451              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);              if (recno != 0)
5452                  called = _pcre_find_bracket(cd->start_code, utf8, recno);
5453    
5454              /* Forward reference */              /* Forward reference */
5455    
5456              if (called == NULL)              if (called == NULL)
5457                {                {
5458                if (find_parens(ptr, cd, NULL, recno,                if (find_parens(cd, NULL, recno,
5459                      (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
5460                  {                  {
5461                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5462                  goto FAILED;                  goto FAILED;
5463                  }                  }
5464    
5465                  /* Fudge the value of "called" so that when it is inserted as an
5466                  offset below, what it actually inserted is the reference number
5467                  of the group. */
5468    
5469                called = cd->start_code + recno;                called = cd->start_code + recno;
5470                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, (int)(code + 2 + LINK_SIZE - cd->start_code));
5471                }                }
5472    
5473              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
# Line 4905  we set the flag only if there is a liter Line 5475  we set the flag only if there is a liter
5475              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. */
5476    
5477              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 &&
5478                       could_be_empty(called, code, bcptr, utf8))                       could_be_empty(called, code, bcptr, utf8, cd))
5479                {                {
5480                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
5481                goto FAILED;                goto FAILED;
# Line 4921  we set the flag only if there is a liter Line 5491  we set the flag only if there is a liter
5491            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
5492    
5493            *code = OP_RECURSE;            *code = OP_RECURSE;
5494            PUT(code, 1, called - cd->start_code);            PUT(code, 1, (int)(called - cd->start_code));
5495            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
5496    
5497            *code = OP_KET;            *code = OP_KET;
# Line 5000  we set the flag only if there is a liter Line 5570  we set the flag only if there is a liter
5570              {              {
5571              cd->external_options = newoptions;              cd->external_options = newoptions;
5572              }              }
5573           else            else
5574              {              {
5575              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))
5576                {                {
# Line 5032  we set the flag only if there is a liter Line 5602  we set the flag only if there is a liter
5602          }     /* End of switch for character following (? */          }     /* End of switch for character following (? */
5603        }       /* End of (? handling */        }       /* End of (? handling */
5604    
5605      /* Opening parenthesis not followed by '?'. If PCRE_NO_AUTO_CAPTURE is set,      /* Opening parenthesis not followed by '*' or '?'. If PCRE_NO_AUTO_CAPTURE
5606      all unadorned brackets become non-capturing and behave like (?:...)      is set, all unadorned brackets become non-capturing and behave like (?:...)
5607      brackets. */      brackets. */
5608    
5609      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)
# Line 5225  we set the flag only if there is a liter Line 5795  we set the flag only if there is a liter
5795    
5796      /* ===================================================================*/      /* ===================================================================*/
5797      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values
5798      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
5799      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
5800      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
5801      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
5802      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
5803        ever created. */
5804    
5805      case CHAR_BACKSLASH:      case CHAR_BACKSLASH:
5806      tempptr = ptr;      tempptr = ptr;
# Line 5240  we set the flag only if there is a liter Line 5811  we set the flag only if there is a liter
5811        {        {
5812        if (-c == ESC_Q)            /* Handle start of quoted string */        if (-c == ESC_Q)            /* Handle start of quoted string */
5813          {          {
5814          if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)          if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5815            ptr += 2;               /* avoid empty string */            ptr += 2;               /* avoid empty string */
5816              else inescq = TRUE;              else inescq = TRUE;
5817          continue;          continue;
# Line 5270  we set the flag only if there is a liter Line 5841  we set the flag only if there is a liter
5841          {          {
5842          const uschar *p;          const uschar *p;
5843          save_hwm = cd->hwm;   /* Normally this is set when '(' is read */          save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5844          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5845            CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;            CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5846    
5847          /* These two statements stop the compiler for warning about possibly          /* These two statements stop the compiler for warning about possibly
# Line 5321  we set the flag only if there is a liter Line 5892  we set the flag only if there is a liter
5892        /* \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).
