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

Diff of /code/trunk/pcre_compile.c

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

revision 406 by ph10, Mon Mar 23 12:05:43 2009 UTC revision 552 by ph10, Wed Oct 13 10:15:41 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      count        pointer to the current capturing subpattern number (updated)
1114    
1115  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1116  */  */
1117    
1118  static int  static int
1119  find_parens(const uschar *ptr, compile_data *cd, const uschar *name, int lorn,  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1120    BOOL xmode)    BOOL xmode, int *count)
1121  {  {
1122  const uschar *thisname;  uschar *ptr = *ptrptr;
1123  int count = cd->bracount;  int start_count = *count;
1124    int hwm_count = start_count;
1125    BOOL dup_parens = FALSE;
1126    
1127  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1128    dealing with. The very first call may not start with a parenthesis. */
1129    
1130    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1131    {    {
1132    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1133    
1134      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1135    
1136      /* Handle a normal, unnamed capturing parenthesis. */
1137    
1138      else if (ptr[1] != CHAR_QUESTION_MARK)
1139        {
1140        *count += 1;
1141        if (name == NULL && *count == lorn) return *count;
1142        ptr++;
1143        }
1144    
1145      /* All cases now have (? at the start. Remember when we are in a group
1146      where the parenthesis numbers are duplicated. */
1147    
1148      else if (ptr[2] == CHAR_VERTICAL_LINE)
1149        {
1150        ptr += 3;
1151        dup_parens = TRUE;
1152        }
1153    
1154      /* Handle comments; all characters are allowed until a ket is reached. */
1155    
1156      else if (ptr[2] == CHAR_NUMBER_SIGN)
1157        {
1158        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1159        goto FAIL_EXIT;
1160        }
1161    
1162      /* Handle a condition. If it is an assertion, just carry on so that it
1163      is processed as normal. If not, skip to the closing parenthesis of the
1164      condition (there can't be any nested parens). */
1165    
1166      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1167        {
1168        ptr += 2;
1169        if (ptr[1] != CHAR_QUESTION_MARK)
1170          {
1171          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1172          if (*ptr != 0) ptr++;
1173          }
1174        }
1175    
1176      /* Start with (? but not a condition. */
1177    
1178      else
1179        {
1180        ptr += 2;
1181        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1182    
1183        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1184    
1185        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1186            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1187          {
1188          int term;
1189          const uschar *thisname;
1190          *count += 1;
1191          if (name == NULL && *count == lorn) return *count;
1192          term = *ptr++;
1193          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1194          thisname = ptr;
1195          while (*ptr != term) ptr++;
1196          if (name != NULL && lorn == ptr - thisname &&
1197              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1198            return *count;
1199          term++;
1200          }
1201        }
1202      }
1203    
1204    /* Past any initial parenthesis handling, scan for parentheses or vertical
1205    bars. */
1206    
1207    for (; *ptr != 0; ptr++)
1208      {
1209    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1210    
1211    if (*ptr == CHAR_BACKSLASH)    if (*ptr == CHAR_BACKSLASH)
1212      {      {
1213      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1214      if (*ptr == CHAR_Q) for (;;)      if (*ptr == CHAR_Q) for (;;)
1215        {        {
1216        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1217        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1218        if (*(++ptr) == CHAR_E) break;        if (*(++ptr) == CHAR_E) break;
1219        }        }
1220      continue;      continue;
# Line 1065  for (; *ptr != 0; ptr++) Line 1231  for (; *ptr != 0; ptr++)
1231      BOOL negate_class = FALSE;      BOOL negate_class = FALSE;
1232      for (;;)      for (;;)
1233        {        {
1234        int c = *(++ptr);        if (ptr[1] == CHAR_BACKSLASH)
       if (c == CHAR_BACKSLASH)  
1235          {          {
1236          if (ptr[1] == CHAR_E)          if (ptr[2] == CHAR_E)
1237            ptr++;            ptr+= 2;
1238          else if (strncmp((const char *)ptr+1,          else if (strncmp((const char *)ptr+2,
1239                   STR_Q STR_BACKSLASH STR_E, 3) == 0)                   STR_Q STR_BACKSLASH STR_E, 3) == 0)
1240            ptr += 3;            ptr += 4;
1241          else          else
1242            break;            break;
1243          }          }
1244        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)        else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1245            {
1246          negate_class = TRUE;          negate_class = TRUE;
1247            ptr++;
1248            }
1249        else break;        else break;
1250        }        }
1251    
# Line 1093  for (; *ptr != 0; ptr++) Line 1261  for (; *ptr != 0; ptr++)
1261        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1262        if (*ptr == CHAR_BACKSLASH)        if (*ptr == CHAR_BACKSLASH)
1263          {          {
1264          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1265          if (*ptr == CHAR_Q) for (;;)          if (*ptr == CHAR_Q) for (;;)
1266            {            {
1267            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1268            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1269            if (*(++ptr) == CHAR_E) break;            if (*(++ptr) == CHAR_E) break;
1270            }            }
1271          continue;          continue;
# Line 1111  for (; *ptr != 0; ptr++) Line 1279  for (; *ptr != 0; ptr++)
1279    if (xmode && *ptr == CHAR_NUMBER_SIGN)    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1280      {      {
1281      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1282      if (*ptr == 0) return -1;      if (*ptr == 0) goto FAIL_EXIT;
1283      continue;      continue;
1284      }      }
1285    
1286    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1287    
1288    if (*ptr != CHAR_LEFT_PARENTHESIS) continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)  
1289      {      {
1290      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1291      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1292      continue;      if (*ptr == 0) goto FAIL_EXIT;
1293      }      }
1294    
1295    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1296    if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */      {
1297        if (dup_parens && *count < hwm_count) *count = hwm_count;
1298        goto FAIL_EXIT;
1299        }
1300    
1301    /* We have to disambiguate (?<! and (?<= from (?<name> */    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1302        {
1303        if (*count > hwm_count) hwm_count = *count;
1304        *count = start_count;
1305        }
1306      }
1307    
1308    FAIL_EXIT:
1309    *ptrptr = ptr;
1310    return -1;
1311    }
1312    
   if ((*ptr != CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_EXCLAMATION_MARK ||  
       ptr[1] == CHAR_EQUALS_SIGN) && *ptr != CHAR_APOSTROPHE)  
     continue;  
1313    
   count++;  
1314    
1315    if (name == NULL && count == lorn) return count;  
1316    term = *ptr++;  /*************************************************
1317    if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;  *       Find forward referenced subpattern       *
1318    thisname = ptr;  *************************************************/
1319    while (*ptr != term) ptr++;  
1320    if (name != NULL && lorn == ptr - thisname &&  /* This function scans along a pattern's text looking for capturing
1321        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  subpatterns, and counting them. If it finds a named pattern that matches the
1322      return count;  name it is given, it returns its number. Alternatively, if the name is NULL, it
1323    returns when it reaches a given numbered subpattern. This is used for forward
1324    references to subpatterns. We used to be able to start this scan from the
1325    current compiling point, using the current count value from cd->bracount, and
1326    do it all in a single loop, but the addition of the possibility of duplicate
1327    subpattern numbers means that we have to scan from the very start, in order to
1328    take account of such duplicates, and to use a recursive function to keep track
1329    of the different types of group.
1330    
1331    Arguments:
1332      cd           compile background data
1333      name         name to seek, or NULL if seeking a numbered subpattern
1334      lorn         name length, or subpattern number if name is NULL
1335      xmode        TRUE if we are in /x mode
1336    
1337    Returns:       the number of the found subpattern, or -1 if not found
1338    */
1339    
1340    static int
1341    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1342    {
1343    uschar *ptr = (uschar *)cd->start_pattern;
1344    int count = 0;
1345    int rc;
1346    
1347    /* If the pattern does not start with an opening parenthesis, the first call
1348    to find_parens_sub() will scan right to the end (if necessary). However, if it
1349    does start with a parenthesis, find_parens_sub() will return when it hits the
1350    matching closing parens. That is why we have to have a loop. */
1351    
1352    for (;;)
1353      {
1354      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1355      if (rc > 0 || *ptr++ == 0) break;
1356    }    }
1357    
1358  return -1;  return rc;
1359  }  }
1360    
1361    
1362    
1363    
1364  /*************************************************  /*************************************************
1365  *      Find first significant op code            *  *      Find first significant op code            *
1366  *************************************************/  *************************************************/
# Line 1200  for (;;) Line 1410  for (;;)
1410    
1411      case OP_CALLOUT:      case OP_CALLOUT:
1412      case OP_CREF:      case OP_CREF:
1413        case OP_NCREF:
1414      case OP_RREF:      case OP_RREF:
1415        case OP_NRREF:
1416      case OP_DEF:      case OP_DEF:
1417      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1418      break;      break;
# Line 1216  for (;;) Line 1428  for (;;)
1428    
1429    
1430  /*************************************************  /*************************************************
1431  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1432  *************************************************/  *************************************************/
1433    
1434  /* 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,
1435  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.
