/[pcre]/code/branches/pcre16/pcre_compile.c
ViewVC logotype

Diff of /code/branches/pcre16/pcre_compile.c

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

revision 406 by ph10, Mon Mar 23 12:05:43 2009 UTC revision 556 by ph10, Tue Oct 26 11:06:44 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 118  static const short int escapes[] = { Line 124  static const short int escapes[] = {
124       -ESC_H,                  0,       -ESC_H,                  0,
125       0,                       -ESC_K,       0,                       -ESC_K,
126       0,                       0,       0,                       0,
127       0,                       0,       -ESC_N,                  0,
128       -ESC_P,                  -ESC_Q,       -ESC_P,                  -ESC_Q,
129       -ESC_R,                  -ESC_S,       -ESC_R,                  -ESC_S,
130       0,                       0,       0,                       0,
# Line 165  static const short int escapes[] = { Line 171  static const short int escapes[] = {
171  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
172  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
173  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
175  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
177  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
# Line 182  string is built from string macros so th Line 188  string is built from string macros so th
188  platforms. */  platforms. */
189    
190  typedef struct verbitem {  typedef struct verbitem {
191    int   len;    int   len;                 /* Length of verb name */
192    int   op;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
193      int   op_arg;              /* Op when arg present, or -1 if not allowed */
194  } verbitem;  } verbitem;
195    
196  static const char verbnames[] =  static const char verbnames[] =
197      "\0"                       /* Empty name is a shorthand for MARK */
198      STRING_MARK0
199    STRING_ACCEPT0    STRING_ACCEPT0
200    STRING_COMMIT0    STRING_COMMIT0
201    STRING_F0    STRING_F0
# Line 196  static const char verbnames[] = Line 205  static const char verbnames[] =
205    STRING_THEN;    STRING_THEN;
206    
207  static const verbitem verbs[] = {  static const verbitem verbs[] = {
208    { 6, OP_ACCEPT },    { 0, -1,        OP_MARK },
209    { 6, OP_COMMIT },    { 4, -1,        OP_MARK },
210    { 1, OP_FAIL },    { 6, OP_ACCEPT, -1 },
211    { 4, OP_FAIL },    { 6, OP_COMMIT, -1 },
212    { 5, OP_PRUNE },    { 1, OP_FAIL,   -1 },
213    { 4, OP_SKIP  },    { 4, OP_FAIL,   -1 },
214    { 4, OP_THEN  }    { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217  };  };
218    
219  static const int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
# Line 250  static const int posix_class_maps[] = { Line 261  static const int posix_class_maps[] = {
261    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
262  };  };
263    
264    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
265    substitutes must be in the order of the names, defined above, and there are
266    both positive and negative cases. NULL means no substitute. */
267    
268    #ifdef SUPPORT_UCP
269    static const uschar *substitutes[] = {
270      (uschar *)"\\P{Nd}",    /* \D */
271      (uschar *)"\\p{Nd}",    /* \d */
272      (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */
273      (uschar *)"\\p{Xsp}",   /* \s */
274      (uschar *)"\\P{Xwd}",   /* \W */
275      (uschar *)"\\p{Xwd}"    /* \w */
276    };
277    
278    static const uschar *posix_substitutes[] = {
279      (uschar *)"\\p{L}",     /* alpha */
280      (uschar *)"\\p{Ll}",    /* lower */
281      (uschar *)"\\p{Lu}",    /* upper */
282      (uschar *)"\\p{Xan}",   /* alnum */
283      NULL,                   /* ascii */
284      (uschar *)"\\h",        /* blank */
285      NULL,                   /* cntrl */
286      (uschar *)"\\p{Nd}",    /* digit */
287      NULL,                   /* graph */
288      NULL,                   /* print */
289      NULL,                   /* punct */
290      (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */
291      (uschar *)"\\p{Xwd}",   /* word */
292      NULL,                   /* xdigit */
293      /* Negated cases */
294      (uschar *)"\\P{L}",     /* ^alpha */
295      (uschar *)"\\P{Ll}",    /* ^lower */
296      (uschar *)"\\P{Lu}",    /* ^upper */
297      (uschar *)"\\P{Xan}",   /* ^alnum */
298      NULL,                   /* ^ascii */
299      (uschar *)"\\H",        /* ^blank */
300      NULL,                   /* ^cntrl */
301      (uschar *)"\\P{Nd}",    /* ^digit */
302      NULL,                   /* ^graph */
303      NULL,                   /* ^print */
304      NULL,                   /* ^punct */
305      (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */
306      (uschar *)"\\P{Xwd}",   /* ^word */
307      NULL                    /* ^xdigit */
308    };
309    #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))
310    #endif
311    
312  #define STRING(a)  # a  #define STRING(a)  # a
313  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 262  the number of relocations needed when a Line 320  the number of relocations needed when a
320  it is now one long string. We cannot use a table of offsets, because the  it is now one long string. We cannot use a table of offsets, because the
321  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
322  simply count through to the one we want - this isn't a performance issue  simply count through to the one we want - this isn't a performance issue
323  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
324    
325    Each substring ends with \0 to insert a null character. This includes the final
326    substring, so that the whole string ends with \0\0, which can be detected when
327    counting through. */
328    
329  static const char error_texts[] =  static const char error_texts[] =
330    "no error\0"    "no error\0"
# Line 309  static const char error_texts[] = Line 371  static const char error_texts[] =
371    /* 35 */    /* 35 */
372    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
373    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
374    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
375    "number after (?C is > 255\0"    "number after (?C is > 255\0"
376    "closing ) for (?C expected\0"    "closing ) for (?C expected\0"
377    /* 40 */    /* 40 */
# Line 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 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      utf8         TRUE if we are in UTF-8 mode
1114      count        pointer to the current capturing subpattern number (updated)
1115    
1116  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1117  */  */
1118    
1119  static int  static int
1120  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,
1121    BOOL xmode)    BOOL xmode, BOOL utf8, int *count)
1122  {  {
1123  const uschar *thisname;  uschar *ptr = *ptrptr;
1124  int count = cd->bracount;  int start_count = *count;
1125    int hwm_count = start_count;
1126    BOOL dup_parens = FALSE;
1127    
1128  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1129    dealing with. The very first call may not start with a parenthesis. */
1130    
1131    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1132    {    {
1133    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1134    
1135      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1136    
1137      /* Handle a normal, unnamed capturing parenthesis. */
1138    
1139      else if (ptr[1] != CHAR_QUESTION_MARK)
1140        {
1141        *count += 1;
1142        if (name == NULL && *count == lorn) return *count;
1143        ptr++;
1144        }
1145    
1146      /* All cases now have (? at the start. Remember when we are in a group
1147      where the parenthesis numbers are duplicated. */
1148    
1149      else if (ptr[2] == CHAR_VERTICAL_LINE)
1150        {
1151        ptr += 3;
1152        dup_parens = TRUE;
1153        }
1154    
1155      /* Handle comments; all characters are allowed until a ket is reached. */
1156    
1157      else if (ptr[2] == CHAR_NUMBER_SIGN)
1158        {
1159        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1160        goto FAIL_EXIT;
1161        }
1162    
1163      /* Handle a condition. If it is an assertion, just carry on so that it
1164      is processed as normal. If not, skip to the closing parenthesis of the
1165      condition (there can't be any nested parens). */
1166    
1167      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1168        {
1169        ptr += 2;
1170        if (ptr[1] != CHAR_QUESTION_MARK)
1171          {
1172          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1173          if (*ptr != 0) ptr++;
1174          }
1175        }
1176    
1177      /* Start with (? but not a condition. */
1178    
1179      else
1180        {
1181        ptr += 2;
1182        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1183    
1184        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1185    
1186        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1187            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1188          {
1189          int term;
1190          const uschar *thisname;
1191          *count += 1;
1192          if (name == NULL && *count == lorn) return *count;
1193          term = *ptr++;
1194          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1195          thisname = ptr;
1196          while (*ptr != term) ptr++;
1197          if (name != NULL && lorn == ptr - thisname &&
1198              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1199            return *count;
1200          term++;
1201          }
1202        }
1203      }
1204    
1205    /* Past any initial parenthesis handling, scan for parentheses or vertical
1206    bars. */
1207    
1208    for (; *ptr != 0; ptr++)
1209      {
1210    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1211    
1212    if (*ptr == CHAR_BACKSLASH)    if (*ptr == CHAR_BACKSLASH)
1213      {      {
1214      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1215      if (*ptr == CHAR_Q) for (;;)      if (*ptr == CHAR_Q) for (;;)
1216        {        {
1217        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1218        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1219        if (*(++ptr) == CHAR_E) break;        if (*(++ptr) == CHAR_E) break;
1220        }        }
1221      continue;      continue;
# Line 1065  for (; *ptr != 0; ptr++) Line 1232  for (; *ptr != 0; ptr++)
1232      BOOL negate_class = FALSE;      BOOL negate_class = FALSE;
1233      for (;;)      for (;;)
1234        {        {
1235        int c = *(++ptr);        if (ptr[1] == CHAR_BACKSLASH)
       if (c == CHAR_BACKSLASH)  
1236          {          {
1237          if (ptr[1] == CHAR_E)          if (ptr[2] == CHAR_E)
1238            ptr++;            ptr+= 2;
1239          else if (strncmp((const char *)ptr+1,          else if (strncmp((const char *)ptr+2,
1240                   STR_Q STR_BACKSLASH STR_E, 3) == 0)                   STR_Q STR_BACKSLASH STR_E, 3) == 0)
1241            ptr += 3;            ptr += 4;
1242          else          else
1243            break;            break;
1244          }          }
1245        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)        else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1246            {
