/[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 459 by ph10, Sun Oct 4 09:21:39 2009 UTC
# Line 341  static const char error_texts[] = Line 341  static const char error_texts[] =
341    "number is too big\0"    "number is too big\0"
342    "subpattern name expected\0"    "subpattern name expected\0"
343    "digit expected after (?+\0"    "digit expected after (?+\0"
344    "] is an invalid data character in JavaScript compatibility mode";    "] is an invalid data character in JavaScript compatibility mode\0"
345      /* 65 */
346      "different names for subpatterns of the same number are not allowed";
347    
348    
349  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 1009  return p; Line 1011  return p;
1011    
1012    
1013  /*************************************************  /*************************************************
1014  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1015  *************************************************/  *************************************************/
1016    
1017  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1018    top-level call starts at the beginning of the pattern. All other calls must
1019    start at a parenthesis. It scans along a pattern's text looking for capturing
1020  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
1021  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
1022  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
1023  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
1024  be terminated by '>' because that is checked in the first pass.  first pass. Recursion is used to keep track of subpatterns that reset the
1025    capturing group numbers - the (?| feature.
1026    
1027  Arguments:  Arguments:
1028    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1029    cd           compile background data    cd           compile background data
1030    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1031    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1032    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1033      count        pointer to the current capturing subpattern number (updated)
1034    
1035  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1036  */  */
1037    
1038  static int  static int
1039  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,
1040    BOOL xmode)    BOOL xmode, int *count)
1041  {  {
1042  const uschar *thisname;  uschar *ptr = *ptrptr;
1043  int count = cd->bracount;  int start_count = *count;
1044    int hwm_count = start_count;
1045    BOOL dup_parens = FALSE;
1046    
1047  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1048    dealing with. The very first call may not start with a parenthesis. */
1049    
1050    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1051    {    {
1052    int term;    if (ptr[1] == CHAR_QUESTION_MARK &&
1053          ptr[2] == CHAR_VERTICAL_LINE)
1054        {
1055        ptr += 3;
1056        dup_parens = TRUE;
1057        }
1058    
1059      /* Handle a normal, unnamed capturing parenthesis */
1060    
1061      else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1062        {
1063        *count += 1;
1064        if (name == NULL && *count == lorn) return *count;
1065        ptr++;
1066        }
1067    
1068      /* Handle a condition. If it is an assertion, just carry on so that it
1069      is processed as normal. If not, skip to the closing parenthesis of the
1070      condition (there can't be any nested parens. */
1071    
1072      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1073        {
1074        ptr += 2;
1075        if (ptr[1] != CHAR_QUESTION_MARK)
1076          {
1077          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1078          if (*ptr != 0) ptr++;
1079          }
1080        }
1081    
1082      /* We have either (? or (* and not a condition */
1083    
1084      else
1085        {
1086        ptr += 2;
1087        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1088    
1089        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1090    
1091        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1092            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1093          {
1094          int term;
1095          const uschar *thisname;
1096          *count += 1;
1097          if (name == NULL && *count == lorn) return *count;
1098          term = *ptr++;
1099          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1100          thisname = ptr;
1101          while (*ptr != term) ptr++;
1102          if (name != NULL && lorn == ptr - thisname &&
1103              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1104            return *count;
1105          term++;
1106          }
1107        }
1108      }
1109    
1110    /* Past any initial parenthesis handling, scan for parentheses or vertical
1111    bars. */
1112    
1113    for (; *ptr != 0; ptr++)
1114      {
1115    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1116    
1117    if (*ptr == CHAR_BACKSLASH)    if (*ptr == CHAR_BACKSLASH)
1118      {      {