1436  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
1437    temporarily terminated with OP_END when this function is called.
1438    
1439    This function is called when a backward assertion is encountered, so that if it
1440    fails, the error message can point to the correct place in the pattern.
1441    However, we cannot do this when the assertion contains subroutine calls,
1442    because they can be forward references. We solve this by remembering this case
1443    and doing the check at the end; a flag specifies which mode we are running in.
1444    
1445  Arguments:  Arguments:
1446    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1447    options  the compiling options    options  the compiling options
1448      atend    TRUE if called when the pattern is complete
1449      cd       the "compile data" structure
1450    
1451  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1452                 or -1 if there is no fixed length,
1453               or -2 if \C was encountered               or -2 if \C was encountered
1454                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1455  */  */
1456    
1457  static int  static int
1458  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1459  {  {
1460  int length = -1;  int length = -1;
1461    
# Line 1245  branch, check the length against that of Line 1468  branch, check the length against that of
1468  for (;;)  for (;;)
1469    {    {
1470    int d;    int d;
1471      uschar *ce, *cs;
1472    register int op = *cc;    register int op = *cc;
1473    switch (op)    switch (op)
1474      {      {
# Line 1252  for (;;) Line 1476  for (;;)
1476      case OP_BRA:      case OP_BRA:
1477      case OP_ONCE:      case OP_ONCE:
1478      case OP_COND:      case OP_COND:
1479      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1480      if (d < 0) return d;      if (d < 0) return d;
1481      branchlength += d;      branchlength += d;
1482      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1275  for (;;) Line 1499  for (;;)
1499      branchlength = 0;      branchlength = 0;
1500      break;      break;
1501    
1502        /* A true recursion implies not fixed length, but a subroutine call may
1503        be OK. If the subroutine is a forward reference, we can't deal with
1504        it until the end of the pattern, so return -3. */
1505    
1506        case OP_RECURSE:
1507        if (!atend) return -3;
1508        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1509        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1510        if (cc > cs && cc < ce) return -1;                /* Recursion */
1511        d = find_fixedlength(cs + 2, options, atend, cd);
1512        if (d < 0) return d;
1513        branchlength += d;
1514        cc += 1 + LINK_SIZE;
1515        break;
1516    
1517      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1518    
1519      case OP_ASSERT:      case OP_ASSERT:
# Line 1288  for (;;) Line 1527  for (;;)
1527    
1528      case OP_REVERSE:      case OP_REVERSE:
1529      case OP_CREF:      case OP_CREF:
1530        case OP_NCREF:
1531      case OP_RREF:      case OP_RREF:
1532        case OP_NRREF:
1533      case OP_DEF:      case OP_DEF:
1534      case OP_OPT:      case OP_OPT:
1535      case OP_CALLOUT:      case OP_CALLOUT:
1536      case OP_SOD:      case OP_SOD:
1537      case OP_SOM:      case OP_SOM:
1538        case OP_SET_SOM:
1539      case OP_EOD:      case OP_EOD:
1540      case OP_EODN:      case OP_EODN:
1541      case OP_CIRC:      case OP_CIRC:
# Line 1311  for (;;) Line 1553  for (;;)
1553      branchlength++;      branchlength++;
1554      cc += 2;      cc += 2;
1555  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1556      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1557        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1558  #endif  #endif
1559      break;      break;
1560    
# Line 1325  for (;;) Line 1565  for (;;)
1565      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1566      cc += 4;      cc += 4;
1567  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1568      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1569        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1570  #endif  #endif
1571      break;      break;
1572    
# Line 1407  for (;;) Line 1645  for (;;)
1645    
1646    
1647  /*************************************************  /*************************************************
1648  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1649  *************************************************/  *************************************************/
1650    
1651  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1652  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1653    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1654    so that it can be called from pcre_study() when finding the minimum matching
1655    length.
1656    
1657  Arguments:  Arguments:
1658    code        points to start of expression    code        points to start of expression
1659    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1660    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1661    
1662  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
1663  */  */
1664    
1665  static const uschar *  const uschar *
1666  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1667  {  {
1668  for (;;)  for (;;)
1669    {    {
# Line 1435  for (;;) Line 1676  for (;;)
1676    
1677    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1678    
1679      /* Handle recursion */
1680    
1681      else if (c == OP_REVERSE)
1682        {
1683        if (number < 0) return (uschar *)code;
1684        code += _pcre_OP_lengths[c];
1685        }
1686    
1687    /* Handle capturing bracket */    /* Handle capturing bracket */
1688    
1689    else if (c == OP_CBRA)    else if (c == OP_CBRA)
# Line 1446  for (;;) Line 1695  for (;;)
1695    
1696    /* 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
1697    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
1698    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1699      must add in its length. */
1700    
1701    else    else
1702      {      {
# Line 1470  for (;;) Line 1720  for (;;)
1720        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1721        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1722        break;        break;
1723    
1724          case OP_MARK:
1725          case OP_PRUNE_ARG:
1726          case OP_SKIP_ARG:
1727          code += code[1];
1728          break;
1729    
1730          case OP_THEN_ARG:
1731          code += code[1+LINK_SIZE];
1732          break;
1733        }        }
1734    
1735      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1541  for (;;) Line 1801  for (;;)
1801    
1802    /* 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
1803    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
1804    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1805      must add in its length. */
1806    
1807    else    else
1808      {      {
# Line 1565  for (;;) Line 1826  for (;;)
1826        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1827        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1828        break;        break;
1829    
1830          case OP_MARK:
1831          case OP_PRUNE_ARG:
1832          case OP_SKIP_ARG:
1833          code += code[1];
1834          break;
1835    
1836          case OP_THEN_ARG:
1837          code += code[1+LINK_SIZE];
1838          break;
1839        }        }
1840    
1841      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1621  Arguments: Line 1892  Arguments:
1892    code        points to start of search    code        points to start of search
1893    endcode     points to where to stop    endcode     points to where to stop
1894    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1895      cd          contains pointers to tables etc.
1896    
1897  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1898  */  */
1899    
1900  static BOOL  static BOOL
1901  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1902      compile_data *cd)
1903  {  {
1904  register int c;  register int c;
1905  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 1930  for (code = first_significant_code(code
1930      continue;      continue;
1931      }      }
1932    
1933      /* For a recursion/subroutine call, if its end has been reached, which
1934      implies a subroutine call, we can scan it. */
1935    
1936      if (c == OP_RECURSE)
1937        {
1938        BOOL empty_branch = FALSE;
1939        const uschar *scode = cd->start_code + GET(code, 1);
1940        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1941        do
1942          {
1943          if (could_be_empty_branch(scode, endcode, utf8, cd))
1944            {
1945            empty_branch = TRUE;
1946            break;
1947            }
1948          scode += GET(scode, 1);
1949          }
1950        while (*scode == OP_ALT);
1951        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1952        continue;
1953        }
1954    
1955    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
1956    
1957    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 1970  for (code = first_significant_code(code
1970        empty_branch = FALSE;        empty_branch = FALSE;
1971        do        do
1972          {          {
1973          if (!empty_branch && could_be_empty_branch(code, endcode, utf8))          if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1974            empty_branch = TRUE;            empty_branch = TRUE;
1975          code += GET(code, 1);          code += GET(code, 1);
1976          }          }
# Line 1800  for (code = first_significant_code(code Line 2095  for (code = first_significant_code(code
2095      case OP_QUERY:      case OP_QUERY:
2096      case OP_MINQUERY:      case OP_MINQUERY:
2097      case OP_POSQUERY:      case OP_POSQUERY:
2098        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2099        break;
2100    
2101      case OP_UPTO:      case OP_UPTO:
2102      case OP_MINUPTO:      case OP_MINUPTO:
2103      case OP_POSUPTO:      case OP_POSUPTO:
2104      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2105      break;      break;
2106  #endif  #endif
2107    
2108        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2109        string. */
2110    
2111        case OP_MARK:
2112        case OP_PRUNE_ARG:
2113        case OP_SKIP_ARG:
2114        code += code[1];
2115        break;
2116    
2117        case OP_THEN_ARG:
2118        code += code[1+LINK_SIZE];
2119        break;
2120    
2121        /* None of the remaining opcodes are required to match a character. */
2122    
2123        default:
2124        break;
2125      }      }
2126    }    }
2127    
# Line 1828  Arguments: Line 2144  Arguments:
2144    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2145    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2146    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2147      cd          pointers to tables etc
2148    
2149  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2150  */  */
2151    
2152  static BOOL  static BOOL
2153  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2154    BOOL utf8)    BOOL utf8, compile_data *cd)
2155  {  {
2156  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2157    {    {
2158    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2159        return FALSE;
2160    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2161    }    }
2162  return TRUE;  return TRUE;
# Line 2020  auto_callout(uschar *code, const uschar Line 2338  auto_callout(uschar *code, const uschar
2338  {  {
2339  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2340  *code++ = 255;  *code++ = 255;
2341  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2342  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2343  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2344  }  }
2345    
# Line 2046  Returns: nothing Line 2364  Returns: nothing
2364  static void  static void
2365  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2366  {  {
2367  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2368  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2369  }  }
2370    
# Line 2096  for (++c; c <= d; c++) Line 2414  for (++c; c <= d; c++)
2414    
2415  return TRUE;  return TRUE;
2416  }  }
2417    
2418    
2419    
2420    /*************************************************
2421    *        Check a character and a property        *
2422    *************************************************/
2423    
2424    /* This function is called by check_auto_possessive() when a property item
2425    is adjacent to a fixed character.