1247          negate_class = TRUE;          negate_class = TRUE;
1248            ptr++;
1249            }
1250        else break;        else break;
1251        }        }
1252    
# Line 1093  for (; *ptr != 0; ptr++) Line 1262  for (; *ptr != 0; ptr++)
1262        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1263        if (*ptr == CHAR_BACKSLASH)        if (*ptr == CHAR_BACKSLASH)
1264          {          {
1265          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1266          if (*ptr == CHAR_Q) for (;;)          if (*ptr == CHAR_Q) for (;;)
1267            {            {
1268            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1269            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1270            if (*(++ptr) == CHAR_E) break;            if (*(++ptr) == CHAR_E) break;
1271            }            }
1272          continue;          continue;
# Line 1110  for (; *ptr != 0; ptr++) Line 1279  for (; *ptr != 0; ptr++)
1279    
1280    if (xmode && *ptr == CHAR_NUMBER_SIGN)    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1281      {      {
1282      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};      ptr++;
1283      if (*ptr == 0) return -1;      while (*ptr != 0)
1284          {
1285          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1286          ptr++;
1287    #ifdef SUPPORT_UTF8
1288          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1289    #endif
1290          }
1291        if (*ptr == 0) goto FAIL_EXIT;
1292      continue;      continue;
1293      }      }
1294    
1295    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1296    
1297    if (*ptr != CHAR_LEFT_PARENTHESIS) continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)  
1298      {      {
1299      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1300      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1301      continue;      if (*ptr == 0) goto FAIL_EXIT;
1302      }      }
1303    
1304    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1305    if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */      {
1306        if (dup_parens && *count < hwm_count) *count = hwm_count;
1307        goto FAIL_EXIT;
1308        }
1309    
1310    /* We have to disambiguate (?<! and (?<= from (?<name> */    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1311        {
1312        if (*count > hwm_count) hwm_count = *count;
1313        *count = start_count;
1314        }
1315      }
1316    
1317    if ((*ptr != CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_EXCLAMATION_MARK ||  FAIL_EXIT:
1318        ptr[1] == CHAR_EQUALS_SIGN) && *ptr != CHAR_APOSTROPHE)  *ptrptr = ptr;
1319      continue;  return -1;
1320    }
1321    
   count++;  
1322    
1323    if (name == NULL && count == lorn) return count;  
1324    term = *ptr++;  
1325    if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;  /*************************************************
1326    thisname = ptr;  *       Find forward referenced subpattern       *
1327    while (*ptr != term) ptr++;  *************************************************/
1328    if (name != NULL && lorn == ptr - thisname &&  
1329        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  /* This function scans along a pattern's text looking for capturing
1330      return count;  subpatterns, and counting them. If it finds a named pattern that matches the
1331    name it is given, it returns its number. Alternatively, if the name is NULL, it
1332    returns when it reaches a given numbered subpattern. This is used for forward
1333    references to subpatterns. We used to be able to start this scan from the
1334    current compiling point, using the current count value from cd->bracount, and
1335    do it all in a single loop, but the addition of the possibility of duplicate
1336    subpattern numbers means that we have to scan from the very start, in order to
1337    take account of such duplicates, and to use a recursive function to keep track
1338    of the different types of group.
1339    
1340    Arguments:
1341      cd           compile background data
1342      name         name to seek, or NULL if seeking a numbered subpattern
1343      lorn         name length, or subpattern number if name is NULL
1344      xmode        TRUE if we are in /x mode
1345      utf8         TRUE if we are in UTF-8 mode
1346    
1347    Returns:       the number of the found subpattern, or -1 if not found
1348    */
1349    
1350    static int
1351    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1352      BOOL utf8)
1353    {
1354    uschar *ptr = (uschar *)cd->start_pattern;
1355    int count = 0;
1356    int rc;
1357    
1358    /* If the pattern does not start with an opening parenthesis, the first call
1359    to find_parens_sub() will scan right to the end (if necessary). However, if it
1360    does start with a parenthesis, find_parens_sub() will return when it hits the
1361    matching closing parens. That is why we have to have a loop. */
1362    
1363    for (;;)
1364      {
1365      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1366      if (rc > 0 || *ptr++ == 0) break;
1367    }    }
1368    
1369  return -1;  return rc;
1370  }  }
1371    
1372    
1373    
1374    
1375  /*************************************************  /*************************************************
1376  *      Find first significant op code            *  *      Find first significant op code            *
1377  *************************************************/  *************************************************/
# Line 1200  for (;;) Line 1421  for (;;)
1421    
1422      case OP_CALLOUT:      case OP_CALLOUT:
1423      case OP_CREF:      case OP_CREF:
1424        case OP_NCREF:
1425      case OP_RREF:      case OP_RREF:
1426        case OP_NRREF:
1427      case OP_DEF:      case OP_DEF:
1428      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1429      break;      break;
# Line 1216  for (;;) Line 1439  for (;;)
1439    
1440    
1441  /*************************************************  /*************************************************
1442  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1443  *************************************************/  *************************************************/
1444    
1445  /* 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,
1446  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.
1447  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
1448    temporarily terminated with OP_END when this function is called.
1449    
1450    This function is called when a backward assertion is encountered, so that if it
1451    fails, the error message can point to the correct place in the pattern.
1452    However, we cannot do this when the assertion contains subroutine calls,
1453    because they can be forward references. We solve this by remembering this case
1454    and doing the check at the end; a flag specifies which mode we are running in.
1455    
1456  Arguments:  Arguments:
1457    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1458    options  the compiling options    options  the compiling options
1459      atend    TRUE if called when the pattern is complete
1460      cd       the "compile data" structure
1461    
1462  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1463                 or -1 if there is no fixed length,
1464               or -2 if \C was encountered               or -2 if \C was encountered
1465                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1466  */  */
1467    
1468  static int  static int
1469  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1470  {  {
1471  int length = -1;  int length = -1;
1472    
# Line 1245  branch, check the length against that of Line 1479  branch, check the length against that of
1479  for (;;)  for (;;)
1480    {    {
1481    int d;    int d;
1482      uschar *ce, *cs;
1483    register int op = *cc;    register int op = *cc;
1484    switch (op)    switch (op)
1485      {      {
# Line 1252  for (;;) Line 1487  for (;;)
1487      case OP_BRA:      case OP_BRA:
1488      case OP_ONCE:      case OP_ONCE:
1489      case OP_COND:      case OP_COND:
1490      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1491      if (d < 0) return d;      if (d < 0) return d;
1492      branchlength += d;      branchlength += d;
1493      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1275  for (;;) Line 1510  for (;;)
1510      branchlength = 0;      branchlength = 0;
1511      break;      break;
1512    
1513        /* A true recursion implies not fixed length, but a subroutine call may
1514        be OK. If the subroutine is a forward reference, we can't deal with
1515        it until the end of the pattern, so return -3. */
1516    
1517        case OP_RECURSE:
1518        if (!atend) return -3;
1519        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1520        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1521        if (cc > cs && cc < ce) return -1;                /* Recursion */
1522        d = find_fixedlength(cs + 2, options, atend, cd);
1523        if (d < 0) return d;
1524        branchlength += d;
1525        cc += 1 + LINK_SIZE;
1526        break;
1527    
1528      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1529    
1530      case OP_ASSERT:      case OP_ASSERT:
# Line 1288  for (;;) Line 1538  for (;;)
1538    
1539      case OP_REVERSE:      case OP_REVERSE:
1540      case OP_CREF:      case OP_CREF:
1541        case OP_NCREF:
1542      case OP_RREF:      case OP_RREF:
1543        case OP_NRREF:
1544      case OP_DEF:      case OP_DEF:
1545      case OP_OPT:      case OP_OPT:
1546      case OP_CALLOUT:      case OP_CALLOUT:
1547      case OP_SOD:      case OP_SOD:
1548      case OP_SOM:      case OP_SOM:
1549        case OP_SET_SOM:
1550      case OP_EOD:      case OP_EOD:
1551      case OP_EODN:      case OP_EODN:
1552      case OP_CIRC:      case OP_CIRC:
# Line 1311  for (;;) Line 1564  for (;;)
1564      branchlength++;      branchlength++;
1565      cc += 2;      cc += 2;
1566  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1567      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1568        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1569  #endif  #endif
1570      break;      break;
1571    
# Line 1325  for (;;) Line 1576  for (;;)
1576      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1577      cc += 4;      cc += 4;
1578  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1579      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1580        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1581  #endif  #endif
1582      break;      break;
1583    
# Line 1407  for (;;) Line 1656  for (;;)
1656    
1657    
1658  /*************************************************  /*************************************************
1659  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1660  *************************************************/  *************************************************/
1661    
1662  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1663  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1664    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1665    so that it can be called from pcre_study() when finding the minimum matching
1666    length.