1119      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1120      if (*ptr == CHAR_Q) for (;;)      if (*ptr == CHAR_Q) for (;;)
1121        {        {
1122        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1123        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1124        if (*(++ptr) == CHAR_E) break;        if (*(++ptr) == CHAR_E) break;
1125        }        }
1126      continue;      continue;
# Line 1065  for (; *ptr != 0; ptr++) Line 1137  for (; *ptr != 0; ptr++)
1137      BOOL negate_class = FALSE;      BOOL negate_class = FALSE;
1138      for (;;)      for (;;)
1139        {        {
1140        int c = *(++ptr);        if (ptr[1] == CHAR_BACKSLASH)
       if (c == CHAR_BACKSLASH)  
1141          {          {
1142          if (ptr[1] == CHAR_E)          if (ptr[2] == CHAR_E)
1143            ptr++;            ptr+= 2;
1144          else if (strncmp((const char *)ptr+1,          else if (strncmp((const char *)ptr+2,
1145                   STR_Q STR_BACKSLASH STR_E, 3) == 0)                   STR_Q STR_BACKSLASH STR_E, 3) == 0)
1146            ptr += 3;            ptr += 4;
1147          else          else
1148            break;            break;
1149          }          }
1150        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)        else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1151            {
1152          negate_class = TRUE;          negate_class = TRUE;
1153            ptr++;
1154            }
1155        else break;        else break;
1156        }        }
1157    
# Line 1093  for (; *ptr != 0; ptr++) Line 1167  for (; *ptr != 0; ptr++)
1167        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1168        if (*ptr == CHAR_BACKSLASH)        if (*ptr == CHAR_BACKSLASH)
1169          {          {
1170          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1171          if (*ptr == CHAR_Q) for (;;)          if (*ptr == CHAR_Q) for (;;)
1172            {            {
1173            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1174            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1175            if (*(++ptr) == CHAR_E) break;            if (*(++ptr) == CHAR_E) break;
1176            }            }
1177          continue;          continue;
# Line 1111  for (; *ptr != 0; ptr++) Line 1185  for (; *ptr != 0; ptr++)
1185    if (xmode && *ptr == CHAR_NUMBER_SIGN)    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1186      {      {
1187      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1188      if (*ptr == 0) return -1;      if (*ptr == 0) goto FAIL_EXIT;
1189      continue;      continue;
1190      }      }
1191    
1192    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1193    
1194    if (*ptr != CHAR_LEFT_PARENTHESIS) continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)  
1195      {      {
1196      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1197      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1198      continue;      if (*ptr == 0) goto FAIL_EXIT;
1199        }
1200    
1201      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1202        {
1203        if (dup_parens && *count < hwm_count) *count = hwm_count;
1204        *ptrptr = ptr;
1205        return -1;
1206      }      }
1207    
1208    ptr += 2;    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1209    if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */      {
1210        if (*count > hwm_count) hwm_count = *count;
1211        *count = start_count;
1212        }
1213      }
1214    
1215    /* We have to disambiguate (?<! and (?<= from (?<name> */  FAIL_EXIT:
1216    *ptrptr = ptr;
1217    return -1;
1218    }
1219    
   if ((*ptr != CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_EXCLAMATION_MARK ||  
       ptr[1] == CHAR_EQUALS_SIGN) && *ptr != CHAR_APOSTROPHE)  
     continue;  
1220    
   count++;  
1221    
1222    if (name == NULL && count == lorn) return count;  
1223    term = *ptr++;  /*************************************************
1224    if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;  *       Find forward referenced subpattern       *
1225    thisname = ptr;  *************************************************/
1226    while (*ptr != term) ptr++;  
1227    if (name != NULL && lorn == ptr - thisname &&  /* This function scans along a pattern's text looking for capturing
1228        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  subpatterns, and counting them. If it finds a named pattern that matches the
1229      return count;  name it is given, it returns its number. Alternatively, if the name is NULL, it
1230    returns when it reaches a given numbered subpattern. This is used for forward
1231    references to subpatterns. We used to be able to start this scan from the
1232    current compiling point, using the current count value from cd->bracount, and
1233    do it all in a single loop, but the addition of the possibility of duplicate
1234    subpattern numbers means that we have to scan from the very start, in order to
1235    take account of such duplicates, and to use a recursive function to keep track
1236    of the different types of group.