2426    
2427    Arguments:
2428      c            the character
2429      ptype        the property type
2430      pdata        the data for the type
2431      negated      TRUE if it's a negated property (\P or \p{^)
2432    
2433    Returns:       TRUE if auto-possessifying is OK
2434    */
2435    
2436    static BOOL
2437    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2438    {
2439    const ucd_record *prop = GET_UCD(c);
2440    switch(ptype)
2441      {
2442      case PT_LAMP:
2443      return (prop->chartype == ucp_Lu ||
2444              prop->chartype == ucp_Ll ||
2445              prop->chartype == ucp_Lt) == negated;
2446    
2447      case PT_GC:
2448      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2449    
2450      case PT_PC:
2451      return (pdata == prop->chartype) == negated;
2452    
2453      case PT_SC:
2454      return (pdata == prop->script) == negated;
2455    
2456      /* These are specials */
2457    
2458      case PT_ALNUM:
2459      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2460              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2461    
2462      case PT_SPACE:    /* Perl space */
2463      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2464              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2465              == negated;
2466    
2467      case PT_PXSPACE:  /* POSIX space */
2468      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2469              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2470              c == CHAR_FF || c == CHAR_CR)
2471              == negated;
2472    
2473      case PT_WORD:
2474      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2475              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2476              c == CHAR_UNDERSCORE) == negated;
2477      }
2478    return FALSE;
2479    }
2480  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2481    
2482    
# Line 2109  whether the next thing could possibly ma Line 2490  whether the next thing could possibly ma
2490  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2491    
2492  Arguments:  Arguments:
2493    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2494    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2495    ptr           next character in pattern    ptr           next character in pattern
2496    options       options bits    options       options bits
2497    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2121  Returns: TRUE if possessifying is Line 2500  Returns: TRUE if possessifying is
2500  */  */
2501    
2502  static BOOL  static BOOL
2503  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2504    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2505  {  {
2506  int next;  int c, next;
2507    int op_code = *previous++;
2508    
2509  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2510    
# Line 2185  if (*ptr == CHAR_ASTERISK || *ptr == CHA Line 2565  if (*ptr == CHAR_ASTERISK || *ptr == CHA
2565    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)
2566      return FALSE;      return FALSE;
2567    
2568  /* 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
2569  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. */  
2570    
2571  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2572    {    {
2573    case OP_CHAR:    case OP_CHAR:
2574  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2575    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2576  #else  #else
2577    (void)(utf8_char);  /* Keep compiler happy by referencing function argument */    c = *previous;
2578  #endif  #endif
2579    return item != next;    return c != next;
2580    
2581    /* 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
2582    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 2584  if (next >= 0) switch(op_code)
2584    
2585    case OP_CHARNC:    case OP_CHARNC:
2586  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2587    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2588    #else
2589      c = *previous;
2590  #endif  #endif
2591    if (item == next) return FALSE;    if (c == next) return FALSE;
2592  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2593    if (utf8)    if (utf8)
2594      {      {
# Line 2222  if (next >= 0) switch(op_code) Line 2599  if (next >= 0) switch(op_code)
2599  #else  #else
2600      othercase = NOTACHAR;      othercase = NOTACHAR;
2601  #endif  #endif
2602      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2603      }      }
2604    else    else
2605  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2606    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2607    
2608    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, its data is always a single-byte character. */
2609    
2610    case OP_NOT:    case OP_NOT:
2611    if (item == next) return TRUE;    if ((c = *previous) == next) return TRUE;
2612    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2613  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2614    if (utf8)    if (utf8)
# Line 2243  if (next >= 0) switch(op_code) Line 2620  if (next >= 0) switch(op_code)
2620  #else  #else
2621      othercase = NOTACHAR;      othercase = NOTACHAR;
2622  #endif  #endif
2623      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2624      }      }
2625    else    else
2626  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2627    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2628    
2629      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2630      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2631    
2632    case OP_DIGIT:    case OP_DIGIT:
2633    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 2670  if (next >= 0) switch(op_code)
2670      case 0x202f:      case 0x202f:
2671      case 0x205f:      case 0x205f:
2672      case 0x3000:      case 0x3000:
2673      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2674      default:      default:
2675      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2676      }      }
2677    
2678      case OP_ANYNL:
2679    case OP_VSPACE:    case OP_VSPACE:
2680    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2681    switch(next)    switch(next)
# Line 2306  if (next >= 0) switch(op_code) Line 2687  if (next >= 0) switch(op_code)
2687      case 0x85:      case 0x85:
2688      case 0x2028:      case 0x2028:
2689      case 0x2029:      case 0x2029:
2690      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2691      default:      default:
2692      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2693      }      }
2694    
2695    #ifdef SUPPORT_UCP
2696      case OP_PROP:
2697      return check_char_prop(next, previous[0], previous[1], FALSE);
2698    
2699      case OP_NOTPROP:
2700      return check_char_prop(next, previous[0], previous[1], TRUE);
2701    #endif
2702    
2703    default:    default:
2704    return FALSE;    return FALSE;
2705    }    }
2706    
2707    
2708  /* 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
2709    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2710    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2711    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2712    replaced by OP_PROP codes when PCRE_UCP is set. */
2713    
2714  switch(op_code)  switch(op_code)
2715    {    {
2716    case OP_CHAR:    case OP_CHAR:
2717    case OP_CHARNC:    case OP_CHARNC:
2718  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2719    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2720    #else
2721      c = *previous;
2722  #endif  #endif
2723    switch(-next)    switch(-next)
2724      {      {
2725      case ESC_d:      case ESC_d:
2726      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2727    
2728      case ESC_D:      case ESC_D:
2729      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2730    
2731      case ESC_s:      case ESC_s:
2732      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2733    
2734      case ESC_S:      case ESC_S:
2735      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2736    
2737      case ESC_w:      case ESC_w:
2738      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2739    
2740      case ESC_W:      case ESC_W:
2741      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2742    
2743      case ESC_h:      case ESC_h:
2744      case ESC_H:      case ESC_H:
2745      switch(item)      switch(c)
2746        {        {
2747        case 0x09:        case 0x09:
2748        case 0x20:        case 0x20:
# Line 2375  switch(op_code) Line 2770  switch(op_code)
2770    
2771      case ESC_v:      case ESC_v:
2772      case ESC_V:      case ESC_V:
2773      switch(item)      switch(c)
2774        {        {
2775        case 0x0a:        case 0x0a:
2776        case 0x0b:        case 0x0b:
# Line 2389  switch(op_code) Line 2784  switch(op_code)
2784        return -next == ESC_v;        return -next == ESC_v;
2785        }        }
2786    
2787        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2788        their substitutions and process them. The result will always be either
2789        -ESC_p or -ESC_P. Then fall through to process those values. */
2790    
2791    #ifdef SUPPORT_UCP
2792        case ESC_du:
2793        case ESC_DU:
2794        case ESC_wu:
2795        case ESC_WU:
2796        case ESC_su:
2797        case ESC_SU:
2798          {
2799          int temperrorcode = 0;
2800          ptr = substitutes[-next - ESC_DU];
2801          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2802          if (temperrorcode != 0) return FALSE;
2803          ptr++;    /* For compatibility */
2804          }
2805        /* Fall through */
2806    
2807        case ESC_p:
2808        case ESC_P:
2809          {
2810          int ptype, pdata, errorcodeptr;
2811          BOOL negated;
2812    
2813          ptr--;      /* Make ptr point at the p or P */
2814          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2815          if (ptype < 0) return FALSE;
2816          ptr++;      /* Point past the final curly ket */
2817    
2818          /* If the property item is optional, we have to give up. (When generated
2819          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2820          to the original \d etc. At this point, ptr will point to a zero byte. */
2821    
2822          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2823            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2824              return FALSE;
2825    
2826          /* Do the property check. */
2827    
2828          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2829          }
2830    #endif
2831    
2832      default:      default:
2833      return FALSE;      return FALSE;
2834      }      }
2835    
2836      /* In principle, support for Unicode properties should be integrated here as
2837      well. It means re-organizing the above code so as to get hold of the property
2838      values before switching on the op-code. However, I wonder how many patterns
2839      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2840      these op-codes are never generated.) */
2841    
2842    case OP_DIGIT:    case OP_DIGIT:
2843    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2844           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
2845    
2846    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2847    return next == -ESC_d;    return next == -ESC_d;
2848    
2849    case OP_WHITESPACE:    case OP_WHITESPACE:
2850    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2851    
2852    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2853    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2854    
2855    case OP_HSPACE:    case OP_HSPACE:
2856    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2857             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2858    
2859    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
2860    return next == -ESC_h;    return next == -ESC_h;
2861    
2862    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
2863      case OP_ANYNL:
2864    case OP_VSPACE:    case OP_VSPACE:
2865    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2866    
2867    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2868    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