1667    
1668  Arguments:  Arguments:
1669    code        points to start of expression    code        points to start of expression
1670    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1671    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1672    
1673  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
1674  */  */
1675    
1676  static const uschar *  const uschar *
1677  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1678  {  {
1679  for (;;)  for (;;)
1680    {    {
# Line 1435  for (;;) Line 1687  for (;;)
1687    
1688    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1689    
1690      /* Handle recursion */
1691    
1692      else if (c == OP_REVERSE)
1693        {
1694        if (number < 0) return (uschar *)code;
1695        code += _pcre_OP_lengths[c];
1696        }
1697    
1698    /* Handle capturing bracket */    /* Handle capturing bracket */
1699    
1700    else if (c == OP_CBRA)    else if (c == OP_CBRA)
# Line 1446  for (;;) Line 1706  for (;;)
1706    
1707    /* 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
1708    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
1709    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1710      must add in its length. */
1711    
1712    else    else
1713      {      {
# Line 1470  for (;;) Line 1731  for (;;)
1731        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1732        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1733        break;        break;
1734    
1735          case OP_MARK:
1736          case OP_PRUNE_ARG:
1737          case OP_SKIP_ARG:
1738          code += code[1];
1739          break;
1740    
1741          case OP_THEN_ARG:
1742          code += code[1+LINK_SIZE];
1743          break;
1744        }        }
1745    
1746      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1541  for (;;) Line 1812  for (;;)
1812    
1813    /* 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
1814    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
1815    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1816      must add in its length. */
1817    
1818    else    else
1819      {      {
# Line 1565  for (;;) Line 1837  for (;;)
1837        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1838        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1839        break;        break;
1840    
1841          case OP_MARK:
1842          case OP_PRUNE_ARG:
1843          case OP_SKIP_ARG:
1844          code += code[1];
1845          break;
1846    
1847          case OP_THEN_ARG:
1848          code += code[1+LINK_SIZE];
1849          break;
1850        }        }
1851    
1852      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1621  Arguments: Line 1903  Arguments:
1903    code        points to start of search    code        points to start of search
1904    endcode     points to where to stop    endcode     points to where to stop
1905    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1906      cd          contains pointers to tables etc.
1907    
1908  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1909  */  */
1910    
1911  static BOOL  static BOOL
1912  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1913      compile_data *cd)
1914  {  {
1915  register int c;  register int c;
1916  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 1941  for (code = first_significant_code(code
1941      continue;      continue;
1942      }      }
1943    
1944      /* For a recursion/subroutine call, if its end has been reached, which
1945      implies a subroutine call, we can scan it. */
1946    
1947      if (c == OP_RECURSE)
1948        {
1949        BOOL empty_branch = FALSE;
1950        const uschar *scode = cd->start_code + GET(code, 1);
1951        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1952        do
1953          {
1954          if (could_be_empty_branch(scode, endcode, utf8, cd))
1955            {
1956            empty_branch = TRUE;
1957            break;
1958            }
1959          scode += GET(scode, 1);
1960          }
1961        while (*scode == OP_ALT);
1962        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1963        continue;
1964        }
1965    
1966    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
1967    
1968    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 1675  for (code = first_significant_code(code Line 1981  for (code = first_significant_code(code
1981        empty_branch = FALSE;        empty_branch = FALSE;
1982        do        do
1983          {          {
1984          if (!empty_branch && could_be_empty_branch(code, endcode, utf8))          if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1985            empty_branch = TRUE;            empty_branch = TRUE;
1986          code += GET(code, 1);          code += GET(code, 1);
1987          }          }
# Line 1800  for (code = first_significant_code(code Line 2106  for (code = first_significant_code(code
2106      case OP_QUERY:      case OP_QUERY:
2107      case OP_MINQUERY:      case OP_MINQUERY:
2108      case OP_POSQUERY:      case OP_POSQUERY:
2109        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2110        break;
2111    
2112      case OP_UPTO:      case OP_UPTO:
2113      case OP_MINUPTO:      case OP_MINUPTO:
2114      case OP_POSUPTO:      case OP_POSUPTO:
2115      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2116      break;      break;
2117  #endif  #endif
2118    
2119        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2120        string. */
2121    
2122        case OP_MARK:
2123        case OP_PRUNE_ARG:
2124        case OP_SKIP_ARG:
2125        code += code[1];
2126        break;
2127    
2128        case OP_THEN_ARG:
2129        code += code[1+LINK_SIZE];
2130        break;
2131    
2132        /* None of the remaining opcodes are required to match a character. */
2133    
2134        default:
2135        break;
2136      }      }
2137    }    }
2138    
# Line 1828  Arguments: Line 2155  Arguments:
2155    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2156    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2157    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2158      cd          pointers to tables etc
2159    
2160  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2161  */  */
2162    
2163  static BOOL  static BOOL
2164  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2165    BOOL utf8)    BOOL utf8, compile_data *cd)
2166  {  {
2167  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2168    {    {
2169    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2170        return FALSE;
2171    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2172    }    }
2173  return TRUE;  return TRUE;
# Line 2020  auto_callout(uschar *code, const uschar Line 2349  auto_callout(uschar *code, const uschar
2349  {  {
2350  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2351  *code++ = 255;  *code++ = 255;
2352  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2353  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2354  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2355  }  }
2356    
# Line 2046  Returns:             nothing Line 2375  Returns:             nothing
2375  static void  static void
2376  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2377  {  {
2378  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2379  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2380  }  }
2381    
# Line 2096  for (++c; c <= d; c++) Line 2425  for (++c; c <= d; c++)
2425    
2426  return TRUE;  return TRUE;
2427  }  }
2428    
2429    
2430    
2431    /*************************************************
2432    *        Check a character and a property        *
2433    *************************************************/
2434    
2435    /* This function is called by check_auto_possessive() when a property item
2436    is adjacent to a fixed character.
2437    
2438    Arguments:
2439      c            the character
2440      ptype        the property type
2441      pdata        the data for the type
2442      negated      TRUE if it's a negated property (\P or \p{^)
2443    
2444    Returns:       TRUE if auto-possessifying is OK
2445    */
2446    
2447    static BOOL
2448    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2449    {
2450    const ucd_record *prop = GET_UCD(c);
2451    switch(ptype)
2452      {
2453      case PT_LAMP:
2454      return (prop->chartype == ucp_Lu ||
2455              prop->chartype == ucp_Ll ||
2456              prop->chartype == ucp_Lt) == negated;
2457    
2458      case PT_GC:
2459      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2460    
2461      case PT_PC:
2462      return (pdata == prop->chartype) == negated;
2463    
2464      case PT_SC:
2465      return (pdata == prop->script) == negated;
2466    
2467      /* These are specials */
2468    
2469      case PT_ALNUM:
2470      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2471              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2472    
2473      case PT_SPACE:    /* Perl space */
2474      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2475              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2476              == negated;
2477    
2478      case PT_PXSPACE:  /* POSIX space */
2479      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2480              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2481              c == CHAR_FF || c == CHAR_CR)
2482              == negated;
2483    
2484      case PT_WORD:
2485      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2486              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2487              c == CHAR_UNDERSCORE) == negated;
2488      }
2489    return FALSE;
2490    }
2491  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2492    
2493    
# Line 2109  whether the next thing could possibly ma Line 2501  whether the next thing could possibly ma
2501  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2502    
2503  Arguments:  Arguments:
2504    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2505    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2506    ptr           next character in pattern    ptr           next character in pattern
2507    options       options bits    options       options bits
2508    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2121  Returns:        TRUE if possessifying is Line 2511  Returns:        TRUE if possessifying is
2511  */  */
2512    
2513  static BOOL  static BOOL
2514  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2515    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2516  {  {
2517  int next;  int c, next;
2518    int op_code = *previous++;
2519    
2520  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2521    
# Line 2135  if ((options & PCRE_EXTENDED) != 0) Line 2526  if ((options & PCRE_EXTENDED) != 0)
2526      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2527      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2528        {        {
2529        while (*(++ptr) != 0)        ptr++;
2530          while (*ptr != 0)
2531            {
2532          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2533            ptr++;
2534    #ifdef SUPPORT_UTF8
2535            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2536    #endif
2537            }
2538        }        }
2539      else break;      else break;
2540      }      }
# Line 2172  if ((options & PCRE_EXTENDED) != 0) Line 2570  if ((options & PCRE_EXTENDED) != 0)
2570      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2571      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2572        {        {
2573        while (*(++ptr) != 0)        ptr++;
2574          while (*ptr != 0)
2575            {
2576          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2577            ptr++;
2578    #ifdef SUPPORT_UTF8
2579            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2580    #endif
2581            }
2582        }        }
2583      else break;      else break;
2584      }      }
# Line 2185  if (*ptr == CHAR_ASTERISK || *ptr == CHA Line 2590  if (*ptr == CHAR_ASTERISK || *ptr == CHA
2590    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)
2591      return FALSE;      return FALSE;
2592    
2593  /* 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
2594  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. */  
2595    
2596  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2597    {    {
2598    case OP_CHAR:    case OP_CHAR:
2599  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2600    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2601  #else  #else
2602    (void)(utf8_char);  /* Keep compiler happy by referencing function argument */    c = *previous;
2603  #endif  #endif
2604    return item != next;    return c != next;
2605    
2606    /* 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
2607    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 2209  if (next >= 0) switch(op_code) Line 2609  if (next >= 0) switch(op_code)
2609    
2610    case OP_CHARNC:    case OP_CHARNC:
2611  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2612    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2613    #else
2614      c = *previous;
2615  #endif  #endif
2616    if (item == next) return FALSE;    if (c == next) return FALSE;