5893        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax) */
5894    
5895        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||
5896            ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))            ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))
5897          {          {
5898          is_recurse = FALSE;          is_recurse = FALSE;
5899          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5900            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
5901            CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;            CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
5902          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
5903          }          }
# Line 5337  we set the flag only if there is a liter Line 5908  we set the flag only if there is a liter
5908    
5909        if (-c >= ESC_REF)        if (-c >= ESC_REF)
5910          {          {
5911            open_capitem *oc;
5912          recno = -c - ESC_REF;          recno = -c - ESC_REF;
5913    
5914          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
# Line 5346  we set the flag only if there is a liter Line 5918  we set the flag only if there is a liter
5918          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
5919          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
5920          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
5921    
5922            /* Check to see if this back reference is recursive, that it, it
5923            is inside the group that it references. A flag is set so that the
5924            group can be made atomic. */
5925    
5926            for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5927              {
5928              if (oc->number == recno)
5929                {
5930                oc->flag = TRUE;
5931                break;
5932                }
5933              }
5934          }          }
5935    
5936        /* So are Unicode property matches, if supported. */        /* So are Unicode property matches, if supported. */
# Line 5375  we set the flag only if there is a liter Line 5960  we set the flag only if there is a liter
5960  #endif  #endif
5961    
5962        /* For the rest (including \X when Unicode properties are supported), we        /* For the rest (including \X when Unicode properties are supported), we
5963        can obtain the OP value by negating the escape value. */        can obtain the OP value by negating the escape value in the default
5964          situation when PCRE_UCP is not set. When it *is* set, we substitute
5965          Unicode property tests. */
5966    
5967        else        else
5968          {          {
5969          previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;  #ifdef SUPPORT_UCP
5970          *code++ = -c;          if (-c >= ESC_DU && -c <= ESC_wu)
5971              {
5972              nestptr = ptr + 1;                   /* Where to resume */
5973              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
5974              }
5975            else
5976    #endif
5977              {
5978              previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;
5979              *code++ = -c;
5980              }
5981          }          }
5982        continue;        continue;
5983        }        }
# Line 5528  uschar *code = *codeptr; Line 6125  uschar *code = *codeptr;
6125  uschar *last_branch = code;  uschar *last_branch = code;
6126  uschar *start_bracket = code;  uschar *start_bracket = code;
6127  uschar *reverse_count = NULL;  uschar *reverse_count = NULL;
6128    open_capitem capitem;
6129    int capnumber = 0;
6130  int firstbyte, reqbyte;  int firstbyte, reqbyte;
6131  int branchfirstbyte, branchreqbyte;  int branchfirstbyte, branchreqbyte;
6132  int length;  int length;
6133  int orig_bracount;  int orig_bracount;
6134  int max_bracount;  int max_bracount;
6135    int old_external_options = cd->external_options;
6136  branch_chain bc;  branch_chain bc;
6137    
6138  bc.outer = bcptr;  bc.outer = bcptr;
6139  bc.current = code;  bc.current_branch = code;
6140    
6141  firstbyte = reqbyte = REQ_UNSET;  firstbyte = reqbyte = REQ_UNSET;
6142    
# Line 5554  the code that abstracts option settings Line 6154  the code that abstracts option settings
6154  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
6155  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. */
6156    
6157    /* If this is a capturing subpattern, add to the chain of open capturing items