2869    
2870    case OP_WORDCHAR:    case OP_WORDCHAR:
2871    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2872             next == -ESC_v || next == -ESC_R;
2873    
2874    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2875    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2484  BOOL inescq = FALSE; Line 2933  BOOL inescq = FALSE;
2933  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
2934  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
2935  const uschar *tempptr;  const uschar *tempptr;
2936    const uschar *nestptr = NULL;
2937  uschar *previous = NULL;  uschar *previous = NULL;
2938  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2939  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
# Line 2500  BOOL utf8 = FALSE; Line 2950  BOOL utf8 = FALSE;
2950  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2951  #endif  #endif
2952    
2953  #ifdef DEBUG  #ifdef PCRE_DEBUG
2954  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2955  #endif  #endif
2956    
# Line 2554  for (;; ptr++) Line 3004  for (;; ptr++)
3004    
3005    c = *ptr;    c = *ptr;
3006    
3007      /* If we are at the end of a nested substitution, revert to the outer level
3008      string. Nesting only happens one level deep. */
3009    
3010      if (c == 0 && nestptr != NULL)
3011        {
3012        ptr = nestptr;
3013        nestptr = NULL;
3014        c = *ptr;
3015        }
3016    
3017    /* 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
3018    previous cycle of this loop. */    previous cycle of this loop. */
3019    
3020    if (lengthptr != NULL)    if (lengthptr != NULL)
3021      {      {
3022  #ifdef DEBUG  #ifdef PCRE_DEBUG
3023      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3024  #endif  #endif
3025      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3026        {        {
3027        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3028        goto FAILED;        goto FAILED;
# Line 2584  for (;; ptr++) Line 3044  for (;; ptr++)
3044        goto FAILED;        goto FAILED;
3045        }        }
3046    
3047      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3048      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));
3049    
3050      /* 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 3071  for (;; ptr++)
3071    /* 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
3072    reference list. */    reference list. */
3073    
3074    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3075      {      {
3076      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3077      goto FAILED;      goto FAILED;
# Line 2702  for (;; ptr++) Line 3162  for (;; ptr++)
3162          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3163          goto FAILED;          goto FAILED;
3164          }          }
3165        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3166        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3167        }        }
3168      return TRUE;      return TRUE;
# Line 2907  for (;; ptr++) Line 3367  for (;; ptr++)
3367            ptr++;            ptr++;
3368            }            }
3369    
3370          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3371          if (posix_class < 0)          if (posix_class < 0)
3372            {            {
3373            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2921  for (;; ptr++) Line 3381  for (;; ptr++)
3381          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3382            posix_class = 0;            posix_class = 0;
3383    
3384          /* 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
3385          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3386          subtract bits that may be in the main map already. At the end we or the  
3387          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3388            if ((options & PCRE_UCP) != 0)
3389              {
3390              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3391              if (posix_substitutes[pc] != NULL)
3392                {
3393                nestptr = tempptr + 1;
3394                ptr = posix_substitutes[pc] - 1;
3395                continue;
3396                }
3397              }
3398    #endif
3399            /* In the non-UCP case, we build the bit map for the POSIX class in a
3400            chunk of local store because we may be adding and subtracting from it,
3401            and we don't want to subtract bits that may be in the main map already.
3402            At the end we or the result into the bit map that is being built. */
3403    
3404          posix_class *= 3;          posix_class *= 3;
3405    
# Line 2968  for (;; ptr++) Line 3443  for (;; ptr++)
3443    
3444        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3445        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
3446        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
3447        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
3448        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
3449        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3450          PCRE_EXTRA is set. */
3451    
3452        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
3453          {          {
3454          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3455          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3456    
3457          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 */  
3458          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3459            {            {
3460            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
# Line 2997  for (;; ptr++) Line 3471  for (;; ptr++)
3471            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3472            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3473    
3474            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3475              {              {
3476    #ifdef SUPPORT_UCP
3477                case ESC_du:     /* These are the values given for \d etc */
3478                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3479                case ESC_wu:     /* escape sequence with an appropriate \p */
3480                case ESC_WU:     /* or \P to test Unicode properties instead */
3481                case ESC_su:     /* of the default ASCII testing. */
3482                case ESC_SU:
3483                nestptr = ptr;
3484                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3485                class_charcount -= 2;                /* Undo! */
3486                continue;
3487    #endif
3488              case ESC_d:              case ESC_d:
3489              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3490              continue;              continue;
# Line 3019  for (;; ptr++) Line 3503  for (;; ptr++)
3503              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3504              continue;              continue;
3505    
3506                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3507                if it was previously set by something earlier in the character
3508                class. */
3509    
3510              case ESC_s:              case ESC_s:
3511              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3512              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3513                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3514              continue;              continue;
3515    
3516              case ESC_S:              case ESC_S:
# Line 3030  for (;; ptr++) Line 3519  for (;; ptr++)
3519              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3520              continue;              continue;
3521    
3522              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)  
             {  
3523              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3524              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3525              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 3067  for (;; ptr++) Line 3543  for (;; ptr++)
3543                }                }
3544  #endif  #endif
3545              continue;              continue;
             }  
3546    
3547            if (-c == ESC_H)              case ESC_H:
             {  
3548              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3549                {                {
3550                int x = 0xff;                int x = 0xff;
# Line 3112  for (;; ptr++) Line 3586  for (;; ptr++)
3586                }                }
3587  #endif  #endif
3588              continue;              continue;
             }  
3589    
3590            if (-c == ESC_v)              case ESC_v:
             {  
3591              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3592              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3593              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 3131  for (;; ptr++) Line 3603  for (;; ptr++)
3603                }                }
3604  #endif  #endif
3605              continue;              continue;
             }  
3606    
3607            if (-c == ESC_V)              case ESC_V:
             {  
3608              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3609                {                {
3610                int x = 0xff;                int x = 0xff;
# Line 3164  for (;; ptr++) Line 3634  for (;; ptr++)
3634                }                }
3635  #endif  #endif
3636              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3637    
3638  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3639            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3640              {              case ESC_P:
3641              BOOL negated;                {
3642              int pdata;                BOOL negated;
3643              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3644              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3645              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3646              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3647                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3648              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3649              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3650              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3651              continue;                class_charcount -= 2;   /* Not a < 256 character */
3652              }                continue;
3653                  }
3654  #endif  #endif
3655            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3656            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3657            treated as literals. */              treated as literals. */
3658    
3659            if ((options & PCRE_EXTRA) != 0)              default:
3660              {              if ((options & PCRE_EXTRA) != 0)
3661              *errorcodeptr = ERR7;                {
3662              goto FAILED;                *errorcodeptr = ERR7;
3663                  goto FAILED;
3664                  }
3665                class_charcount -= 2;  /* Undo the default count from above */
3666                c = *ptr;              /* Get the final character and fall through */
3667                break;
3668              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3669            }            }
3670    
3671          /* 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 3735  for (;; ptr++)
3735            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3736            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3737    
3738            /* \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 */  
3739    
3740            if (d < 0)            if (d < 0)
3741              {              {
3742              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  
3743                {                {
3744                ptr = oldptr;                ptr = oldptr;
3745                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3438  for (;; ptr++) Line 3905  for (;; ptr++)
3905          }          }
3906        }        }
3907    
3908      /* 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.