2617  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2618    if (utf8)    if (utf8)
2619      {      {
# Line 2222  if (next >= 0) switch(op_code) Line 2624  if (next >= 0) switch(op_code)
2624  #else  #else
2625      othercase = NOTACHAR;      othercase = NOTACHAR;
2626  #endif  #endif
2627      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2628      }      }
2629    else    else
2630  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2631    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2632    
2633    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, its data is always a single-byte character. */
2634    
2635    case OP_NOT:    case OP_NOT:
2636    if (item == next) return TRUE;    if ((c = *previous) == next) return TRUE;
2637    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2638  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2639    if (utf8)    if (utf8)
# Line 2243  if (next >= 0) switch(op_code) Line 2645  if (next >= 0) switch(op_code)
2645  #else  #else
2646      othercase = NOTACHAR;      othercase = NOTACHAR;
2647  #endif  #endif
2648      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2649      }      }
2650    else    else
2651  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2652    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2653    
2654      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2655      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2656    
2657    case OP_DIGIT:    case OP_DIGIT:
2658    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2290  if (next >= 0) switch(op_code) Line 2695  if (next >= 0) switch(op_code)
2695      case 0x202f:      case 0x202f:
2696      case 0x205f:      case 0x205f:
2697      case 0x3000:      case 0x3000:
2698      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2699      default:      default:
2700      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2701      }      }
2702    
2703      case OP_ANYNL:
2704    case OP_VSPACE:    case OP_VSPACE:
2705    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2706    switch(next)    switch(next)
# Line 2306  if (next >= 0) switch(op_code) Line 2712  if (next >= 0) switch(op_code)
2712      case 0x85:      case 0x85:
2713      case 0x2028:      case 0x2028:
2714      case 0x2029:      case 0x2029:
2715      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2716      default:      default:
2717      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2718      }      }
2719    
2720    #ifdef SUPPORT_UCP
2721      case OP_PROP:
2722      return check_char_prop(next, previous[0], previous[1], FALSE);
2723    
2724      case OP_NOTPROP:
2725      return check_char_prop(next, previous[0], previous[1], TRUE);
2726    #endif
2727    
2728    default:    default:
2729    return FALSE;    return FALSE;
2730    }    }
2731    
2732    
2733  /* 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
2734    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2735    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2736    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2737    replaced by OP_PROP codes when PCRE_UCP is set. */
2738    
2739  switch(op_code)  switch(op_code)
2740    {    {
2741    case OP_CHAR:    case OP_CHAR:
2742    case OP_CHARNC:    case OP_CHARNC:
2743  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2744    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2745    #else
2746      c = *previous;
2747  #endif  #endif
2748    switch(-next)    switch(-next)
2749      {      {
2750      case ESC_d:      case ESC_d:
2751      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2752    
2753      case ESC_D:      case ESC_D:
2754      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2755    
2756      case ESC_s:      case ESC_s:
2757      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2758    
2759      case ESC_S:      case ESC_S:
2760      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2761    
2762      case ESC_w:      case ESC_w:
2763      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2764    
2765      case ESC_W:      case ESC_W:
2766      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2767    
2768      case ESC_h:      case ESC_h:
2769      case ESC_H:      case ESC_H:
2770      switch(item)      switch(c)
2771        {        {
2772        case 0x09:        case 0x09:
2773        case 0x20:        case 0x20:
# Line 2375  switch(op_code) Line 2795  switch(op_code)
2795    
2796      case ESC_v:      case ESC_v:
2797      case ESC_V:      case ESC_V:
2798      switch(item)      switch(c)
2799        {        {
2800        case 0x0a:        case 0x0a:
2801        case 0x0b:        case 0x0b:
# Line 2389  switch(op_code) Line 2809  switch(op_code)
2809        return -next == ESC_v;        return -next == ESC_v;
2810        }        }
2811    
2812        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2813        their substitutions and process them. The result will always be either
2814        -ESC_p or -ESC_P. Then fall through to process those values. */
2815    
2816    #ifdef SUPPORT_UCP
2817        case ESC_du:
2818        case ESC_DU:
2819        case ESC_wu:
2820        case ESC_WU:
2821        case ESC_su:
2822        case ESC_SU:
2823          {
2824          int temperrorcode = 0;
2825          ptr = substitutes[-next - ESC_DU];
2826          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2827          if (temperrorcode != 0) return FALSE;
2828          ptr++;    /* For compatibility */
2829          }
2830        /* Fall through */
2831    
2832        case ESC_p:
2833        case ESC_P:
2834          {
2835          int ptype, pdata, errorcodeptr;
2836          BOOL negated;
2837    
2838          ptr--;      /* Make ptr point at the p or P */
2839          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2840          if (ptype < 0) return FALSE;
2841          ptr++;      /* Point past the final curly ket */
2842    
2843          /* If the property item is optional, we have to give up. (When generated
2844          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2845          to the original \d etc. At this point, ptr will point to a zero byte. */
2846    
2847          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2848            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2849              return FALSE;
2850    
2851          /* Do the property check. */
2852    
2853          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2854          }
2855    #endif
2856    
2857      default:      default:
2858      return FALSE;      return FALSE;
2859      }      }
2860    
2861      /* In principle, support for Unicode properties should be integrated here as
2862      well. It means re-organizing the above code so as to get hold of the property
2863      values before switching on the op-code. However, I wonder how many patterns
2864      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2865      these op-codes are never generated.) */
2866    
2867    case OP_DIGIT:    case OP_DIGIT:
2868    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2869           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
2870    
2871    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2872    return next == -ESC_d;    return next == -ESC_d;
2873    
2874    case OP_WHITESPACE:    case OP_WHITESPACE:
2875    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2876    
2877    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2878    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2879    
2880    case OP_HSPACE:    case OP_HSPACE:
2881    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2882             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2883    
2884    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
2885    return next == -ESC_h;    return next == -ESC_h;
2886    
2887    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
2888      case OP_ANYNL:
2889    case OP_VSPACE:    case OP_VSPACE:
2890    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2891    
2892    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2893    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
2894    
2895    case OP_WORDCHAR:    case OP_WORDCHAR:
2896    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2897             next == -ESC_v || next == -ESC_R;
2898    
2899    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2900    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2484  BOOL inescq = FALSE; Line 2958  BOOL inescq = FALSE;
2958  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
2959  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
2960  const uschar *tempptr;  const uschar *tempptr;
2961    const uschar *nestptr = NULL;
2962  uschar *previous = NULL;  uschar *previous = NULL;
2963  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2964  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
# Line 2500  BOOL utf8 = FALSE; Line 2975  BOOL utf8 = FALSE;
2975  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2976  #endif  #endif
2977    
2978  #ifdef DEBUG  #ifdef PCRE_DEBUG
2979  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2980  #endif  #endif
2981    
# Line 2554  for (;; ptr++) Line 3029  for (;; ptr++)
3029    
3030    c = *ptr;    c = *ptr;
3031    
3032      /* If we are at the end of a nested substitution, revert to the outer level
3033      string. Nesting only happens one level deep. */
3034    
3035      if (c == 0 && nestptr != NULL)
3036        {
3037        ptr = nestptr;
3038        nestptr = NULL;
3039        c = *ptr;
3040        }
3041    
3042    /* 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
3043    previous cycle of this loop. */    previous cycle of this loop. */
3044    
3045    if (lengthptr != NULL)    if (lengthptr != NULL)
3046      {      {
3047  #ifdef DEBUG  #ifdef PCRE_DEBUG
3048      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3049  #endif  #endif
3050      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3051        {        {
3052        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3053        goto FAILED;        goto FAILED;
# Line 2584  for (;; ptr++) Line 3069  for (;; ptr++)
3069        goto FAILED;        goto FAILED;
3070        }        }
3071    
3072      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3073      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));
3074    
3075      /* 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 2611  for (;; ptr++) Line 3096  for (;; ptr++)
3096    /* 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
3097    reference list. */    reference list. */
3098    
3099    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3100      {      {
3101      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3102      goto FAILED;      goto FAILED;
# Line 2666  for (;; ptr++) Line 3151  for (;; ptr++)
3151      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3152      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
3153        {        {
3154        while (*(++ptr) != 0)        ptr++;
3155          while (*ptr != 0)
3156          {          {
3157          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3158            ptr++;
3159    #ifdef SUPPORT_UTF8
3160            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3161    #endif
3162          }          }
3163        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3164    
# Line 2702  for (;; ptr++) Line 3192  for (;; ptr++)
3192          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3193          goto FAILED;          goto FAILED;
3194          }          }
3195        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3196        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3197        }        }
3198      return TRUE;      return TRUE;
# Line 2907  for (;; ptr++) Line 3397  for (;; ptr++)
3397            ptr++;            ptr++;
3398            }            }
3399    
3400          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3401          if (posix_class < 0)          if (posix_class < 0)
3402            {            {
3403            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2921  for (;; ptr++) Line 3411  for (;; ptr++)
3411          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3412            posix_class = 0;            posix_class = 0;
3413    
3414          /* 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
3415          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3416          subtract bits that may be in the main map already. At the end we or the  
3417          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3418            if ((options & PCRE_UCP) != 0)
3419              {
3420              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3421              if (posix_substitutes[pc] != NULL)
3422                {
3423                nestptr = tempptr + 1;
3424                ptr = posix_substitutes[pc] - 1;
3425                continue;
3426                }
3427              }
3428    #endif
3429            /* In the non-UCP case, we build the bit map for the POSIX class in a
3430            chunk of local store because we may be adding and subtracting from it,
3431            and we don't want to subtract bits that may be in the main map already.