6158    so that we can detect them if (*ACCEPT) is encountered. This is also used to
6159    detect groups that contain recursive back references to themselves. */
6160    
6161    if (*code == OP_CBRA)
6162      {
6163      capnumber = GET2(code, 1 + LINK_SIZE);
6164      capitem.number = capnumber;
6165      capitem.next = cd->open_caps;
6166      capitem.flag = FALSE;
6167      cd->open_caps = &capitem;
6168      }
6169    
6170  /* Offset is set zero to mark that this bracket is still open */  /* Offset is set zero to mark that this bracket is still open */
6171    
6172  PUT(code, 1, 0);  PUT(code, 1, 0);
# Line 5598  for (;;) Line 6211  for (;;)
6211      return FALSE;      return FALSE;
6212      }      }
6213    
6214      /* If the external options have changed during this branch, it means that we
6215      are at the top level, and a leading option setting has been encountered. We
6216      need to re-set the original option values to take account of this so that,
6217      during the pre-compile phase, we know to allow for a re-set at the start of
6218      subsequent branches. */
6219    
6220      if (old_external_options != cd->external_options)
6221        oldims = cd->external_options & PCRE_IMS;
6222    
6223    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6224    has fewer than the rest. */    has fewer than the rest. */
6225    
# Line 5648  for (;;) Line 6270  for (;;)
6270    
6271      /* If lookbehind, check that this branch matches a fixed-length string, and      /* If lookbehind, check that this branch matches a fixed-length string, and
6272      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
6273      branch with OP_END. */      branch with OP_END. If the branch contains OP_RECURSE, the result is -3
6274        because there may be forward references that we can't check here. Set a
6275        flag to cause another lookbehind check at the end. Why not do it all at the
6276        end? Because common, erroneous checks are picked up here and the offset of
6277        the problem can be shown. */
6278    
6279      if (lookbehind)      if (lookbehind)
6280        {        {
6281        int fixed_length;        int fixed_length;
6282        *code = OP_END;        *code = OP_END;
6283        fixed_length = find_fixedlength(last_branch, options);        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);
6284        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6285        if (fixed_length < 0)        if (fixed_length == -3)
6286            {
6287            cd->check_lookbehind = TRUE;
6288            }
6289          else if (fixed_length < 0)
6290          {          {
6291          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;
6292          *ptrptr = ptr;          *ptrptr = ptr;
6293          return FALSE;          return FALSE;
6294          }          }
6295        PUT(reverse_count, 0, fixed_length);        else { PUT(reverse_count, 0, fixed_length); }
6296        }        }
6297      }      }
6298    
# Line 5679  for (;;) Line 6309  for (;;)
6309      {      {
6310      if (lengthptr == NULL)      if (lengthptr == NULL)
6311        {        {
6312        int branch_length = code - last_branch;        int branch_length = (int)(code - last_branch);
6313        do        do
6314          {          {
6315          int prev_length = GET(last_branch, 1);          int prev_length = GET(last_branch, 1);
# Line 5693  for (;;) Line 6323  for (;;)
6323      /* Fill in the ket */      /* Fill in the ket */
6324    
6325      *code = OP_KET;      *code = OP_KET;
6326      PUT(code, 1, code - start_bracket);      PUT(code, 1, (int)(code - start_bracket));
6327      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6328    
6329      /* Resetting option if needed */      /* If it was a capturing subpattern, check to see if it contained any
6330        recursive back references. If so, we must wrap it in atomic brackets.
6331        In any event, remove the block from the chain. */
6332    
6333        if (capnumber > 0)
6334          {
6335          if (cd->open_caps->flag)
6336            {
6337            memmove(start_bracket + 1 + LINK_SIZE, start_bracket,
6338              code - start_bracket);
6339            *start_bracket = OP_ONCE;
6340            code += 1 + LINK_SIZE;
6341            PUT(start_bracket, 1, (int)(code - start_bracket));
6342            *code = OP_KET;
6343            PUT(code, 1, (int)(code - start_bracket));
6344            code += 1 + LINK_SIZE;
6345            length += 2 + 2*LINK_SIZE;