3909        If we are at the end of an internal nested string, revert to the outer
3910        string. */
3911    
3912        while (((c = *(++ptr)) != 0 ||
3913               (nestptr != NULL &&
3914                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
3915               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3916    
3917      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));      /* Check for missing terminating ']' */
3918    
3919      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
3920        {        {
3921        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
3922        goto FAILED;        goto FAILED;
3923        }        }
3924    
   
 /* 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  
   
   
3925      /* 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
3926      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
3927      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 3985  we set the flag only if there is a liter
3985    
3986      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3987      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
3988      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
3989      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
3990      (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
3991      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
3992        actual compiled code. */
3993    
3994  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3995      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
3996        {        {
3997        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3998        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3562  we set the flag only if there is a liter Line 4018  we set the flag only if there is a liter
4018        }        }
4019  #endif  #endif
4020    
4021      /* 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
4022      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
4023      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
4024      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4025        negating it if necessary. */
4026    
4027      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4028      if (negate_class)      if (negate_class)
# Line 3689  we set the flag only if there is a liter Line 4146  we set the flag only if there is a liter
4146    
4147        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4148            repeat_max < 0 &&            repeat_max < 0 &&
4149            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4150          {          {
4151          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4152          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3711  we set the flag only if there is a liter Line 4167  we set the flag only if there is a liter
4167        c = previous[1];        c = previous[1];
4168        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4169            repeat_max < 0 &&            repeat_max < 0 &&
4170            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4171          {          {
4172          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4173          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3735  we set the flag only if there is a liter Line 4191  we set the flag only if there is a liter
4191    
4192        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4193            repeat_max < 0 &&            repeat_max < 0 &&
4194            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4195          {          {
4196          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4197          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3757  we set the flag only if there is a liter Line 4213  we set the flag only if there is a liter
4213    
4214        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4215    
4216          /*--------------------------------------------------------------------*/
4217          /* This code is obsolete from release 8.00; the restriction was finally
4218          removed: */
4219    
4220        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4221        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4222    
4223        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4224          /*--------------------------------------------------------------------*/
4225    
4226        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4227    
# Line 3907  we set the flag only if there is a liter Line 4368  we set the flag only if there is a liter
4368          goto END_REPEAT;          goto END_REPEAT;
4369          }          }
4370    
4371          /*--------------------------------------------------------------------*/
4372          /* This code is obsolete from release 8.00; the restriction was finally
4373          removed: */
4374    
4375        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4376        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4377    
4378        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4379          /*--------------------------------------------------------------------*/
4380    
4381        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4382          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3935  we set the flag only if there is a liter Line 4401  we set the flag only if there is a liter
4401        {        {
4402        register int i;        register int i;
4403        int ketoffset = 0;        int ketoffset = 0;
4404        int len = code - previous;        int len = (int)(code - previous);
4405        uschar *bralink = NULL;        uschar *bralink = NULL;
4406    
4407        /* 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 4422  we set the flag only if there is a liter
4422          {          {
4423          register uschar *ket = previous;          register uschar *ket = previous;
4424          do ket += GET(ket, 1); while (*ket != OP_KET);          do ket += GET(ket, 1); while (*ket != OP_KET);
4425          ketoffset = code - ket;          ketoffset = (int)(code - ket);
4426          }          }
4427    
4428        /* 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 4490  we set the flag only if there is a liter
4490            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4491            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4492    
4493            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4494            bralink = previous;            bralink = previous;
4495            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4496            }            }
# Line 4045  we set the flag only if there is a liter Line 4511  we set the flag only if there is a liter
4511            {            {
4512            /* 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
4513            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
4514            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4515              integer type when available, otherwise double. */
4516    
4517            if (lengthptr != NULL)            if (lengthptr != NULL)
4518              {              {
4519              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4520              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4521                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4522                        (INT64_OR_DOUBLE)INT_MAX ||
4523                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4524                {                {
4525                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 4097  we set the flag only if there is a liter Line 4565  we set the flag only if there is a liter
4565          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
4566          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
4567          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
4568          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4569            a 64-bit integer type when available, otherwise double. */
4570    
4571          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4572            {            {
4573            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4574                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4575            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4576                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4577                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4578                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4579              {              {
4580              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 4130  we set the flag only if there is a liter Line 4599  we set the flag only if there is a liter
4599              {              {
4600              int offset;              int offset;
4601              *code++ = OP_BRA;              *code++ = OP_BRA;
4602              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4603              bralink = code;              bralink = code;
4604              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4605              }              }
# Line 4151  we set the flag only if there is a liter Line 4620  we set the flag only if there is a liter
4620          while (bralink != NULL)          while (bralink != NULL)
4621            {            {
4622            int oldlinkoffset;            int oldlinkoffset;
4623            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4624            uschar *bra = code - offset;            uschar *bra = code - offset;
4625            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4626            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 4651  we set the flag only if there is a liter
4651            uschar *scode = bracode;            uschar *scode = bracode;
4652            do            do
4653              {              {
4654              if (could_be_empty_branch(scode, ketcode, utf8))              if (could_be_empty_branch(scode, ketcode, utf8, cd))
4655                {                {
4656                *bracode += OP_SBRA - OP_BRA;                *bracode += OP_SBRA - OP_BRA;
4657                break;                break;
# Line 4225  we set the flag only if there is a liter Line 4694  we set the flag only if there is a liter
4694      if (possessive_quantifier)      if (possessive_quantifier)
4695        {        {
4696        int len;        int len;
4697        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4698            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4699          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += _pcre_OP_lengths[*tempcode] +
4700            ((*tempcode == OP_TYPEEXACT &&            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4701               (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);  
4702        len = code - tempcode;        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4703            {
4704            tempcode += _pcre_OP_lengths[*tempcode];
4705    #ifdef SUPPORT_UTF8
4706            if (utf8 && tempcode[-1] >= 0xc0)
4707              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4708    #endif
4709            }
4710    
4711          len = (int)(code - tempcode);
4712        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4713          {          {
4714          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 4726  we set the flag only if there is a liter
4726          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4727          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4728    
4729            /* Because we are moving code along, we must ensure that any
4730            pending recursive references are updated. */
4731    
4732          default:          default:
4733            *code = OP_END;
4734            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4735          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4736          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4737          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4284  we set the flag only if there is a liter Line 4767  we set the flag only if there is a liter
4767    
4768      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4769    
4770      if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
4771             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
4772        {        {
4773        int i, namelen;        int i, namelen;
4774          int arglen = 0;
4775        const char *vn = verbnames;        const char *vn = verbnames;
4776        const uschar *name = ++ptr;        const uschar *name = ptr + 1;
4777          const uschar *arg = NULL;
4778        previous = NULL;        previous = NULL;
4779        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4780          namelen = (int)(ptr - name);
4781    
4782        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
4783          {          {
4784          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
4785          goto FAILED;          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
4786              || *ptr == '_') ptr++;
4787            arglen = (int)(ptr - arg);
4788          }          }
4789    
4790        if (*ptr != CHAR_RIGHT_PARENTHESIS)        if (*ptr != CHAR_RIGHT_PARENTHESIS)
4791          {          {
4792          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4793          goto FAILED;          goto FAILED;
4794          }          }
4795        namelen = ptr - name;  
4796          /* Scan the table of verb names */
4797    
4798        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
4799          {          {
4800          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4801              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4802            {            {
4803            *code = verbs[i].op;            /* Check for open captures before ACCEPT */
4804            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;  
4805            break;            if (verbs[i].op == OP_ACCEPT)
4806                {
4807                open_capitem *oc;
4808                cd->had_accept = TRUE;
4809                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4810                  {
4811                  *code++ = OP_CLOSE;
4812                  PUT2INC(code, 0, oc->number);
4813                  }