3432            At the end we or the result into the bit map that is being built. */
3433    
3434          posix_class *= 3;          posix_class *= 3;
3435    
# Line 2968  for (;; ptr++) Line 3473  for (;; ptr++)
3473    
3474        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3475        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
3476        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
3477        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
3478        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
3479        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3480          PCRE_EXTRA is set. */
3481    
3482        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
3483          {          {
3484          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3485          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3486    
3487          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 */  
3488          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3489            {            {
3490            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
# Line 2997  for (;; ptr++) Line 3501  for (;; ptr++)
3501            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3502            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3503    
3504            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3505              {              {
3506    #ifdef SUPPORT_UCP
3507                case ESC_du:     /* These are the values given for \d etc */
3508                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3509                case ESC_wu:     /* escape sequence with an appropriate \p */
3510                case ESC_WU:     /* or \P to test Unicode properties instead */
3511                case ESC_su:     /* of the default ASCII testing. */
3512                case ESC_SU:
3513                nestptr = ptr;
3514                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3515                class_charcount -= 2;                /* Undo! */
3516                continue;
3517    #endif
3518              case ESC_d:              case ESC_d:
3519              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3520              continue;              continue;
# Line 3019  for (;; ptr++) Line 3533  for (;; ptr++)
3533              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3534              continue;              continue;
3535    
3536                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3537                if it was previously set by something earlier in the character
3538                class. */
3539    
3540              case ESC_s:              case ESC_s:
3541              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3542              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3543                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3544              continue;              continue;
3545    
3546              case ESC_S:              case ESC_S:
# Line 3030  for (;; ptr++) Line 3549  for (;; ptr++)
3549              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3550              continue;              continue;
3551    
3552              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)  
             {  
3553              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3554              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3555              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 3067  for (;; ptr++) Line 3573  for (;; ptr++)
3573                }                }
3574  #endif  #endif
3575              continue;              continue;
             }  
3576    
3577            if (-c == ESC_H)              case ESC_H:
             {  
3578              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3579                {                {
3580                int x = 0xff;                int x = 0xff;
# Line 3112  for (;; ptr++) Line 3616  for (;; ptr++)
3616                }                }
3617  #endif  #endif
3618              continue;              continue;
             }  
3619    
3620            if (-c == ESC_v)              case ESC_v:
             {  
3621              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3622              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3623              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 3131  for (;; ptr++) Line 3633  for (;; ptr++)
3633                }                }
3634  #endif  #endif
3635              continue;              continue;
             }  
3636    
3637            if (-c == ESC_V)              case ESC_V:
             {  
3638              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3639                {                {
3640                int x = 0xff;                int x = 0xff;
# Line 3164  for (;; ptr++) Line 3664  for (;; ptr++)
3664                }                }
3665  #endif  #endif
3666              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3667    
3668  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3669            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3670              {              case ESC_P:
3671              BOOL negated;                {
3672              int pdata;                BOOL negated;
3673              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3674              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3675              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3676              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3677                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3678              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3679              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3680              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3681              continue;                class_charcount -= 2;   /* Not a < 256 character */
3682              }                continue;
3683                  }
3684  #endif  #endif
3685            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3686            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3687            treated as literals. */              treated as literals. */
3688    
3689            if ((options & PCRE_EXTRA) != 0)              default:
3690              {              if ((options & PCRE_EXTRA) != 0)
3691              *errorcodeptr = ERR7;                {
3692              goto FAILED;                *errorcodeptr = ERR7;
3693                  goto FAILED;
3694                  }
3695                class_charcount -= 2;  /* Undo the default count from above */
3696                c = *ptr;              /* Get the final character and fall through */
3697                break;
3698              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3699            }            }
3700    
3701          /* 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 3265  for (;; ptr++) Line 3765  for (;; ptr++)
3765            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3766            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3767    
3768            /* \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 */  
3769    
3770            if (d < 0)            if (d < 0)
3771              {              {
3772              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  
3773                {                {
3774                ptr = oldptr;                ptr = oldptr;
3775                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3438  for (;; ptr++) Line 3935  for (;; ptr++)
3935          }          }
3936        }        }
3937    
3938      /* 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.
3939        If we are at the end of an internal nested string, revert to the outer
3940        string. */
3941    
3942        while (((c = *(++ptr)) != 0 ||
3943               (nestptr != NULL &&
3944                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
3945               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3946    
3947      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));      /* Check for missing terminating ']' */
3948    
3949      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
3950        {        {
3951        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
3952        goto FAILED;        goto FAILED;
3953        }        }
3954    
   
 /* 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  
   
   
3955      /* 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
3956      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
3957      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 3530  we set the flag only if there is a liter Line 4015  we set the flag only if there is a liter
4015    
4016      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4017      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
4018      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
4019      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
4020      (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
4021      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
4022        actual compiled code. */
4023    
4024  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4025      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4026        {        {
4027        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4028        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3562  we set the flag only if there is a liter Line 4048  we set the flag only if there is a liter
4048        }        }
4049  #endif  #endif
4050    
4051      /* 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
4052      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
4053      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
4054      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4055        negating it if necessary. */
4056    
4057      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4058      if (negate_class)      if (negate_class)
# Line 3689  we set the flag only if there is a liter Line 4176  we set the flag only if there is a liter
4176    
4177        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4178            repeat_max < 0 &&            repeat_max < 0 &&
4179            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4180          {          {
4181          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4182          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3711  we set the flag only if there is a liter Line 4197  we set the flag only if there is a liter
4197        c = previous[1];        c = previous[1];
4198        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4199            repeat_max < 0 &&            repeat_max < 0 &&
4200            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4201          {          {
4202          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4203          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3735  we set the flag only if there is a liter Line 4221  we set the flag only if there is a liter
4221    
4222        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4223            repeat_max < 0 &&            repeat_max < 0 &&
4224            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4225          {          {
4226          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4227          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3757  we set the flag only if there is a liter Line 4243  we set the flag only if there is a liter
4243    
4244        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4245    
4246          /*--------------------------------------------------------------------*/
4247          /* This code is obsolete from release 8.00; the restriction was finally
4248          removed: */
4249    
4250        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4251        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4252    
4253        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4254          /*--------------------------------------------------------------------*/
4255    
4256        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4257    
# Line 3907  we set the flag only if there is a liter Line 4398  we set the flag only if there is a liter
4398          goto END_REPEAT;          goto END_REPEAT;
4399          }          }
4400    
4401          /*--------------------------------------------------------------------*/
4402          /* This code is obsolete from release 8.00; the restriction was finally
4403          removed: */
4404    
4405        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4406        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4407    
4408        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4409          /*--------------------------------------------------------------------*/
4410    
4411        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4412          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3935  we set the flag only if there is a liter Line 4431  we set the flag only if there is a liter
4431        {        {
4432        register int i;        register int i;
4433        int ketoffset = 0;        int ketoffset = 0;
4434        int len = code - previous;        int len = (int)(code - previous);
4435        uschar *bralink = NULL;        uschar *bralink = NULL;
4436    
4437        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless */
# Line 3956  we set the flag only if there is a liter Line 4452  we set the flag only if there is a liter
4452          {          {
4453          register uschar *ket = previous;          register uschar *ket = previous;
4454          do ket += GET(ket, 1); while (*ket != OP_KET);          do ket += GET(ket, 1); while (*ket != OP_KET);
4455          ketoffset = code - ket;          ketoffset = (int)(code - ket);
4456          }          }
4457    
4458        /* 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 4024  we set the flag only if there is a liter Line 4520  we set the flag only if there is a liter
4520            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4521            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4522    
4523            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4524            bralink = previous;            bralink = previous;
4525            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4526            }            }
# Line 4045  we set the flag only if there is a liter Line 4541  we set the flag only if there is a liter
4541            {            {
4542            /* 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
4543            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
4544            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4545              integer type when available, otherwise double. */
4546    
4547            if (lengthptr != NULL)            if (lengthptr != NULL)
4548              {              {
4549              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4550              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4551                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4552                        (INT64_OR_DOUBLE)INT_MAX ||
4553                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4554                {                {
4555                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 4097  we set the flag only if there is a liter Line 4595  we set the flag only if there is a liter
4595          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
4596          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
4597          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
4598          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4599            a 64-bit integer type when available, otherwise double. */
4600    
4601          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4602            {            {
4603            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4604                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4605            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4606                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4607                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4608                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4609              {              {
4610              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 4130  we set the flag only if there is a liter Line 4629  we set the flag only if there is a liter
4629              {              {
4630              int offset;              int offset;
4631              *code++ = OP_BRA;              *code++ = OP_BRA;
4632              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4633              bralink = code;              bralink = code;
4634              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4635              }              }
# Line 4151  we set the flag only if there is a liter Line 4650  we set the flag only if there is a liter
4650          while (bralink != NULL)          while (bralink != NULL)
4651            {            {
4652            int oldlinkoffset;            int oldlinkoffset;
4653            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4654            uschar *bra = code - offset;            uschar *bra = code - offset;
4655            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4656            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 4182  we set the flag only if there is a liter Line 4681  we set the flag only if there is a liter
4681            uschar *scode = bracode;            uschar *scode = bracode;
4682            do            do
4683              {              {
4684              if (could_be_empty_branch(scode, ketcode, utf8))              if (could_be_empty_branch(scode, ketcode, utf8, cd))
4685                {                {
4686                *bracode += OP_SBRA - OP_BRA;                *bracode += OP_SBRA - OP_BRA;
4687                break;                break;
# Line 4225  we set the flag only if there is a liter Line 4724  we set the flag only if there is a liter
4724      if (possessive_quantifier)      if (possessive_quantifier)
4725        {        {
4726        int len;        int len;
4727        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4728            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4729          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += _pcre_OP_lengths[*tempcode] +