6346            }
6347          cd->open_caps = cd->open_caps->next;
6348          }
6349    
6350        /* Reset options if needed. */
6351    
6352      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)
6353        {        {
# Line 5744  for (;;) Line 6395  for (;;)
6395    else    else
6396      {      {
6397      *code = OP_ALT;      *code = OP_ALT;
6398      PUT(code, 1, code - last_branch);      PUT(code, 1, (int)(code - last_branch));
6399      bc.current = last_branch = code;      bc.current_branch = last_branch = code;
6400      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6401      }      }
6402    
# Line 5879  do { Line 6530  do {
6530     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6531       NULL, 0, FALSE);       NULL, 0, FALSE);
6532     register int op = *scode;     register int op = *scode;
6533    
6534     /* 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
6535     conditional assertion *and* what follows the condition must satisfy the test     conditional assertion *and* what follows the condition must satisfy the test
6536     for start of line. Other kinds of condition fail. Note that there may be an     for start of line. Other kinds of condition fail. Note that there may be an
# Line 5887  do { Line 6538  do {
6538    
6539     if (op == OP_COND)     if (op == OP_COND)
6540       {       {
6541       scode += 1 + LINK_SIZE;       scode += 1 + LINK_SIZE;
6542       if (*scode == OP_CALLOUT) scode += _pcre_OP_lengths[OP_CALLOUT];       if (*scode == OP_CALLOUT) scode += _pcre_OP_lengths[OP_CALLOUT];
6543       switch (*scode)       switch (*scode)
6544         {         {
6545         case OP_CREF:         case OP_CREF:
6546           case OP_NCREF:
6547         case OP_RREF:         case OP_RREF:
6548           case OP_NRREF:
6549         case OP_DEF:         case OP_DEF:
6550         return FALSE;         return FALSE;
6551    
6552         default:     /* Assertion */         default:     /* Assertion */
6553         if (!is_startline(scode, bracket_map, backref_map)) return FALSE;         if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6554         do scode += GET(scode, 1); while (*scode == OP_ALT);         do scode += GET(scode, 1); while (*scode == OP_ALT);
6555         scode += 1 + LINK_SIZE;         scode += 1 + LINK_SIZE;
6556         break;         break;
6557         }         }
6558       scode = first_significant_code(scode, NULL, 0, FALSE);       scode = first_significant_code(scode, NULL, 0, FALSE);
6559       op = *scode;       op = *scode;
6560       }       }
6561    
6562     /* Non-capturing brackets */     /* Non-capturing brackets */
6563    
# Line 5925  do { Line 6578  do {
6578     /* Other brackets */     /* Other brackets */
6579    
6580     else if (op == OP_ASSERT || op == OP_ONCE)     else if (op == OP_ASSERT || op == OP_ONCE)
6581       {       {
6582       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6583       }       }
6584    
6585     /* .* means "start at start or after \n" if it isn't in brackets that     /* .* means "start at start or after \n" if it isn't in brackets that
# Line 6061  int length = 1;  /* For final END opcode Line 6714  int length = 1;  /* For final END opcode
6714  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
6715  int errorcode = 0;  int errorcode = 0;
6716  int skipatstart = 0;  int skipatstart = 0;
 #ifdef SUPPORT_UTF8  
6717  BOOL utf8;  BOOL utf8;
 #endif  
6718  size_t size;  size_t size;
6719  uschar *code;  uschar *code;
6720  const uschar *codestart;  const uschar *codestart;
# Line 6106  if (erroroffset == NULL) Line 6757  if (erroroffset == NULL)
6757    
6758  *erroroffset = 0;  *erroroffset = 0;
6759    
 /* Can't support UTF8 unless PCRE has been compiled to include the code. */  
   
 #ifdef SUPPORT_UTF8  
 utf8 = (options & PCRE_UTF8) != 0;  
 if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&  
      (*erroroffset = _pcre_valid_utf8((uschar *)pattern, -1)) >= 0)  
   {  
   errorcode = ERR44;  
   goto PCRE_EARLY_ERROR_RETURN2;  
   }  
 #else  
 if ((options & PCRE_UTF8) != 0)  
   {  
   errorcode = ERR32;  
   goto PCRE_EARLY_ERROR_RETURN;  
   }  
 #endif  
   
 if ((options & ~PUBLIC_COMPILE_OPTIONS) != 0)  
   {  
   errorcode = ERR17;  
   goto PCRE_EARLY_ERROR_RETURN;  
   }  
   
6760  /* Set up pointers to the individual character tables */  /* Set up pointers to the individual character tables */
6761    
6762  if (tables == NULL) tables = _pcre_default_tables;  if (tables == NULL) tables = _pcre_default_tables;