1237    
1238    Arguments:
1239      cd           compile background data
1240      name         name to seek, or NULL if seeking a numbered subpattern
1241      lorn         name length, or subpattern number if name is NULL
1242      xmode        TRUE if we are in /x mode
1243    
1244    Returns:       the number of the found subpattern, or -1 if not found
1245    */
1246    
1247    static int
1248    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1249    {
1250    uschar *ptr = (uschar *)cd->start_pattern;
1251    int count = 0;
1252    int rc;
1253    
1254    /* If the pattern does not start with an opening parenthesis, the first call
1255    to find_parens_sub() will scan right to the end (if necessary). However, if it
1256    does start with a parenthesis, find_parens_sub() will return when it hits the
1257    matching closing parens. That is why we have to have a loop. */
1258    
1259    for (;;)
1260      {
1261      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1262      if (rc > 0 || *ptr++ == 0) break;
1263    }    }
1264    
1265  return -1;  return rc;
1266  }  }
1267    
1268    
1269    
1270    
1271  /*************************************************  /*************************************************
1272  *      Find first significant op code            *  *      Find first significant op code            *
1273  *************************************************/  *************************************************/
# Line 1200  for (;;) Line 1317  for (;;)
1317    
1318      case OP_CALLOUT:      case OP_CALLOUT:
1319      case OP_CREF:      case OP_CREF:
1320        case OP_NCREF:
1321      case OP_RREF:      case OP_RREF:
1322        case OP_NRREF:
1323      case OP_DEF:      case OP_DEF:
1324      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1325      break;      break;
# Line 1216  for (;;) Line 1335  for (;;)
1335    
1336    
1337  /*************************************************  /*************************************************
1338  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1339  *************************************************/  *************************************************/
1340    
1341  /* 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,
1342  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.
1343  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
1344    temporarily terminated with OP_END when this function is called.
1345    
1346    This function is called when a backward assertion is encountered, so that if it
1347    fails, the error message can point to the correct place in the pattern.
1348    However, we cannot do this when the assertion contains subroutine calls,
1349    because they can be forward references. We solve this by remembering this case
1350    and doing the check at the end; a flag specifies which mode we are running in.
1351    
1352  Arguments:  Arguments:
1353    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1354    options  the compiling options    options  the compiling options
1355      atend    TRUE if called when the pattern is complete
1356      cd       the "compile data" structure
1357    
1358  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1359                 or -1 if there is no fixed length,
1360               or -2 if \C was encountered               or -2 if \C was encountered
1361                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1362  */  */
1363    
1364  static int  static int
1365  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1366  {  {
1367  int length = -1;  int length = -1;
1368    
# Line 1245  branch, check the length against that of Line 1375  branch, check the length against that of
1375  for (;;)  for (;;)
1376    {    {
1377    int d;    int d;
1378      uschar *ce, *cs;
1379    register int op = *cc;    register int op = *cc;
1380    switch (op)    switch (op)
1381      {      {
# Line 1252  for (;;) Line 1383  for (;;)
1383      case OP_BRA:      case OP_BRA:
1384      case OP_ONCE:      case OP_ONCE:
1385      case OP_COND:      case OP_COND:
1386      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1387      if (d < 0) return d;      if (d < 0) return d;
1388      branchlength += d;      branchlength += d;
1389      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1274  for (;;) Line 1405  for (;;)
1405      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1406      branchlength = 0;      branchlength = 0;
1407      break;      break;
1408    
1409        /* A true recursion implies not fixed length, but a subroutine call may
1410        be OK. If the subroutine is a forward reference, we can't deal with
1411        it until the end of the pattern, so return -3. */
1412    
1413        case OP_RECURSE:
1414        if (!atend) return -3;
1415        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1416        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1417        if (cc > cs && cc < ce) return -1;                /* Recursion */
1418        d = find_fixedlength(cs + 2, options, atend, cd);