4730            ((*tempcode == OP_TYPEEXACT &&            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4731               (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);  
4732        len = code - tempcode;        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4733            {
4734            tempcode += _pcre_OP_lengths[*tempcode];
4735    #ifdef SUPPORT_UTF8
4736            if (utf8 && tempcode[-1] >= 0xc0)
4737              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4738    #endif
4739            }
4740    
4741          len = (int)(code - tempcode);
4742        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4743          {          {
4744          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 4248  we set the flag only if there is a liter Line 4756  we set the flag only if there is a liter
4756          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4757          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4758    
4759            /* Because we are moving code along, we must ensure that any
4760            pending recursive references are updated. */
4761    
4762          default:          default:
4763            *code = OP_END;
4764            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4765          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4766          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4767          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4284  we set the flag only if there is a liter Line 4797  we set the flag only if there is a liter
4797    
4798      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4799    
4800      if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
4801             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
4802        {        {
4803        int i, namelen;        int i, namelen;
4804          int arglen = 0;
4805        const char *vn = verbnames;        const char *vn = verbnames;
4806        const uschar *name = ++ptr;        const uschar *name = ptr + 1;
4807          const uschar *arg = NULL;
4808        previous = NULL;        previous = NULL;
4809        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4810          namelen = (int)(ptr - name);
4811    
4812        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
4813          {          {
4814          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
4815          goto FAILED;          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
4816              || *ptr == '_') ptr++;
4817            arglen = (int)(ptr - arg);
4818          }          }
4819    
4820        if (*ptr != CHAR_RIGHT_PARENTHESIS)        if (*ptr != CHAR_RIGHT_PARENTHESIS)
4821          {          {
4822          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4823          goto FAILED;          goto FAILED;
4824          }          }
4825        namelen = ptr - name;  
4826          /* Scan the table of verb names */
4827    
4828        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
4829          {          {
4830          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4831              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4832            {            {
4833            *code = verbs[i].op;            /* Check for open captures before ACCEPT */
4834            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;  
4835            break;            if (verbs[i].op == OP_ACCEPT)
4836                {
4837                open_capitem *oc;
4838                cd->had_accept = TRUE;
4839                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4840                  {
4841                  *code++ = OP_CLOSE;
4842                  PUT2INC(code, 0, oc->number);
4843                  }
4844                }
4845    
4846              /* Handle the cases with/without an argument */
4847    
4848              if (arglen == 0)
4849                {
4850                if (verbs[i].op < 0)   /* Argument is mandatory */
4851                  {
4852                  *errorcodeptr = ERR66;
4853                  goto FAILED;
4854                  }
4855                *code = verbs[i].op;
4856                if (*code++ == OP_THEN)
4857                  {
4858                  PUT(code, 0, code - bcptr->current_branch - 1);
4859                  code += LINK_SIZE;
4860                  }
4861                }
4862    
4863              else
4864                {
4865                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
4866                  {
4867                  *errorcodeptr = ERR59;
4868                  goto FAILED;
4869                  }
4870                *code = verbs[i].op_arg;
4871                if (*code++ == OP_THEN_ARG)
4872                  {
4873                  PUT(code, 0, code - bcptr->current_branch - 1);
4874                  code += LINK_SIZE;
4875                  }
4876                *code++ = arglen;
4877                memcpy(code, arg, arglen);
4878                code += arglen;
4879                *code++ = 0;
4880                }
4881    
4882              break;  /* Found verb, exit loop */
4883            }            }
4884    
4885          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
4886          }          }
4887        if (i < verbcount) continue;  
4888        *errorcodeptr = ERR60;        if (i < verbcount) continue;    /* Successfully handled a verb */
4889          *errorcodeptr = ERR60;          /* Verb not recognized */
4890        goto FAILED;        goto FAILED;
4891        }        }
4892    
# Line 4433  we set the flag only if there is a liter Line 5005  we set the flag only if there is a liter
5005                recno * 10 + *ptr - CHAR_0 : -1;                recno * 10 + *ptr - CHAR_0 : -1;
5006            ptr++;            ptr++;
5007            }            }
5008          namelen = ptr - name;          namelen = (int)(ptr - name);
5009    
5010          if ((terminator > 0 && *ptr++ != terminator) ||          if ((terminator > 0 && *ptr++ != terminator) ||
5011              *ptr++ != CHAR_RIGHT_PARENTHESIS)              *ptr++ != CHAR_RIGHT_PARENTHESIS)
# Line 4470  we set the flag only if there is a liter Line 5042  we set the flag only if there is a liter
5042            }            }
5043    
5044          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
5045          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
5046            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
5047            except they record that the reference was originally to a name. The
5048            information is used to check duplicate names. */
5049    
5050          slot = cd->name_table;          slot = cd->name_table;
5051          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 4485  we set the flag only if there is a liter Line 5060  we set the flag only if there is a liter
5060            {            {
5061            recno = GET2(slot, 0);            recno = GET2(slot, 0);
5062            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
5063              code[1+LINK_SIZE]++;
5064            }            }
5065    
5066          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5067    
5068          else if ((i = find_parens(ptr, cd, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5069                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0, utf8)) > 0)
5070            {            {
5071            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5072              code[1+LINK_SIZE]++;
5073            }            }
5074    
5075          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 4624  we set the flag only if there is a liter Line 5201  we set the flag only if there is a liter
5201              goto FAILED;              goto FAILED;
5202              }              }
5203            *code++ = n;            *code++ = n;
5204            PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */            PUT(code, 0, (int)(ptr - cd->start_pattern + 1)); /* Pattern offset */
5205            PUT(code, LINK_SIZE, 0);                    /* Default length */            PUT(code, LINK_SIZE, 0);                          /* Default length */
5206            code += 2 * LINK_SIZE;            code += 2 * LINK_SIZE;
5207            }            }
5208          previous = NULL;          previous = NULL;
# Line 4658  we set the flag only if there is a liter Line 5235  we set the flag only if there is a liter
5235            name = ++ptr;            name = ++ptr;
5236    
5237            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5238            namelen = ptr - name;            namelen = (int)(ptr - name);
5239    
5240            /* In the pre-compile phase, just do a syntax check. */            /* In the pre-compile phase, just do a syntax check. */
5241    
# Line 4685  we set the flag only if there is a liter Line 5262  we set the flag only if there is a liter
5262                }                }
5263              }              }
5264    
5265            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
5266              alphabetical order. Duplicate names for different numbers are
5267              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
5268              number are always OK. (An existing number can be re-used if (?|
5269              appears in the pattern.) In either event, a duplicate name results in
5270              a duplicate entry in the table, even if the number is the same. This
5271              is because the number of names, and hence the table size, is computed
5272              in the pre-compile, and it affects various numbers and pointers which
5273              would all have to be modified, and the compiled code moved down, if
5274              duplicates with the same number were omitted from the table. This
5275              doesn't seem worth the hassle. However, *different* names for the
5276              same number are not permitted. */
5277    
5278            else            else
5279              {              {
5280                BOOL dupname = FALSE;
5281              slot = cd->name_table;              slot = cd->name_table;
5282    
5283              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
5284                {                {
5285                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 4697  we set the flag only if there is a liter Line 5287  we set the flag only if there is a liter
5287                  {                  {
5288                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
5289                    {                    {
5290                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
5291                          (options & PCRE_DUPNAMES) == 0)
5292                      {                      {
5293                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
5294                      goto FAILED;                      goto FAILED;
5295                      }                      }
5296                      else dupname = TRUE;
5297                    }                    }
5298                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
5299                  }                  }
5300    
5301                  /* Make space in the table and break the loop for an earlier
5302                  name. For a duplicate or later name, carry on. We do this for
5303                  duplicates so that in the simple case (when ?(| is not used) they
5304                  are in order of their numbers. */
5305    
5306                if (crc < 0)                if (crc < 0)
5307                  {                  {
5308                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
5309                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
5310                  break;                  break;
5311                  }                  }
5312    
5313                  /* Continue the loop for a later or duplicate name */
5314    
5315                slot += cd->name_entry_size;                slot += cd->name_entry_size;
5316                }                }
5317    
5318                /* For non-duplicate names, check for a duplicate number before
5319                adding the new name. */
5320    
5321                if (!dupname)
5322                  {
5323                  uschar *cslot = cd->name_table;
5324                  for (i = 0; i < cd->names_found; i++)
5325                    {
5326                    if (cslot != slot)
5327                      {
5328                      if (GET2(cslot, 0) == cd->bracount + 1)
5329                        {
5330                        *errorcodeptr = ERR65;
5331                        goto FAILED;
5332                        }
5333                      }
5334                    else i--;
5335                    cslot += cd->name_entry_size;
5336                    }
5337                  }
5338    
5339              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
5340              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
5341              slot[2+namelen] = 0;              slot[2+namelen] = 0;
5342              }              }
5343            }            }
5344    
5345          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
5346            encountered. */
5347    
         ptr++;                    /* Move past > or ' */  
5348          cd->names_found++;          cd->names_found++;
5349            ptr++;                    /* Move past > or ' */
5350          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
5351    
5352    
# Line 4742  we set the flag only if there is a liter Line 5365  we set the flag only if there is a liter
5365          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
5366          name = ++ptr;          name = ++ptr;
5367          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5368          namelen = ptr - name;          namelen = (int)(ptr - name);
5369    
5370          /* 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
5371          reference number. */          reference number. */
# Line 4788  we set the flag only if there is a liter Line 5411  we set the flag only if there is a liter
5411              recno = GET2(slot, 0);              recno = GET2(slot, 0);
5412              }              }
5413            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5414                      find_parens(ptr, cd, name, namelen,                      find_parens(cd, name, namelen,
5415                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0, utf8)) <= 0)
5416              {              {
5417              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
5418              goto FAILED;              goto FAILED;
# Line 4892  we set the flag only if there is a liter Line 5515  we set the flag only if there is a liter
5515            if (lengthptr == NULL)            if (lengthptr == NULL)
5516              {              {
5517              *code = OP_END;              *code = OP_END;
5518              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);              if (recno != 0)
5519                  called = _pcre_find_bracket(cd->start_code, utf8, recno);
5520    
5521              /* Forward reference */              /* Forward reference */
5522    
5523              if (called == NULL)              if (called == NULL)
5524                {                {
5525                if (find_parens(ptr, cd, NULL, recno,                if (find_parens(cd, NULL, recno,
5526                      (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0, utf8) < 0)
5527                  {                  {
5528                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5529                  goto FAILED;                  goto FAILED;
5530                  }                  }
5531    
5532                  /* Fudge the value of "called" so that when it is inserted as an
5533                  offset below, what it actually inserted is the reference number
5534                  of the group. */
5535    
5536                called = cd->start_code + recno;                called = cd->start_code + recno;
5537                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, (int)(code + 2 + LINK_SIZE - cd->start_code));
5538                }                }
5539    
5540              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
# Line 4913  we set the flag only if there is a liter Line 5542  we set the flag only if there is a liter
5542              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. */
5543    
5544              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 &&
5545                       could_be_empty(called, code, bcptr, utf8))                       could_be_empty(called, code, bcptr, utf8, cd))
5546                {                {
5547                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
5548                goto FAILED;                goto FAILED;
# Line 4929  we set the flag only if there is a liter Line 5558  we set the flag only if there is a liter
5558            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
5559    
5560            *code = OP_RECURSE;            *code = OP_RECURSE;
5561            PUT(code, 1, called - cd->start_code);            PUT(code, 1, (int)(called - cd->start_code));
5562            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
5563    
5564            *code = OP_KET;            *code = OP_KET;
# Line 5008  we set the flag only if there is a liter Line 5637  we set the flag only if there is a liter
5637              {              {
5638              cd->external_options = newoptions;              cd->external_options = newoptions;
5639              }              }
5640           else            else
5641              {              {
5642              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))
5643                {                {
# Line 5040  we set the flag only if there is a liter Line 5669  we set the flag only if there is a liter
5669          }     /* End of switch for character following (? */          }     /* End of switch for character following (? */
5670        }       /* End of (? handling */        }       /* End of (? handling */
5671    
5672      /* Opening parenthesis not followed by '?'. If PCRE_NO_AUTO_CAPTURE is set,      /* Opening parenthesis not followed by '*' or '?'. If PCRE_NO_AUTO_CAPTURE
5673      all unadorned brackets become non-capturing and behave like (?:...)      is set, all unadorned brackets become non-capturing and behave like (?:...)