4814                }
4815    
4816              /* Handle the cases with/without an argument */
4817    
4818              if (arglen == 0)
4819                {
4820                if (verbs[i].op < 0)   /* Argument is mandatory */
4821                  {
4822                  *errorcodeptr = ERR66;
4823                  goto FAILED;
4824                  }
4825                *code = verbs[i].op;
4826                if (*code++ == OP_THEN)
4827                  {
4828                  PUT(code, 0, code - bcptr->current_branch - 1);
4829                  code += LINK_SIZE;
4830                  }
4831                }
4832    
4833              else
4834                {
4835                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
4836                  {
4837                  *errorcodeptr = ERR59;
4838                  goto FAILED;
4839                  }
4840                *code = verbs[i].op_arg;
4841                if (*code++ == OP_THEN_ARG)
4842                  {
4843                  PUT(code, 0, code - bcptr->current_branch - 1);
4844                  code += LINK_SIZE;
4845                  }
4846                *code++ = arglen;
4847                memcpy(code, arg, arglen);
4848                code += arglen;
4849                *code++ = 0;
4850                }
4851    
4852              break;  /* Found verb, exit loop */
4853            }            }
4854    
4855          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
4856          }          }
4857        if (i < verbcount) continue;  
4858        *errorcodeptr = ERR60;        if (i < verbcount) continue;    /* Successfully handled a verb */
4859          *errorcodeptr = ERR60;          /* Verb not recognized */
4860        goto FAILED;        goto FAILED;
4861        }        }
4862    
# Line 4433  we set the flag only if there is a liter Line 4975  we set the flag only if there is a liter
4975                recno * 10 + *ptr - CHAR_0 : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4976            ptr++;            ptr++;
4977            }            }
4978          namelen = ptr - name;          namelen = (int)(ptr - name);
4979    
4980          if ((terminator > 0 && *ptr++ != terminator) ||          if ((terminator > 0 && *ptr++ != terminator) ||
4981              *ptr++ != CHAR_RIGHT_PARENTHESIS)              *ptr++ != CHAR_RIGHT_PARENTHESIS)
# Line 4470  we set the flag only if there is a liter Line 5012  we set the flag only if there is a liter
5012            }            }
5013    
5014          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
5015          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
5016            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
5017            except they record that the reference was originally to a name. The
5018            information is used to check duplicate names. */
5019    
5020          slot = cd->name_table;          slot = cd->name_table;
5021          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 5030  we set the flag only if there is a liter
5030            {            {
5031            recno = GET2(slot, 0);            recno = GET2(slot, 0);
5032            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
5033              code[1+LINK_SIZE]++;
5034            }            }
5035    
5036          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5037    
5038          else if ((i = find_parens(ptr, cd, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5039                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
5040            {            {
5041            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5042              code[1+LINK_SIZE]++;
5043            }            }
5044    
5045          /* 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 5171  we set the flag only if there is a liter
5171              goto FAILED;              goto FAILED;
5172              }              }
5173            *code++ = n;            *code++ = n;
5174            PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */            PUT(code, 0, (int)(ptr - cd->start_pattern + 1)); /* Pattern offset */
5175            PUT(code, LINK_SIZE, 0);                    /* Default length */            PUT(code, LINK_SIZE, 0);                          /* Default length */
5176            code += 2 * LINK_SIZE;            code += 2 * LINK_SIZE;
5177            }            }
5178          previous = NULL;          previous = NULL;
# Line 4658  we set the flag only if there is a liter Line 5205  we set the flag only if there is a liter
5205            name = ++ptr;            name = ++ptr;
5206    
5207            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5208            namelen = ptr - name;            namelen = (int)(ptr - name);
5209    
5210            /* In the pre-compile phase, just do a syntax check. */            /* In the pre-compile phase, just do a syntax check. */
5211    
# Line 4685  we set the flag only if there is a liter Line 5232  we set the flag only if there is a liter
5232                }                }
5233              }              }
5234    
5235            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
5236              alphabetical order. Duplicate names for different numbers are
5237              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
5238              number are always OK. (An existing number can be re-used if (?|
5239              appears in the pattern.) In either event, a duplicate name results in
5240              a duplicate entry in the table, even if the number is the same. This
5241              is because the number of names, and hence the table size, is computed
5242              in the pre-compile, and it affects various numbers and pointers which
5243              would all have to be modified, and the compiled code moved down, if
5244              duplicates with the same number were omitted from the table. This
5245              doesn't seem worth the hassle. However, *different* names for the
5246              same number are not permitted. */
5247    
5248            else            else
5249              {              {
5250                BOOL dupname = FALSE;
5251              slot = cd->name_table;              slot = cd->name_table;
5252    
5253              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
5254                {                {
5255                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 5257  we set the flag only if there is a liter
5257                  {                  {
5258                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
5259                    {                    {
5260                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
5261                          (options & PCRE_DUPNAMES) == 0)
5262                      {                      {
5263                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
5264                      goto FAILED;                      goto FAILED;
5265                      }                      }
5266                      else dupname = TRUE;
5267                    }                    }
5268                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
5269                  }                  }
5270    
5271                  /* Make space in the table and break the loop for an earlier
5272                  name. For a duplicate or later name, carry on. We do this for
5273                  duplicates so that in the simple case (when ?(| is not used) they
5274                  are in order of their numbers. */
5275    
5276                if (crc < 0)                if (crc < 0)
5277                  {                  {
5278                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
5279                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
5280                  break;                  break;
5281                  }                  }
5282    
5283                  /* Continue the loop for a later or duplicate name */
5284    
5285                slot += cd->name_entry_size;                slot += cd->name_entry_size;
5286                }                }
5287    
5288                /* For non-duplicate names, check for a duplicate number before
5289                adding the new name. */
5290    
5291                if (!dupname)
5292                  {
5293                  uschar *cslot = cd->name_table;
5294                  for (i = 0; i < cd->names_found; i++)
5295                    {
5296                    if (cslot != slot)
5297                      {
5298                      if (GET2(cslot, 0) == cd->bracount + 1)
5299                        {
5300                        *errorcodeptr = ERR65;
5301                        goto FAILED;
5302                        }
5303                      }
5304                    else i--;
5305                    cslot += cd->name_entry_size;
5306                    }
5307                  }
5308    
5309              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
5310              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
5311              slot[2+namelen] = 0;              slot[2+namelen] = 0;
5312              }              }
5313            }            }
5314    
5315          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
5316            encountered. */
5317    
         ptr++;                    /* Move past > or ' */  
5318          cd->names_found++;          cd->names_found++;
5319            ptr++;                    /* Move past > or ' */
5320          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
5321    
5322    
# Line 4742  we set the flag only if there is a liter Line 5335  we set the flag only if there is a liter
5335          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
5336          name = ++ptr;          name = ++ptr;
5337          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5338          namelen = ptr - name;          namelen = (int)(ptr - name);
5339    
5340          /* 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
5341          reference number. */          reference number. */
# Line 4788  we set the flag only if there is a liter Line 5381  we set the flag only if there is a liter
5381              recno = GET2(slot, 0);              recno = GET2(slot, 0);
5382              }              }
5383            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5384                      find_parens(ptr, cd, name, namelen,                      find_parens(cd, name, namelen,
5385                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
5386              {              {
5387              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 4892  we set the flag only if there is a liter Line 5485  we set the flag only if there is a liter
5485            if (lengthptr == NULL)            if (lengthptr == NULL)
5486              {              {
5487              *code = OP_END;              *code = OP_END;
5488              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);              if (recno != 0)
5489                  called = _pcre_find_bracket(cd->start_code, utf8, recno);
5490    
5491              /* Forward reference */              /* Forward reference */
5492    
5493              if (called == NULL)              if (called == NULL)
5494                {                {
5495                if (find_parens(ptr, cd, NULL, recno,                if (find_parens(cd, NULL, recno,
5496                      (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
5497                  {                  {
5498                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5499                  goto FAILED;                  goto FAILED;
5500                  }                  }
5501    
5502                  /* Fudge the value of "called" so that when it is inserted as an
5503                  offset below, what it actually inserted is the reference number
5504                  of the group. */
5505    
5506                called = cd->start_code + recno;                called = cd->start_code + recno;
5507                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, (int)(code + 2 + LINK_SIZE - cd->start_code));
5508                }                }
5509    
5510              /* 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 5512  we set the flag only if there is a liter
5512              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. */
5513    
5514              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 &&
5515                       could_be_empty(called, code, bcptr, utf8))                       could_be_empty(called, code, bcptr, utf8, cd))
5516                {                {
5517                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
5518                goto FAILED;                goto FAILED;
# Line 4929  we set the flag only if there is a liter Line 5528  we set the flag only if there is a liter
5528            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
5529    
5530            *code = OP_RECURSE;            *code = OP_RECURSE;
5531            PUT(code, 1, called - cd->start_code);            PUT(code, 1, (int)(called - cd->start_code));
5532            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
5533    
5534            *code = OP_KET;            *code = OP_KET;
# Line 5008  we set the flag only if there is a liter Line 5607  we set the flag only if there is a liter
5607              {              {
5608              cd->external_options = newoptions;              cd->external_options = newoptions;
5609              }              }
5610           else            else
5611              {              {
5612              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))
5613                {                {
# Line 5040  we set the flag only if there is a liter Line 5639  we set the flag only if there is a liter
5639          }     /* End of switch for character following (? */          }     /* End of switch for character following (? */
5640        }       /* End of (? handling */        }       /* End of (? handling */
5641    
5642      /* Opening parenthesis not followed by '?'. If PCRE_NO_AUTO_CAPTURE is set,      /* Opening parenthesis not followed by '*' or '?'. If PCRE_NO_AUTO_CAPTURE
5643      all unadorned brackets become non-capturing and behave like (?:...)      is set, all unadorned brackets become non-capturing and behave like (?:...)