# Line 6138  cd->fcc = tables + fcc_offset; Line 6765  cd->fcc = tables + fcc_offset;
6765  cd->cbits = tables + cbits_offset;  cd->cbits = tables + cbits_offset;
6766  cd->ctypes = tables + ctypes_offset;  cd->ctypes = tables + ctypes_offset;
6767    
6768    /* Check that all undefined public option bits are zero */
6769    
6770    if ((options & ~PUBLIC_COMPILE_OPTIONS) != 0)
6771      {
6772      errorcode = ERR17;
6773      goto PCRE_EARLY_ERROR_RETURN;
6774      }
6775    
6776  /* Check for global one-time settings at the start of the pattern, and remember  /* Check for global one-time settings at the start of the pattern, and remember
6777  the offset for later. */  the offset for later. */
6778    
6779  while (ptr[skipatstart] == CHAR_LEFT_PARENTHESIS &&  while (ptr[skipatstart] == CHAR_LEFT_PARENTHESIS &&
6780         ptr[skipatstart+1] == CHAR_ASTERISK)         ptr[skipatstart+1] == CHAR_ASTERISK)
6781    {    {
6782    int newnl = 0;    int newnl = 0;
6783    int newbsr = 0;    int newbsr = 0;
6784    
6785      if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)
6786        { skipatstart += 7; options |= PCRE_UTF8; continue; }
6787      else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)
6788        { skipatstart += 6; options |= PCRE_UCP; continue; }
6789    
6790    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
6791      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
6792    else if (strncmp((char *)(ptr+skipatstart+2), STRING_LF_RIGHTPAR, 3)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_LF_RIGHTPAR, 3)  == 0)
# Line 6170  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 6810  while (ptr[skipatstart] == CHAR_LEFT_PAR
6810    else break;    else break;
6811    }    }
6812    
6813    utf8 = (options & PCRE_UTF8) != 0;
6814    
6815    /* Can't support UTF8 unless PCRE has been compiled to include the code. */
6816    
6817    #ifdef SUPPORT_UTF8
6818    if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
6819         (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)
6820      {
6821      errorcode = ERR44;
6822      goto PCRE_EARLY_ERROR_RETURN2;
6823      }
6824    #else
6825    if (utf8)
6826      {
6827      errorcode = ERR32;
6828      goto PCRE_EARLY_ERROR_RETURN;
6829      }
6830    #endif
6831    
6832    /* Can't support UCP unless PCRE has been compiled to include the code. */
6833    
6834    #ifndef SUPPORT_UCP
6835    if ((options & PCRE_UCP) != 0)
6836      {
6837      errorcode = ERR67;
6838      goto PCRE_EARLY_ERROR_RETURN;
6839      }
6840    #endif
6841    
6842  /* Check validity of \R options. */  /* Check validity of \R options. */
6843    
6844  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
# Line 6252  cd->end_pattern = (const uschar *)(patte Line 6921  cd->end_pattern = (const uschar *)(patte
6921  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6922  cd->external_options = options;  cd->external_options = options;
6923  cd->external_flags = 0;  cd->external_flags = 0;
6924    cd->open_caps = NULL;
6925    
6926  /* 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
6927  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 6297  regex compiled on a system with 4-byte p Line 6967  regex compiled on a system with 4-byte p
6967  pointers. */  pointers. */
6968    
6969  re->magic_number = MAGIC_NUMBER;  re->magic_number = MAGIC_NUMBER;
6970  re->size = size;  re->size = (int)size;
6971  re->options = cd->external_options;  re->options = cd->external_options;
6972  re->flags = cd->external_flags;  re->flags = cd->external_flags;
6973  re->dummy1 = 0;  re->dummy1 = 0;
# Line 6326  cd->start_code = codestart; Line 6996  cd->start_code = codestart;
6996  cd->hwm = cworkspace;  cd->hwm = cworkspace;
6997  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6998  cd->had_accept = FALSE;  cd->had_accept = FALSE;
6999    cd->check_lookbehind = FALSE;
7000    cd->open_caps = NULL;
7001    
7002  /* Set up a starting, non-extracting bracket, then compile the expression. On  /* Set up a starting, non-extracting bracket, then compile the expression. On
7003  error, errorcode will be set non-zero, so we don't need to look at the result  error, errorcode will be set non-zero, so we don't need to look at the result
# Line 6351  if debugging, leave the test till after Line 7023  if debugging, leave the test till after
7023    
7024  *code++ = OP_END;  *code++ = OP_END;
7025    
7026  #ifndef DEBUG  #ifndef PCRE_DEBUG
7027  if (code - codestart > length) errorcode = ERR23;  if (code - codestart > length) errorcode = ERR23;
