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

Diff of /code/trunk/pcre_compile.c

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

revision 391 by ph10, Tue Mar 17 21:16:01 2009 UTC revision 503 by ph10, Sun Mar 7 17:35:52 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 100  is invalid. */ Line 101  is invalid. */
101  #ifndef EBCDIC  #ifndef EBCDIC
102    
103  /* This is the "normal" table for ASCII systems or for EBCDIC systems running  /* This is the "normal" table for ASCII systems or for EBCDIC systems running
104  in UTF-8 mode. */  in UTF-8 mode. */
105    
106  static const short int escapes[] = {  static const short int escapes[] = {
107       0,                       0,       0,                       0,
108         0,                       0,
109         0,                       0,
110       0,                       0,       0,                       0,
      0,                       0,  
111       0,                       0,       0,                       0,
      0,                       0,  
112       CHAR_COLON,              CHAR_SEMICOLON,       CHAR_COLON,              CHAR_SEMICOLON,
113       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
114       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
115       CHAR_COMMERCIAL_AT,      -ESC_A,       CHAR_COMMERCIAL_AT,      -ESC_A,
116       -ESC_B,                  -ESC_C,       -ESC_B,                  -ESC_C,
117       -ESC_D,                  -ESC_E,       -ESC_D,                  -ESC_E,
118       0,                       -ESC_G,       0,                       -ESC_G,
119       -ESC_H,                  0,       -ESC_H,                  0,
120       0,                       -ESC_K,       0,                       -ESC_K,
121       0,                       0,       0,                       0,
122       0,                       0,       0,                       0,
123       -ESC_P,                  -ESC_Q,       -ESC_P,                  -ESC_Q,
124       -ESC_R,                  -ESC_S,       -ESC_R,                  -ESC_S,
125       0,                       0,       0,                       0,
126       -ESC_V,                  -ESC_W,       -ESC_V,                  -ESC_W,
127       -ESC_X,                  0,       -ESC_X,                  0,
128       -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,       -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
129       CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,       CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
130       CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,       CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
131       CHAR_GRAVE_ACCENT,       7,       CHAR_GRAVE_ACCENT,       7,
132       -ESC_b,                  0,       -ESC_b,                  0,
133       -ESC_d,                  ESC_e,       -ESC_d,                  ESC_e,
134       ESC_f,                   0,       ESC_f,                   0,
135       -ESC_h,                  0,       -ESC_h,                  0,
136       0,                       -ESC_k,       0,                       -ESC_k,
137       0,                       0,       0,                       0,
138       ESC_n,                   0,       ESC_n,                   0,
139       -ESC_p,                  0,       -ESC_p,                  0,
140       ESC_r,                   -ESC_s,       ESC_r,                   -ESC_s,
141       ESC_tee,                 0,       ESC_tee,                 0,
142       -ESC_v,                  -ESC_w,       -ESC_v,                  -ESC_w,
143       0,                       0,       0,                       0,
144       -ESC_z       -ESC_z
145  };  };
146    
147  #else  #else
148    
149  /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */  /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
150    
# Line 177  static const short int escapes[] = { Line 178  static const short int escapes[] = {
178    
179  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
180  searched linearly. Put all the names into a single string, in order to reduce  searched linearly. Put all the names into a single string, in order to reduce
181  the number of relocations when a shared library is dynamically linked. The  the number of relocations when a shared library is dynamically linked. The
182  string is built from string macros so that it works in UTF-8 mode on EBCDIC  string is built from string macros so that it works in UTF-8 mode on EBCDIC
183  platforms. */  platforms. */
184    
185  typedef struct verbitem {  typedef struct verbitem {
# Line 215  length entry. The first three must be al Line 216  length entry. The first three must be al
216  for handling case independence. */  for handling case independence. */
217    
218  static const char posix_names[] =  static const char posix_names[] =
219    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
220    STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0    STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
221    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
222    STRING_word0  STRING_xdigit;    STRING_word0  STRING_xdigit;
223    
# Line 262  the number of relocations needed when a Line 263  the number of relocations needed when a
263  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
264  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
265  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
266  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
267    
268    Each substring ends with \0 to insert a null character. This includes the final
269    substring, so that the whole string ends with \0\0, which can be detected when
270    counting through. */
271    
272  static const char error_texts[] =  static const char error_texts[] =
273    "no error\0"    "no error\0"
# Line 341  static const char error_texts[] = Line 346  static const char error_texts[] =
346    "number is too big\0"    "number is too big\0"
347    "subpattern name expected\0"    "subpattern name expected\0"
348    "digit expected after (?+\0"    "digit expected after (?+\0"
349    "] is an invalid data character in JavaScript compatibility mode";    "] is an invalid data character in JavaScript compatibility mode\0"
350      /* 65 */
351      "different names for subpatterns of the same number are not allowed\0";
352    
353  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
354  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 360  For convenience, we use the same bit def Line 366  For convenience, we use the same bit def
366    
367  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
368    
369  #ifndef EBCDIC  #ifndef EBCDIC
370    
371  /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in  /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
372  UTF-8 mode. */  UTF-8 mode. */
373    