5644      brackets. */      brackets. */
5645    
5646      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 5832  we set the flag only if there is a liter
5832    
5833      /* ===================================================================*/      /* ===================================================================*/
5834      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values
5835      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
5836      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
5837      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
5838      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
5839      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
5840        ever created. */
5841    
5842      case CHAR_BACKSLASH:      case CHAR_BACKSLASH:
5843      tempptr = ptr;      tempptr = ptr;
# Line 5345  we set the flag only if there is a liter Line 5945  we set the flag only if there is a liter
5945    
5946        if (-c >= ESC_REF)        if (-c >= ESC_REF)
5947          {          {
5948            open_capitem *oc;
5949          recno = -c - ESC_REF;          recno = -c - ESC_REF;
5950    
5951          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 5955  we set the flag only if there is a liter
5955          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
5956          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
5957          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
5958    
5959            /* Check to see if this back reference is recursive, that it, it
5960            is inside the group that it references. A flag is set so that the
5961            group can be made atomic. */
5962    
5963            for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5964              {
5965              if (oc->number == recno)
5966                {
5967                oc->flag = TRUE;
5968                break;
5969                }
5970              }
5971          }          }
5972    
5973        /* 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 5997  we set the flag only if there is a liter
5997  #endif  #endif
5998    
5999        /* For the rest (including \X when Unicode properties are supported), we        /* For the rest (including \X when Unicode properties are supported), we
6000        can obtain the OP value by negating the escape value. */        can obtain the OP value by negating the escape value in the default
6001          situation when PCRE_UCP is not set. When it *is* set, we substitute
6002          Unicode property tests. */
6003    
6004        else        else
6005          {          {
6006          previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;  #ifdef SUPPORT_UCP
6007          *code++ = -c;          if (-c >= ESC_DU && -c <= ESC_wu)
6008              {
6009              nestptr = ptr + 1;                   /* Where to resume */
6010              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
6011              }
6012            else
6013    #endif
6014              {
6015              previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;
6016              *code++ = -c;
6017              }
6018          }          }
6019        continue;        continue;
6020        }        }
# Line 5536  uschar *code = *codeptr; Line 6162  uschar *code = *codeptr;
6162  uschar *last_branch = code;  uschar *last_branch = code;
6163  uschar *start_bracket = code;  uschar *start_bracket = code;
6164  uschar *reverse_count = NULL;  uschar *reverse_count = NULL;
6165    open_capitem capitem;
6166    int capnumber = 0;
6167  int firstbyte, reqbyte;  int firstbyte, reqbyte;
6168  int branchfirstbyte, branchreqbyte;  int branchfirstbyte, branchreqbyte;
6169  int length;  int length;
6170  int orig_bracount;  int orig_bracount;
6171  int max_bracount;  int max_bracount;
6172    int old_external_options = cd->external_options;
6173  branch_chain bc;  branch_chain bc;
6174    
6175  bc.outer = bcptr;  bc.outer = bcptr;
6176  bc.current = code;  bc.current_branch = code;
6177    
6178  firstbyte = reqbyte = REQ_UNSET;  firstbyte = reqbyte = REQ_UNSET;
6179    
# Line 5562  the code that abstracts option settings Line 6191  the code that abstracts option settings
6191  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
6192  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. */
6193    
6194    /* If this is a capturing subpattern, add to the chain of open capturing items
6195    so that we can detect them if (*ACCEPT) is encountered. This is also used to
6196    detect groups that contain recursive back references to themselves. */
6197    
6198    if (*code == OP_CBRA)
6199      {
6200      capnumber = GET2(code, 1 + LINK_SIZE);
6201      capitem.number = capnumber;
6202      capitem.next = cd->open_caps;
6203      capitem.flag = FALSE;
6204      cd->open_caps = &capitem;
6205      }
6206    
6207  /* Offset is set zero to mark that this bracket is still open */  /* Offset is set zero to mark that this bracket is still open */
6208    
6209  PUT(code, 1, 0);  PUT(code, 1, 0);
# Line 5606  for (;;) Line 6248  for (;;)
6248      return FALSE;      return FALSE;
6249      }      }
6250    
6251      /* If the external options have changed during this branch, it means that we
6252      are at the top level, and a leading option setting has been encountered. We
6253      need to re-set the original option values to take account of this so that,
6254      during the pre-compile phase, we know to allow for a re-set at the start of
6255      subsequent branches. */
6256    
6257      if (old_external_options != cd->external_options)
6258        oldims = cd->external_options & PCRE_IMS;
6259    
6260    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6261    has fewer than the rest. */    has fewer than the rest. */
6262    
# Line 5656  for (;;) Line 6307  for (;;)
6307    
6308      /* If lookbehind, check that this branch matches a fixed-length string, and      /* If lookbehind, check that this branch matches a fixed-length string, and
6309      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
6310      branch with OP_END. */      branch with OP_END. If the branch contains OP_RECURSE, the result is -3
6311        because there may be forward references that we can't check here. Set a
6312        flag to cause another lookbehind check at the end. Why not do it all at the
6313        end? Because common, erroneous checks are picked up here and the offset of
6314        the problem can be shown. */
6315    
6316      if (lookbehind)      if (lookbehind)
6317        {        {
6318        int fixed_length;        int fixed_length;
6319        *code = OP_END;        *code = OP_END;
6320        fixed_length = find_fixedlength(last_branch, options);        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);
6321        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6322        if (fixed_length < 0)        if (fixed_length == -3)
6323            {
6324            cd->check_lookbehind = TRUE;
6325            }
6326          else if (fixed_length < 0)
6327          {          {
6328          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;
6329          *ptrptr = ptr;          *ptrptr = ptr;
6330          return FALSE;          return FALSE;
6331          }          }
6332        PUT(reverse_count, 0, fixed_length);        else { PUT(reverse_count, 0, fixed_length); }
6333        }        }
6334      }      }
6335    
# Line 5687  for (;;) Line 6346  for (;;)
6346      {      {
6347      if (lengthptr == NULL)      if (lengthptr == NULL)
6348        {        {
6349        int branch_length = code - last_branch;        int branch_length = (int)(code - last_branch);
6350        do        do
6351          {          {
6352          int prev_length = GET(last_branch, 1);          int prev_length = GET(last_branch, 1);
# Line 5701  for (;;) Line 6360  for (;;)
6360      /* Fill in the ket */      /* Fill in the ket */
6361    
6362      *code = OP_KET;      *code = OP_KET;
6363      PUT(code, 1, code - start_bracket);      PUT(code, 1, (int)(code - start_bracket));
6364      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6365    
6366      /* Resetting option if needed */      /* If it was a capturing subpattern, check to see if it contained any
6367        recursive back references. If so, we must wrap it in atomic brackets.