7028  #endif  #endif
7029    
# Line 6364  while (errorcode == 0 && cd->hwm > cwork Line 7036  while (errorcode == 0 && cd->hwm > cwork
7036    cd->hwm -= LINK_SIZE;    cd->hwm -= LINK_SIZE;
7037    offset = GET(cd->hwm, 0);    offset = GET(cd->hwm, 0);
7038    recno = GET(codestart, offset);    recno = GET(codestart, offset);
7039    groupptr = find_bracket(codestart, (re->options & PCRE_UTF8) != 0, recno);    groupptr = _pcre_find_bracket(codestart, utf8, recno);
7040    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
7041      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, (int)(groupptr - codestart));
7042    }    }
7043    
7044  /* Give an error if there's back reference to a non-existent capturing  /* Give an error if there's back reference to a non-existent capturing
# Line 6374  subpattern. */ Line 7046  subpattern. */
7046    
7047  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;
7048    
7049    /* If there were any lookbehind assertions that contained OP_RECURSE
7050    (recursions or subroutine calls), a flag is set for them to be checked here,
7051    because they may contain forward references. Actual recursions can't be fixed
7052    length, but subroutine calls can. It is done like this so that those without
7053    OP_RECURSE that are not fixed length get a diagnosic with a useful offset. The
7054    exceptional ones forgo this. We scan the pattern to check that they are fixed
7055    length, and set their lengths. */
7056    
7057    if (cd->check_lookbehind)
7058      {
7059      uschar *cc = (uschar *)codestart;
7060    
7061      /* Loop, searching for OP_REVERSE items, and process those that do not have
7062      their length set. (Actually, it will also re-process any that have a length
7063      of zero, but that is a pathological case, and it does no harm.) When we find
7064      one, we temporarily terminate the branch it is in while we scan it. */
7065    
7066      for (cc = (uschar *)_pcre_find_bracket(codestart, utf8, -1);
7067           cc != NULL;
7068           cc = (uschar *)_pcre_find_bracket(cc, utf8, -1))
7069        {
7070        if (GET(cc, 1) == 0)
7071          {
7072          int fixed_length;
7073          uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
7074          int end_op = *be;
7075          *be = OP_END;
7076          fixed_length = find_fixedlength(cc, re->options, TRUE, cd);
7077          *be = end_op;
7078          DPRINTF(("fixed length = %d\n", fixed_length));
7079          if (fixed_length < 0)
7080            {
7081            errorcode = (fixed_length == -2)? ERR36 : ERR25;
7082            break;
7083            }
7084          PUT(cc, 1, fixed_length);
7085          }
7086        cc += 1 + LINK_SIZE;
7087        }
7088      }
7089    
7090  /* Failed to compile, or error while post-processing */  /* Failed to compile, or error while post-processing */
7091    
7092  if (errorcode != 0)  if (errorcode != 0)
7093    {    {
7094    (pcre_free)(re);    (pcre_free)(re);
7095    PCRE_EARLY_ERROR_RETURN:    PCRE_EARLY_ERROR_RETURN:
7096    *erroroffset = ptr - (const uschar *)pattern;    *erroroffset = (int)(ptr - (const uschar *)pattern);
7097    PCRE_EARLY_ERROR_RETURN2:    PCRE_EARLY_ERROR_RETURN2:
7098    *errorptr = find_error_text(errorcode);    *errorptr = find_error_text(errorcode);
7099    if (errorcodeptr != NULL) *errorcodeptr = errorcode;    if (errorcodeptr != NULL) *errorcodeptr = errorcode;
# Line 6434  if (reqbyte >= 0 && Line 7147  if (reqbyte >= 0 &&
7147  /* Print out the compiled data if debugging is enabled. This is never the  /* Print out the compiled data if debugging is enabled. This is never the
7148  case when building a production library. */  case when building a production library. */
7149    
7150  #ifdef DEBUG  #ifdef PCRE_DEBUG
   
7151  printf("Length = %d top_bracket = %d top_backref = %d\n",  printf("Length = %d top_bracket = %d top_backref = %d\n",
7152    length, re->top_bracket, re->top_backref);    length, re->top_bracket, re->top_backref);
7153    
# Line 6472  if (code - codestart > length) Line 7184  if (code - codestart > length)
7184    if (errorcodeptr != NULL) *errorcodeptr = ERR23;    if (errorcodeptr != NULL) *errorcodeptr = ERR23;
7185    return NULL;    return NULL;
7186    }    }
7187  #endif   /* DEBUG */  #endif   /* PCRE_DEBUG */
7188    
7189  return (pcre *)re;  return (pcre *)re;
7190  }  }

Legend:
Removed from v.391  
changed lines
  Added in v.535

  ViewVC Help
Powered by ViewVC 1.1.5