1419        if (d < 0) return d;
1420        branchlength += d;
1421        cc += 1 + LINK_SIZE;
1422        break;
1423    
1424      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1425    
# Line 1288  for (;;) Line 1434  for (;;)
1434    
1435      case OP_REVERSE:      case OP_REVERSE:
1436      case OP_CREF:      case OP_CREF:
1437        case OP_NCREF:
1438      case OP_RREF:      case OP_RREF:
1439        case OP_NRREF:
1440      case OP_DEF:      case OP_DEF:
1441      case OP_OPT:      case OP_OPT:
1442      case OP_CALLOUT:      case OP_CALLOUT:
# Line 1311  for (;;) Line 1459  for (;;)
1459      branchlength++;      branchlength++;
1460      cc += 2;      cc += 2;
1461  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1462      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1463        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1464  #endif  #endif
1465      break;      break;
1466    
# Line 1325  for (;;) Line 1471  for (;;)
1471      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1472      cc += 4;      cc += 4;
1473  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1474      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1475        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1476  #endif  #endif
1477      break;      break;
1478    
# Line 1407  for (;;) Line 1551  for (;;)
1551    
1552    
1553  /*************************************************  /*************************************************
1554  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1555  *************************************************/  *************************************************/
1556    
1557  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1558  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1559    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1560    so that it can be called from pcre_study() when finding the minimum matching
1561    length.
1562    
1563  Arguments:  Arguments:
1564    code        points to start of expression    code        points to start of expression
1565    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1566    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1567    
1568  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
1569  */  */
1570    
1571  static const uschar *  const uschar *
1572  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1573  {  {
1574  for (;;)  for (;;)
1575    {    {
# Line 1434  for (;;) Line 1581  for (;;)
1581    the table is zero; the actual length is stored in the compiled code. */    the table is zero; the actual length is stored in the compiled code. */
1582    
1583    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1584    
1585      /* Handle recursion */
1586    
1587      else if (c == OP_REVERSE)
1588        {
1589        if (number < 0) return (uschar *)code;
1590        code += _pcre_OP_lengths[c];
1591        }
1592    
1593    /* Handle capturing bracket */    /* Handle capturing bracket */
1594    
# Line 1800  for (code = first_significant_code(code Line 1955  for (code = first_significant_code(code
1955      case OP_QUERY:      case OP_QUERY:
1956      case OP_MINQUERY:      case OP_MINQUERY:
1957      case OP_POSQUERY:      case OP_POSQUERY:
1958        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
1959        break;
1960    
1961      case OP_UPTO:      case OP_UPTO:
1962      case OP_MINUPTO:      case OP_MINUPTO:
1963      case OP_POSUPTO:      case OP_POSUPTO:
1964      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
1965      break;      break;
1966  #endif  #endif
1967      }      }
# Line 3757  we set the flag only if there is a liter Line 3915  we set the flag only if there is a liter
3915    
3916        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
3917    
3918          /*--------------------------------------------------------------------*/
3919          /* This code is obsolete from release 8.00; the restriction was finally
3920          removed: */
3921    
3922        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3923        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3924    
3925        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
3926          /*--------------------------------------------------------------------*/
3927    
3928        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3929    
# Line 3907  we set the flag only if there is a liter Line 4070  we set the flag only if there is a liter
4070          goto END_REPEAT;          goto END_REPEAT;
4071          }          }
4072    
4073          /*--------------------------------------------------------------------*/
4074          /* This code is obsolete from release 8.00; the restriction was finally
4075          removed: */
4076    
4077        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4078        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4079    
4080        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4081          /*--------------------------------------------------------------------*/