5674      brackets. */      brackets. */
5675    
5676      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)
# Line 5233  we set the flag only if there is a liter Line 5862  we set the flag only if there is a liter
5862    
5863      /* ===================================================================*/      /* ===================================================================*/
5864      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values
5865      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
5866      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
5867      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
5868      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
5869      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
5870        ever created. */
5871    
5872      case CHAR_BACKSLASH:      case CHAR_BACKSLASH:
5873      tempptr = ptr;      tempptr = ptr;
# Line 5345  we set the flag only if there is a liter Line 5975  we set the flag only if there is a liter
5975    
5976        if (-c >= ESC_REF)        if (-c >= ESC_REF)
5977          {          {
5978            open_capitem *oc;
5979          recno = -c - ESC_REF;          recno = -c - ESC_REF;
5980    
5981          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
# Line 5354  we set the flag only if there is a liter Line 5985  we set the flag only if there is a liter
5985          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
5986          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
5987          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
5988    
5989            /* Check to see if this back reference is recursive, that it, it
5990            is inside the group that it references. A flag is set so that the
5991            group can be made atomic. */
5992    
5993            for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5994              {
5995              if (oc->number == recno)
5996                {
5997                oc->flag = TRUE;
5998                break;
5999                }
6000              }
6001          }          }
6002    
6003        /* So are Unicode property matches, if supported. */        /* So are Unicode property matches, if supported. */
# Line 5383  we set the flag only if there is a liter Line 6027  we set the flag only if there is a liter
6027  #endif  #endif
6028    
6029        /* For the rest (including \X when Unicode properties are supported), we        /* For the rest (including \X when Unicode properties are supported), we
6030        can obtain the OP value by negating the escape value. */        can obtain the OP value by negating the escape value in the default
6031          situation when PCRE_UCP is not set. When it *is* set, we substitute
6032          Unicode property tests. */
6033    
6034        else        else
6035          {          {
6036          previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;  #ifdef SUPPORT_UCP
6037          *code++ = -c;          if (-c >= ESC_DU && -c <= ESC_wu)
6038              {
6039              nestptr = ptr + 1;                   /* Where to resume */
6040              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
6041              }
6042            else
6043    #endif
6044              {
6045              previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;
6046              *code++ = -c;
6047              }
6048          }          }
6049        continue;        continue;
6050        }        }
# Line 5536  uschar *code = *codeptr; Line 6192  uschar *code = *codeptr;
6192  uschar *last_branch = code;  uschar *last_branch = code;
6193  uschar *start_bracket = code;  uschar *start_bracket = code;
6194  uschar *reverse_count = NULL;  uschar *reverse_count = NULL;
6195    open_capitem capitem;
6196    int capnumber = 0;
6197  int firstbyte, reqbyte;  int firstbyte, reqbyte;
6198  int branchfirstbyte, branchreqbyte;  int branchfirstbyte, branchreqbyte;
6199  int length;  int length;
6200  int orig_bracount;  int orig_bracount;
6201  int max_bracount;  int max_bracount;
6202    int old_external_options = cd->external_options;
6203  branch_chain bc;  branch_chain bc;
6204    
6205  bc.outer = bcptr;  bc.outer = bcptr;
6206  bc.current = code;  bc.current_branch = code;
6207    
6208  firstbyte = reqbyte = REQ_UNSET;  firstbyte = reqbyte = REQ_UNSET;
6209    
# Line 5562  the code that abstracts option settings Line 6221  the code that abstracts option settings
6221  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
6222  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. */
6223    
6224    /* If this is a capturing subpattern, add to the chain of open capturing items
6225    so that we can detect them if (*ACCEPT) is encountered. This is also used to
6226    detect groups that contain recursive back references to themselves. */
6227    
6228    if (*code == OP_CBRA)
6229      {
6230      capnumber = GET2(code, 1 + LINK_SIZE);
6231      capitem.number = capnumber;
6232      capitem.next = cd->open_caps;
6233      capitem.flag = FALSE;
6234      cd->open_caps = &capitem;
6235      }
6236    
6237  /* Offset is set zero to mark that this bracket is still open */  /* Offset is set zero to mark that this bracket is still open */
6238    
6239  PUT(code, 1, 0);  PUT(code, 1, 0);
# Line 5606  for (;;) Line 6278  for (;;)
6278      return FALSE;      return FALSE;
6279      }      }
6280    
6281      /* If the external options have changed during this branch, it means that we
6282      are at the top level, and a leading option setting has been encountered. We
6283      need to re-set the original option values to take account of this so that,
6284      during the pre-compile phase, we know to allow for a re-set at the start of
6285      subsequent branches. */
6286    
6287      if (old_external_options != cd->external_options)
6288        oldims = cd->external_options & PCRE_IMS;
6289    
6290    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6291    has fewer than the rest. */    has fewer than the rest. */
6292    
# Line 5656  for (;;) Line 6337  for (;;)
6337    
6338      /* If lookbehind, check that this branch matches a fixed-length string, and      /* If lookbehind, check that this branch matches a fixed-length string, and
6339      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
6340      branch with OP_END. */      branch with OP_END. If the branch contains OP_RECURSE, the result is -3
6341        because there may be forward references that we can't check here. Set a
6342        flag to cause another lookbehind check at the end. Why not do it all at the
6343        end? Because common, erroneous checks are picked up here and the offset of
6344        the problem can be shown. */
6345    
6346      if (lookbehind)      if (lookbehind)
6347        {        {
6348        int fixed_length;        int fixed_length;
6349        *code = OP_END;        *code = OP_END;
6350        fixed_length = find_fixedlength(last_branch, options);        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);
6351        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6352        if (fixed_length < 0)        if (fixed_length == -3)
6353            {
6354            cd->check_lookbehind = TRUE;
6355            }
6356          else if (fixed_length < 0)
6357          {          {
6358          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;
6359          *ptrptr = ptr;          *ptrptr = ptr;
6360          return FALSE;          return FALSE;
6361          }          }
6362        PUT(reverse_count, 0, fixed_length);        else { PUT(reverse_count, 0, fixed_length); }
6363        }        }
6364      }      }
6365    
# Line 5687  for (;;) Line 6376  for (;;)
6376      {      {
6377      if (lengthptr == NULL)      if (lengthptr == NULL)
6378        {        {
6379        int branch_length = code - last_branch;        int branch_length = (int)(code - last_branch);
6380        do        do
6381          {          {
6382          int prev_length = GET(last_branch, 1);          int prev_length = GET(last_branch, 1);
# Line 5701  for (;;) Line 6390  for (;;)
6390      /* Fill in the ket */      /* Fill in the ket */
6391    
6392      *code = OP_KET;      *code = OP_KET;
6393      PUT(code, 1, code - start_bracket);      PUT(code, 1, (int)(code - start_bracket));
6394      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6395    
6396      /* Resetting option if needed */      /* If it was a capturing subpattern, check to see if it contained any
6397        recursive back references. If so, we must wrap it in atomic brackets.