6368        In any event, remove the block from the chain. */
6369    
6370        if (capnumber > 0)
6371          {
6372          if (cd->open_caps->flag)
6373            {
6374            memmove(start_bracket + 1 + LINK_SIZE, start_bracket,
6375              code - start_bracket);
6376            *start_bracket = OP_ONCE;
6377            code += 1 + LINK_SIZE;
6378            PUT(start_bracket, 1, (int)(code - start_bracket));
6379            *code = OP_KET;
6380            PUT(code, 1, (int)(code - start_bracket));
6381            code += 1 + LINK_SIZE;
6382            length += 2 + 2*LINK_SIZE;
6383            }
6384          cd->open_caps = cd->open_caps->next;
6385          }
6386    
6387        /* Reset options if needed. */
6388    
6389      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)
6390        {        {
# Line 5752  for (;;) Line 6432  for (;;)
6432    else    else
6433      {      {
6434      *code = OP_ALT;      *code = OP_ALT;
6435      PUT(code, 1, code - last_branch);      PUT(code, 1, (int)(code - last_branch));
6436      bc.current = last_branch = code;      bc.current_branch = last_branch = code;
6437      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6438      }      }
6439    
# Line 5900  do { Line 6580  do {
6580       switch (*scode)       switch (*scode)
6581         {         {
6582         case OP_CREF:         case OP_CREF:
6583           case OP_NCREF:
6584         case OP_RREF:         case OP_RREF:
6585           case OP_NRREF:
6586         case OP_DEF:         case OP_DEF:
6587         return FALSE;         return FALSE;
6588    
# Line 6069  int length = 1; /* For final END opcode Line 6751  int length = 1; /* For final END opcode
6751  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
6752  int errorcode = 0;  int errorcode = 0;
6753  int skipatstart = 0;  int skipatstart = 0;
 #ifdef SUPPORT_UTF8  
6754  BOOL utf8;  BOOL utf8;
 #endif  
6755  size_t size;  size_t size;
6756  uschar *code;  uschar *code;
6757  const uschar *codestart;  const uschar *codestart;
# Line 6114  if (erroroffset == NULL) Line 6794  if (erroroffset == NULL)
6794    
6795  *erroroffset = 0;  *erroroffset = 0;
6796    
 /* 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;  
   }  
   
6797  /* Set up pointers to the individual character tables */  /* Set up pointers to the individual character tables */
6798    
6799  if (tables == NULL) tables = _pcre_default_tables;  if (tables == NULL) tables = _pcre_default_tables;
# Line 6146  cd->fcc = tables + fcc_offset; Line 6802  cd->fcc = tables + fcc_offset;
6802  cd->cbits = tables + cbits_offset;  cd->cbits = tables + cbits_offset;
6803  cd->ctypes = tables + ctypes_offset;  cd->ctypes = tables + ctypes_offset;
6804    
6805    /* Check that all undefined public option bits are zero */
6806    
6807    if ((options & ~PUBLIC_COMPILE_OPTIONS) != 0)
6808      {
6809      errorcode = ERR17;
6810      goto PCRE_EARLY_ERROR_RETURN;
6811      }
6812    
6813  /* 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
6814  the offset for later. */  the offset for later. */
6815    
# Line 6155  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 6819  while (ptr[skipatstart] == CHAR_LEFT_PAR
6819    int newnl = 0;    int newnl = 0;
6820    int newbsr = 0;    int newbsr = 0;
6821    
6822      if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)
6823        { skipatstart += 7; options |= PCRE_UTF8; continue; }
6824      else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)
6825        { skipatstart += 6; options |= PCRE_UCP; continue; }
6826    
6827    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
6828      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
6829    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 6847  while (ptr[skipatstart] == CHAR_LEFT_PAR
6847    else break;    else break;
6848    }    }
6849    
6850    utf8 = (options & PCRE_UTF8) != 0;
6851    
6852    /* Can't support UTF8 unless PCRE has been compiled to include the code. */
6853    
6854    #ifdef SUPPORT_UTF8
6855    if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
6856         (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)
6857      {
6858      errorcode = ERR44;
6859      goto PCRE_EARLY_ERROR_RETURN2;
6860      }
6861    #else
6862    if (utf8)
6863      {
6864      errorcode = ERR32;
6865      goto PCRE_EARLY_ERROR_RETURN;
6866      }
6867    #endif
6868    
6869    /* Can't support UCP unless PCRE has been compiled to include the code. */
6870    
6871    #ifndef SUPPORT_UCP
6872    if ((options & PCRE_UCP) != 0)
6873      {
6874      errorcode = ERR67;
6875      goto PCRE_EARLY_ERROR_RETURN;
6876      }
6877    #endif
6878    
6879  /* Check validity of \R options. */  /* Check validity of \R options. */
6880    
6881  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 6958  cd->end_pattern = (const uschar *)(patte
6958  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6959  cd->external_options = options;  cd->external_options = options;
6960  cd->external_flags = 0;  cd->external_flags = 0;
6961    cd->open_caps = NULL;
6962    
6963  /* 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
6964  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 7004  regex compiled on a system with 4-byte p
7004  pointers. */  pointers. */
7005    
7006  re->magic_number = MAGIC_NUMBER;  re->magic_number = MAGIC_NUMBER;
7007  re->size = size;  re->size = (int)size;
7008  re->options = cd->external_options;  re->options = cd->external_options;
7009  re->flags = cd->external_flags;  re->flags = cd->external_flags;
7010  re->dummy1 = 0;  re->dummy1 = 0;
# Line 6334  cd->start_code = codestart; Line 7033  cd->start_code = codestart;
7033  cd->hwm = cworkspace;  cd->hwm = cworkspace;
7034  cd->req_varyopt = 0;  cd->req_varyopt = 0;
7035  cd->had_accept = FALSE;  cd->had_accept = FALSE;
7036    cd->check_lookbehind = FALSE;
7037    cd->open_caps = NULL;
7038    
7039  /* Set up a starting, non-extracting bracket, then compile the expression. On  /* Set up a starting, non-extracting bracket, then compile the expression. On
7040  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 7060  if debugging, leave the test till after
7060    
7061  *code++ = OP_END;  *code++ = OP_END;
7062    
7063  #ifndef DEBUG  #ifndef PCRE_DEBUG
7064  if (code - codestart > length) errorcode = ERR23;  if (code - codestart > length) errorcode = ERR23;
7065  #endif  #endif
7066    
# Line 6372  while (errorcode == 0 && cd->hwm > cwork Line 7073  while (errorcode == 0 && cd->hwm > cwork
7073    cd->hwm -= LINK_SIZE;    cd->hwm -= LINK_SIZE;
7074    offset = GET(cd->hwm, 0);    offset = GET(cd->hwm, 0);
7075    recno = GET(codestart, offset);    recno = GET(codestart, offset);
7076    groupptr = find_bracket(codestart, (re->options & PCRE_UTF8) != 0, recno);    groupptr = _pcre_find_bracket(codestart, utf8, recno);
7077    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
7078      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, (int)(groupptr - codestart));
7079    }    }
7080    
7081  /* 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 7083  subpattern. */
7083    
7084  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;
7085    
7086    /* If there were any lookbehind assertions that contained OP_RECURSE
7087    (recursions or subroutine calls), a flag is set for them to be checked here,
7088    because they may contain forward references. Actual recursions can't be fixed
7089    length, but subroutine calls can. It is done like this so that those without
7090    OP_RECURSE that are not fixed length get a diagnosic with a useful offset. The
7091    exceptional ones forgo this. We scan the pattern to check that they are fixed
7092    length, and set their lengths. */
7093    
7094    if (cd->check_lookbehind)
7095      {
7096      uschar *cc = (uschar *)codestart;
7097    
7098      /* Loop, searching for OP_REVERSE items, and process those that do not have
7099      their length set. (Actually, it will also re-process any that have a length
7100      of zero, but that is a pathological case, and it does no harm.) When we find
7101      one, we temporarily terminate the branch it is in while we scan it. */
7102    
7103      for (cc = (uschar *)_pcre_find_bracket(codestart, utf8, -1);
7104           cc != NULL;
7105           cc = (uschar *)_pcre_find_bracket(cc, utf8, -1))
7106        {
7107        if (GET(cc, 1) == 0)
7108          {
7109          int fixed_length;
7110          uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
7111          int end_op = *be;
7112          *be = OP_END;
7113          fixed_length = find_fixedlength(cc, re->options, TRUE, cd);
7114          *be = end_op;
7115          DPRINTF(("fixed length = %d\n", fixed_length));
7116          if (fixed_length < 0)
7117            {
7118            errorcode = (fixed_length == -2)? ERR36 : ERR25;
7119            break;
7120            }
7121          PUT(cc, 1, fixed_length);
7122          }
7123        cc += 1 + LINK_SIZE;
7124        }
7125      }
7126    
7127  /* Failed to compile, or error while post-processing */  /* Failed to compile, or error while post-processing */
7128    
7129  if (errorcode != 0)  if (errorcode != 0)
7130    {    {
7131    (pcre_free)(re);    (pcre_free)(re);
7132    PCRE_EARLY_ERROR_RETURN:    PCRE_EARLY_ERROR_RETURN:
7133    *erroroffset = ptr - (const uschar *)pattern;    *erroroffset = (int)(ptr - (const uschar *)pattern);
7134    PCRE_EARLY_ERROR_RETURN2:    PCRE_EARLY_ERROR_RETURN2:
7135    *errorptr = find_error_text(errorcode);    *errorptr = find_error_text(errorcode);
7136    if (errorcodeptr != NULL) *errorcodeptr = errorcode;    if (errorcodeptr != NULL) *errorcodeptr = errorcode;
# Line 6442  if (reqbyte >= 0 && Line 7184  if (reqbyte >= 0 &&
7184  /* 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
7185  case when building a production library. */  case when building a production library. */
7186    
7187  #ifdef DEBUG  #ifdef PCRE_DEBUG
   
7188  printf("Length = %d top_bracket = %d top_backref = %d\n",  printf("Length = %d top_bracket = %d top_backref = %d\n",
7189    length, re->top_bracket, re->top_backref);    length, re->top_bracket, re->top_backref);
7190    
# Line 6480  if (code - codestart > length) Line 7221  if (code - codestart > length)
7221    if (errorcodeptr != NULL) *errorcodeptr = ERR23;    if (errorcodeptr != NULL) *errorcodeptr = ERR23;
7222    return NULL;    return NULL;
7223    }    }
7224  #endif   /* DEBUG */  #endif   /* PCRE_DEBUG */
7225    
7226  return (pcre *)re;  return (pcre *)re;
7227  }  }

Legend:
Removed from v.406  
changed lines
  Added in v.552

webmaster@exim.org
ViewVC Help
Powered by ViewVC 1.1.12