4082    
4083        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4084          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 4225  we set the flag only if there is a liter Line 4393  we set the flag only if there is a liter
4393      if (possessive_quantifier)      if (possessive_quantifier)
4394        {        {
4395        int len;        int len;
4396        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4397            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4398          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += _pcre_OP_lengths[*tempcode] +
4399            ((*tempcode == OP_TYPEEXACT &&            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4400               (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);  
4401          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4402            {
4403            tempcode += _pcre_OP_lengths[*tempcode];
4404    #ifdef SUPPORT_UTF8
4405            if (utf8 && tempcode[-1] >= 0xc0)
4406              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4407    #endif
4408            }
4409    
4410        len = code - tempcode;        len = code - tempcode;
4411        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4412          {          {
# Line 4307  we set the flag only if there is a liter Line 4484  we set the flag only if there is a liter
4484          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4485              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4486            {            {
4487            *code = verbs[i].op;            /* Check for open captures before ACCEPT */
4488            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;  
4489              if (verbs[i].op == OP_ACCEPT)
4490                {
4491                open_capitem *oc;
4492                cd->had_accept = TRUE;
4493                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4494                  {
4495                  *code++ = OP_CLOSE;
4496                  PUT2INC(code, 0, oc->number);
4497                  }
4498                }
4499              *code++ = verbs[i].op;
4500            break;            break;
4501            }            }
4502          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
# Line 4470  we set the flag only if there is a liter Line 4658  we set the flag only if there is a liter
4658            }            }
4659    
4660          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
4661          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
4662            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
4663            except they record that the reference was originally to a name. The
4664            information is used to check duplicate names. */
4665    
4666          slot = cd->name_table;          slot = cd->name_table;
4667          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 4676  we set the flag only if there is a liter
4676            {            {
4677            recno = GET2(slot, 0);            recno = GET2(slot, 0);
4678            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4679              code[1+LINK_SIZE]++;
4680            }            }
4681    
4682          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4683    
4684          else if ((i = find_parens(ptr, cd, name, namelen,          else if ((i = find_parens(cd, name, namelen,
4685                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4686            {            {
4687            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
4688              code[1+LINK_SIZE]++;
4689            }            }
4690    
4691          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 4685  we set the flag only if there is a liter Line 4878  we set the flag only if there is a liter
4878                }                }
4879              }              }
4880    
4881            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
4882              alphabetical order. Duplicate names for different numbers are
4883              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
4884              number are always OK. (An existing number can be re-used if (?|
4885              appears in the pattern.) In either event, a duplicate name results in
4886              a duplicate entry in the table, even if the number is the same. This
4887              is because the number of names, and hence the table size, is computed
4888              in the pre-compile, and it affects various numbers and pointers which
4889              would all have to be modified, and the compiled code moved down, if
4890              duplicates with the same number were omitted from the table. This
4891              doesn't seem worth the hassle. However, *different* names for the
4892              same number are not permitted. */
4893    
4894            else            else
4895              {              {
4896                BOOL dupname = FALSE;
4897              slot = cd->name_table;              slot = cd->name_table;