6398        In any event, remove the block from the chain. */
6399    
6400        if (capnumber > 0)
6401          {
6402          if (cd->open_caps->flag)
6403            {
6404            memmove(start_bracket + 1 + LINK_SIZE, start_bracket,
6405              code - start_bracket);
6406            *start_bracket = OP_ONCE;
6407            code += 1 + LINK_SIZE;
6408            PUT(start_bracket, 1, (int)(code - start_bracket));
6409            *code = OP_KET;
6410            PUT(code, 1, (int)(code - start_bracket));
6411            code += 1 + LINK_SIZE;
6412            length += 2 + 2*LINK_SIZE;
6413            }
6414          cd->open_caps = cd->open_caps->next;
6415          }
6416    
6417        /* Reset options if needed. */
6418    
6419      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)
6420        {        {
# Line 5752  for (;;) Line 6462  for (;;)
6462    else    else
6463      {      {
6464      *code = OP_ALT;      *code = OP_ALT;
6465      PUT(code, 1, code - last_branch);      PUT(code, 1, (int)(code - last_branch));
6466      bc.current = last_branch = code;      bc.current_branch = last_branch = code;
6467      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6468      }      }
6469    
# Line 5900  do { Line 6610  do {
6610       switch (*scode)       switch (*scode)
6611         {         {
6612         case OP_CREF:         case OP_CREF:
6613           case OP_NCREF:
6614         case OP_RREF:         case OP_RREF:
6615           case OP_NRREF:
6616         case OP_DEF:         case OP_DEF:
6617         return FALSE;         return FALSE;
6618    
# Line 6069  int length = 1;  /* For final END opcode Line 6781  int length = 1;  /* For final END opcode
6781  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
6782  int errorcode = 0;  int errorcode = 0;
6783  int skipatstart = 0;  int skipatstart = 0;
 #ifdef SUPPORT_UTF8  
6784  BOOL utf8;  BOOL utf8;
 #endif  
6785  size_t size;  size_t size;
6786  uschar *code;  uschar *code;
6787  const uschar *codestart;  const uschar *codestart;
# Line 6114  if (erroroffset == NULL) Line 6824  if (erroroffset == NULL)
6824    
6825  *erroroffset = 0;  *erroroffset = 0;
6826    
 /* 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;  
   }  
   
6827  /* Set up pointers to the individual character tables */  /* Set up pointers to the individual character tables */
6828    
6829  if (tables == NULL) tables = _pcre_default_tables;  if (tables == NULL) tables = _pcre_default_tables;
# Line 6146  cd->fcc = tables + fcc_offset; Line 6832  cd->fcc = tables + fcc_offset;
6832  cd->cbits = tables + cbits_offset;  cd->cbits = tables + cbits_offset;
6833  cd->ctypes = tables + ctypes_offset;  cd->ctypes = tables + ctypes_offset;
6834    
6835    /* Check that all undefined public option bits are zero */
6836    
6837    if ((options & ~PUBLIC_COMPILE_OPTIONS) != 0)
6838      {
6839      errorcode = ERR17;
6840      goto PCRE_EARLY_ERROR_RETURN;
6841      }
6842    
6843  /* 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
6844  the offset for later. */  the offset for later. */
6845    
# Line 6155  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 6849  while (ptr[skipatstart] == CHAR_LEFT_PAR
6849    int newnl = 0;    int newnl = 0;
6850    int newbsr = 0;    int newbsr = 0;
6851    
6852      if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)
6853        { skipatstart += 7; options |= PCRE_UTF8; continue; }
6854      else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)
6855        { skipatstart += 6; options |= PCRE_UCP; continue; }
6856    
6857    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
6858      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
6859    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 6178  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 6877  while (ptr[skipatstart] == CHAR_LEFT_PAR
6877    else break;    else break;
6878    }    }
6879    
6880    utf8 = (options & PCRE_UTF8) != 0;
6881    
6882    /* Can't support UTF8 unless PCRE has been compiled to include the code. */
6883    
6884    #ifdef SUPPORT_UTF8
6885    if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
6886         (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)
6887      {
6888      errorcode = ERR44;
6889      goto PCRE_EARLY_ERROR_RETURN2;
6890      }
6891    #else
6892    if (utf8)
6893      {
6894      errorcode = ERR32;
6895      goto PCRE_EARLY_ERROR_RETURN;
6896      }
6897    #endif
6898    
6899    /* Can't support UCP unless PCRE has been compiled to include the code. */
6900    
6901    #ifndef SUPPORT_UCP
6902    if ((options & PCRE_UCP) != 0)
6903      {
6904      errorcode = ERR67;
6905      goto PCRE_EARLY_ERROR_RETURN;
6906      }
6907    #endif
6908    
6909  /* Check validity of \R options. */  /* Check validity of \R options. */
6910    
6911  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
# Line 6260  cd->end_pattern = (const uschar *)(patte Line 6988  cd->end_pattern = (const uschar *)(patte
6988  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6989  cd->external_options = options;  cd->external_options = options;
6990  cd->external_flags = 0;  cd->external_flags = 0;
6991    cd->open_caps = NULL;
6992    
6993  /* 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
6994  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 6305  regex compiled on a system with 4-byte p Line 7034  regex compiled on a system with 4-byte p
7034  pointers. */  pointers. */
7035    
7036  re->magic_number = MAGIC_NUMBER;  re->magic_number = MAGIC_NUMBER;
7037  re->size = size;  re->size = (int)size;
7038  re->options = cd->external_options;  re->options = cd->external_options;
7039  re->flags = cd->external_flags;  re->flags = cd->external_flags;
7040  re->dummy1 = 0;  re->dummy1 = 0;
# Line 6334  cd->start_code = codestart; Line 7063  cd->start_code = codestart;
7063  cd->hwm = cworkspace;  cd->hwm = cworkspace;
7064  cd->req_varyopt = 0;  cd->req_varyopt = 0;
7065  cd->had_accept = FALSE;  cd->had_accept = FALSE;
7066    cd->check_lookbehind = FALSE;
7067    cd->open_caps = NULL;
7068    
7069  /* Set up a starting, non-extracting bracket, then compile the expression. On  /* Set up a starting, non-extracting bracket, then compile the expression. On
7070  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 6359  if debugging, leave the test till after Line 7090  if debugging, leave the test till after
7090    
7091  *code++ = OP_END;  *code++ = OP_END;
7092    
7093  #ifndef DEBUG  #ifndef PCRE_DEBUG
7094  if (code - codestart > length) errorcode = ERR23;  if (code - codestart > length) errorcode = ERR23;
7095  #endif  #endif
7096    
# Line 6372  while (errorcode == 0 && cd->hwm > cwork Line 7103  while (errorcode == 0 && cd->hwm > cwork
7103    cd->hwm -= LINK_SIZE;    cd->hwm -= LINK_SIZE;
7104    offset = GET(cd->hwm, 0);    offset = GET(cd->hwm, 0);
7105    recno = GET(codestart, offset);    recno = GET(codestart, offset);
7106    groupptr = find_bracket(codestart, (re->options & PCRE_UTF8) != 0, recno);    groupptr = _pcre_find_bracket(codestart, utf8, recno);
7107    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
7108      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, (int)(groupptr - codestart));
7109    }    }
7110    
7111  /* 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 6382  subpattern. */ Line 7113  subpattern. */
7113    
7114  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;
7115    
7116    /* If there were any lookbehind assertions that contained OP_RECURSE
7117    (recursions or subroutine calls), a flag is set for them to be checked here,
7118    because they may contain forward references. Actual recursions can't be fixed
7119    length, but subroutine calls can. It is done like this so that those without
7120    OP_RECURSE that are not fixed length get a diagnosic with a useful offset. The
7121    exceptional ones forgo this. We scan the pattern to check that they are fixed
7122    length, and set their lengths. */
7123    
7124    if (cd->check_lookbehind)
7125      {
7126      uschar *cc = (uschar *)codestart;
7127    
7128      /* Loop, searching for OP_REVERSE items, and process those that do not have
7129      their length set. (Actually, it will also re-process any that have a length
7130      of zero, but that is a pathological case, and it does no harm.) When we find
7131      one, we temporarily terminate the branch it is in while we scan it. */
7132    
7133      for (cc = (uschar *)_pcre_find_bracket(codestart, utf8, -1);
7134           cc != NULL;
7135           cc = (uschar *)_pcre_find_bracket(cc, utf8, -1))
7136        {
7137        if (GET(cc, 1) == 0)
7138          {
7139          int fixed_length;
7140          uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
7141          int end_op = *be;
7142          *be = OP_END;
7143          fixed_length = find_fixedlength(cc, re->options, TRUE, cd);
7144          *be = end_op;
7145          DPRINTF(("fixed length = %d\n", fixed_length));
7146          if (fixed_length < 0)
7147            {
7148            errorcode = (fixed_length == -2)? ERR36 : ERR25;
7149            break;
7150            }
7151          PUT(cc, 1, fixed_length);
7152          }
7153        cc += 1 + LINK_SIZE;
7154        }
7155      }
7156    
7157  /* Failed to compile, or error while post-processing */  /* Failed to compile, or error while post-processing */
7158    
7159  if (errorcode != 0)  if (errorcode != 0)
7160    {    {
7161    (pcre_free)(re);    (pcre_free)(re);
7162    PCRE_EARLY_ERROR_RETURN:    PCRE_EARLY_ERROR_RETURN:
7163    *erroroffset = ptr - (const uschar *)pattern;    *erroroffset = (int)(ptr - (const uschar *)pattern);
7164    PCRE_EARLY_ERROR_RETURN2:    PCRE_EARLY_ERROR_RETURN2:
7165    *errorptr = find_error_text(errorcode);    *errorptr = find_error_text(errorcode);
7166    if (errorcodeptr != NULL) *errorcodeptr = errorcode;    if (errorcodeptr != NULL) *errorcodeptr = errorcode;
# Line 6442  if (reqbyte >= 0 && Line 7214  if (reqbyte >= 0 &&
7214  /* 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
7215  case when building a production library. */  case when building a production library. */
7216    
7217  #ifdef DEBUG  #ifdef PCRE_DEBUG
   
7218  printf("Length = %d top_bracket = %d top_backref = %d\n",  printf("Length = %d top_bracket = %d top_backref = %d\n",
7219    length, re->top_bracket, re->top_backref);    length, re->top_bracket, re->top_backref);
7220    
# Line 6480  if (code - codestart > length) Line 7251  if (code - codestart > length)
7251    if (errorcodeptr != NULL) *errorcodeptr = ERR23;    if (errorcodeptr != NULL) *errorcodeptr = ERR23;
7252    return NULL;    return NULL;
7253    }    }
7254  #endif   /* DEBUG */  #endif   /* PCRE_DEBUG */
7255    
7256  return (pcre *)re;  return (pcre *)re;
7257  }  }

Legend:
Removed from v.406  
changed lines
  Added in v.556

  ViewVC Help
Powered by ViewVC 1.1.5