4898    
4899              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
4900                {                {
4901                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 4903  we set the flag only if there is a liter
4903                  {                  {
4904                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
4905                    {                    {
4906                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
4907                          (options & PCRE_DUPNAMES) == 0)
4908                      {                      {
4909                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
4910                      goto FAILED;                      goto FAILED;
4911                      }                      }
4912                      else dupname = TRUE;
4913                    }                    }
4914                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
4915                  }                  }
4916    
4917                  /* Make space in the table and break the loop for an earlier
4918                  name. For a duplicate or later name, carry on. We do this for
4919                  duplicates so that in the simple case (when ?(| is not used) they
4920                  are in order of their numbers. */
4921    
4922                if (crc < 0)                if (crc < 0)
4923                  {                  {
4924                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
4925                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
4926                  break;                  break;
4927                  }                  }
4928    
4929                  /* Continue the loop for a later or duplicate name */
4930    
4931                slot += cd->name_entry_size;                slot += cd->name_entry_size;
4932                }                }
4933    
4934                /* For non-duplicate names, check for a duplicate number before
4935                adding the new name. */
4936    
4937                if (!dupname)
4938                  {
4939                  uschar *cslot = cd->name_table;
4940                  for (i = 0; i < cd->names_found; i++)
4941                    {
4942                    if (cslot != slot)
4943                      {
4944                      if (GET2(cslot, 0) == cd->bracount + 1)
4945                        {
4946                        *errorcodeptr = ERR65;
4947                        goto FAILED;
4948                        }
4949                      }
4950                    else i--;
4951                    cslot += cd->name_entry_size;
4952                    }
4953                  }
4954    
4955              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
4956              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
# Line 4720  we set the flag only if there is a liter Line 4958  we set the flag only if there is a liter
4958              }              }
4959            }            }
4960    
4961          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
4962            encountered. */
4963    
         ptr++;                    /* Move past > or ' */  
4964          cd->names_found++;          cd->names_found++;
4965            ptr++;                    /* Move past > or ' */
4966          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
4967    
4968    
# Line 4788  we set the flag only if there is a liter Line 5027  we set the flag only if there is a liter
5027              recno = GET2(slot, 0);              recno = GET2(slot, 0);
5028              }              }
5029            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5030                      find_parens(ptr, cd, name, namelen,                      find_parens(cd, name, namelen,
5031                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
5032              {              {
5033              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 4892  we set the flag only if there is a liter Line 5131  we set the flag only if there is a liter
5131            if (lengthptr == NULL)            if (lengthptr == NULL)
5132              {              {
5133              *code = OP_END;              *code = OP_END;
5134              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);              if (recno != 0)
5135                  called = _pcre_find_bracket(cd->start_code, utf8, recno);
5136    
5137              /* Forward reference */              /* Forward reference */
5138    
5139              if (called == NULL)              if (called == NULL)
5140                {                {
5141                if (find_parens(ptr, cd, NULL, recno,                if (find_parens(cd, NULL, recno,
5142                      (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
5143                  {                  {
5144                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
# Line 5536  uschar *code = *codeptr; Line 5776  uschar *code = *codeptr;
5776  uschar *last_branch = code;  uschar *last_branch = code;
5777  uschar *start_bracket = code;  uschar *start_bracket = code;
5778  uschar *reverse_count = NULL;  uschar *reverse_count = NULL;
5779    open_capitem capitem;
5780    int capnumber = 0;
5781  int firstbyte, reqbyte;  int firstbyte, reqbyte;
5782  int branchfirstbyte, branchreqbyte;  int branchfirstbyte, branchreqbyte;
5783  int length;  int length;
# Line 5562  the code that abstracts option settings Line 5804  the code that abstracts option settings
5804  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
5805  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. */
5806    
5807    /* If this is a capturing subpattern, add to the chain of open capturing items
5808    so that we can detect them if (*ACCEPT) is encountered. */
5809    
5810    if (*code == OP_CBRA)
5811      {
5812      capnumber = GET2(code, 1 + LINK_SIZE);
5813      capitem.number = capnumber;
5814      capitem.next = cd->open_caps;
5815      cd->open_caps = &capitem;
5816      }
5817    
5818  /* Offset is set zero to mark that this bracket is still open */  /* Offset is set zero to mark that this bracket is still open */
5819    
5820  PUT(code, 1, 0);  PUT(code, 1, 0);
# Line 5656  for (;;) Line 5909  for (;;)
5909    
5910      /* If lookbehind, check that this branch matches a fixed-length string, and      /* If lookbehind, check that this branch matches a fixed-length string, and
5911      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
5912      branch with OP_END. */      branch with OP_END. If the branch contains OP_RECURSE, the result is -3
5913        because there may be forward references that we can't check here. Set a
5914        flag to cause another lookbehind check at the end. Why not do it all at the
5915        end? Because common, erroneous checks are picked up here and the offset of
5916        the problem can be shown. */
5917    
5918      if (lookbehind)      if (lookbehind)
5919        {        {
5920        int fixed_length;        int fixed_length;
5921        *code = OP_END;        *code = OP_END;
5922        fixed_length = find_fixedlength(last_branch, options);        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);
5923        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
5924        if (fixed_length < 0)        if (fixed_length == -3)
5925            {
5926            cd->check_lookbehind = TRUE;
5927            }
5928          else if (fixed_length < 0)
5929          {          {
5930          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;
5931          *ptrptr = ptr;          *ptrptr = ptr;
5932          return FALSE;          return FALSE;
5933          }          }
5934        PUT(reverse_count, 0, fixed_length);        else { PUT(reverse_count, 0, fixed_length); }
5935        }        }
5936      }      }
5937    
# Line 5697  for (;;) Line 5958  for (;;)
5958          }          }
5959        while (branch_length > 0);        while (branch_length > 0);
5960        }        }
5961    
5962        /* If it was a capturing subpattern, remove it from the chain. */
5963    
5964        if (capnumber > 0) cd->open_caps = cd->open_caps->next;
5965    
5966      /* Fill in the ket */      /* Fill in the ket */
5967    
# Line 5900  do { Line 6165  do {
6165       switch (*scode)       switch (*scode)
6166         {         {
6167         case OP_CREF:         case OP_CREF:
6168           case OP_NCREF:
6169         case OP_RREF:         case OP_RREF:
6170           case OP_NRREF:
6171         case OP_DEF:         case OP_DEF:
6172         return FALSE;         return FALSE;
6173    
# Line 6069  int length = 1;  /* For final END opcode Line 6336  int length = 1;  /* For final END opcode
6336  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
6337  int errorcode = 0;  int errorcode = 0;
6338  int skipatstart = 0;  int skipatstart = 0;
6339  #ifdef SUPPORT_UTF8  BOOL utf8 = (options & PCRE_UTF8) != 0;
 BOOL utf8;  
 #endif  
6340  size_t size;  size_t size;
6341  uschar *code;  uschar *code;
6342  const uschar *codestart;  const uschar *codestart;
# Line 6114  if (erroroffset == NULL) Line 6379  if (erroroffset == NULL)
6379    
6380  *erroroffset = 0;  *erroroffset = 0;
6381    
 /* 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;  
   }  
   
6382  /* Set up pointers to the individual character tables */  /* Set up pointers to the individual character tables */
6383    
6384  if (tables == NULL) tables = _pcre_default_tables;  if (tables == NULL) tables = _pcre_default_tables;
# Line 6146  cd->fcc = tables + fcc_offset; Line 6387  cd->fcc = tables + fcc_offset;
6387  cd->cbits = tables + cbits_offset;  cd->cbits = tables + cbits_offset;
6388  cd->ctypes = tables + ctypes_offset;  cd->ctypes = tables + ctypes_offset;
6389    
6390    /* Check that all undefined public option bits are zero */
6391    
6392    if ((options & ~PUBLIC_COMPILE_OPTIONS) != 0)
6393      {
6394      errorcode = ERR17;
6395      goto PCRE_EARLY_ERROR_RETURN;
6396      }
6397    
6398  /* 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
6399  the offset for later. */  the offset for later. */
6400    
# Line 6155  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 6404  while (ptr[skipatstart] == CHAR_LEFT_PAR
6404    int newnl = 0;    int newnl = 0;
6405    int newbsr = 0;    int newbsr = 0;
6406    
6407      if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)
6408        { skipatstart += 7; options |= PCRE_UTF8; continue; }
6409    
6410    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
6411      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
6412    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 6430  while (ptr[skipatstart] == CHAR_LEFT_PAR
6430    else break;    else break;
6431    }    }
6432    
6433    /* Can't support UTF8 unless PCRE has been compiled to include the code. */
6434    
6435    #ifdef SUPPORT_UTF8
6436    if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
6437         (*erroroffset = _pcre_valid_utf8((uschar *)pattern, -1)) >= 0)
6438      {
6439      errorcode = ERR44;
6440      goto PCRE_EARLY_ERROR_RETURN2;
6441      }
6442    #else
6443    if (utf8)
6444      {
6445      errorcode = ERR32;
6446      goto PCRE_EARLY_ERROR_RETURN;
6447      }
6448    #endif
6449    
6450  /* Check validity of \R options. */  /* Check validity of \R options. */
6451    
6452  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 6529  cd->end_pattern = (const uschar *)(patte
6529  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6530  cd->external_options = options;  cd->external_options = options;
6531  cd->external_flags = 0;  cd->external_flags = 0;
6532    cd->open_caps = NULL;
6533    
6534  /* 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
6535  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 6334  cd->start_code = codestart; Line 6604  cd->start_code = codestart;
6604  cd->hwm = cworkspace;  cd->hwm = cworkspace;
6605  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6606  cd->had_accept = FALSE;  cd->had_accept = FALSE;
6607    cd->check_lookbehind = FALSE;
6608    cd->open_caps = NULL;
6609    
6610  /* Set up a starting, non-extracting bracket, then compile the expression. On  /* Set up a starting, non-extracting bracket, then compile the expression. On
6611  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 6372  while (errorcode == 0 && cd->hwm > cwork Line 6644  while (errorcode == 0 && cd->hwm > cwork
6644    cd->hwm -= LINK_SIZE;    cd->hwm -= LINK_SIZE;
6645    offset = GET(cd->hwm, 0);    offset = GET(cd->hwm, 0);
6646    recno = GET(codestart, offset);    recno = GET(codestart, offset);
6647    groupptr = find_bracket(codestart, (re->options & PCRE_UTF8) != 0, recno);    groupptr = _pcre_find_bracket(codestart, utf8, recno);
6648    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
6649      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, groupptr - codestart);
6650    }    }
# Line 6382  subpattern. */ Line 6654  subpattern. */
6654    
6655  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;
6656    
6657    /* If there were any lookbehind assertions that contained OP_RECURSE
6658    (recursions or subroutine calls), a flag is set for them to be checked here,
6659    because they may contain forward references. Actual recursions can't be fixed
6660    length, but subroutine calls can. It is done like this so that those without
6661    OP_RECURSE that are not fixed length get a diagnosic with a useful offset. The
6662    exceptional ones forgo this. We scan the pattern to check that they are fixed
6663    length, and set their lengths. */
6664    
6665    if (cd->check_lookbehind)
6666      {
6667      uschar *cc = (uschar *)codestart;
6668    
6669      /* Loop, searching for OP_REVERSE items, and process those that do not have
6670      their length set. (Actually, it will also re-process any that have a length
6671      of zero, but that is a pathological case, and it does no harm.) When we find
6672      one, we temporarily terminate the branch it is in while we scan it. */
6673    
6674      for (cc = (uschar *)_pcre_find_bracket(codestart, utf8, -1);
6675           cc != NULL;
6676           cc = (uschar *)_pcre_find_bracket(cc, utf8, -1))
6677        {
6678        if (GET(cc, 1) == 0)
6679          {
6680          int fixed_length;
6681          uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
6682          int end_op = *be;
6683          *be = OP_END;
6684          fixed_length = find_fixedlength(cc, re->options, TRUE, cd);
6685          *be = end_op;
6686          DPRINTF(("fixed length = %d\n", fixed_length));
6687          if (fixed_length < 0)
6688            {
6689            errorcode = (fixed_length == -2)? ERR36 : ERR25;
6690            break;
6691            }
6692          PUT(cc, 1, fixed_length);
6693          }
6694        cc += 1 + LINK_SIZE;
6695        }
6696      }
6697    
6698  /* Failed to compile, or error while post-processing */  /* Failed to compile, or error while post-processing */
6699    
6700  if (errorcode != 0)  if (errorcode != 0)

Legend:
Removed from v.406  
changed lines
  Added in v.459

  ViewVC Help
Powered by ViewVC 1.1.5