374  static const unsigned char digitab[] =  static const unsigned char digitab[] =
# Line 400  static const unsigned char digitab[] = Line 406  static const unsigned char digitab[] =
406    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
407    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
408    
409  #else  #else
410    
411  /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */  /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
412    
# Line 500  static const char * Line 506  static const char *
506  find_error_text(int n)  find_error_text(int n)
507  {  {
508  const char *s = error_texts;  const char *s = error_texts;
509  for (; n > 0; n--) while (*s++ != 0) {};  for (; n > 0; n--)
510      {
511      while (*s++ != 0) {};
512      if (*s == 0) return "Error text not found (please report)";
513      }
514  return s;  return s;
515  }  }
516    
# Line 1009  return p; Line 1019  return p;
1019    
1020    
1021  /*************************************************  /*************************************************
1022  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1023  *************************************************/  *************************************************/
1024    
1025  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1026    top-level call starts at the beginning of the pattern. All other calls must
1027    start at a parenthesis. It scans along a pattern's text looking for capturing
1028  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
1029  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
1030  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
1031  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
1032  be terminated by '>' because that is checked in the first pass.  first pass. Recursion is used to keep track of subpatterns that reset the
1033    capturing group numbers - the (?| feature.
1034    
1035  Arguments:  Arguments:
1036    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1037    cd           compile background data    cd           compile background data
1038    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1039    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1040    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1041      count        pointer to the current capturing subpattern number (updated)
1042    
1043  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1044  */  */
1045    
1046  static int  static int
1047  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,
1048    BOOL xmode)    BOOL xmode, int *count)
1049  {  {
1050  const uschar *thisname;  uschar *ptr = *ptrptr;
1051  int count = cd->bracount;  int start_count = *count;
1052    int hwm_count = start_count;
1053    BOOL dup_parens = FALSE;
1054    
1055  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1056    dealing with. The very first call may not start with a parenthesis. */
1057    
1058    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1059    {    {
1060    int term;    if (ptr[1] == CHAR_QUESTION_MARK &&
1061          ptr[2] == CHAR_VERTICAL_LINE)
1062        {
1063        ptr += 3;
1064        dup_parens = TRUE;
1065        }
1066    
1067      /* Handle a normal, unnamed capturing parenthesis */
1068    
1069      else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1070        {
1071        *count += 1;
1072        if (name == NULL && *count == lorn) return *count;
1073        ptr++;
1074        }
1075    
1076      /* Handle a condition. If it is an assertion, just carry on so that it
1077      is processed as normal. If not, skip to the closing parenthesis of the
1078      condition (there can't be any nested parens. */
1079    
1080      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1081        {
1082        ptr += 2;
1083        if (ptr[1] != CHAR_QUESTION_MARK)
1084          {
1085          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1086          if (*ptr != 0) ptr++;
1087          }
1088        }
1089    
1090      /* We have either (? or (* and not a condition */
1091    
1092      else
1093        {
1094        ptr += 2;
1095        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1096    
1097        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1098    
1099        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1100            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1101          {
1102          int term;
1103          const uschar *thisname;
1104          *count += 1;
1105          if (name == NULL && *count == lorn) return *count;
1106          term = *ptr++;
1107          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1108          thisname = ptr;
1109          while (*ptr != term) ptr++;
1110          if (name != NULL && lorn == ptr - thisname &&
1111              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1112            return *count;
1113          term++;
1114          }
1115        }
1116      }
1117    
1118    /* Past any initial parenthesis handling, scan for parentheses or vertical
1119    bars. */
1120    
1121    for (; *ptr != 0; ptr++)
1122      {
1123    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1124    
1125    if (*ptr == CHAR_BACKSLASH)    if (*ptr == CHAR_BACKSLASH)
1126      {      {
1127      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1128      if (*ptr == CHAR_Q) for (;;)      if (*ptr == CHAR_Q) for (;;)
1129        {        {
1130        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1131        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1132        if (*(++ptr) == CHAR_E) break;        if (*(++ptr) == CHAR_E) break;
1133        }        }
1134      continue;      continue;
# Line 1057  for (; *ptr != 0; ptr++) Line 1137  for (; *ptr != 0; ptr++)
1137    /* Skip over character classes; this logic must be similar to the way they    /* Skip over character classes; this logic must be similar to the way they
1138    are handled for real. If the first character is '^', skip it. Also, if the    are handled for real. If the first character is '^', skip it. Also, if the
1139    first few characters (either before or after ^) are \Q\E or \E we skip them    first few characters (either before or after ^) are \Q\E or \E we skip them
1140    too. This makes for compatibility with Perl. Note the use of STR macros to    too. This makes for compatibility with Perl. Note the use of STR macros to
1141    encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */    encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1142    
1143    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
# Line 1065  for (; *ptr != 0; ptr++) Line 1145  for (; *ptr != 0; ptr++)
1145      BOOL negate_class = FALSE;      BOOL negate_class = FALSE;
1146      for (;;)      for (;;)
1147        {        {
1148        int c = *(++ptr);        if (ptr[1] == CHAR_BACKSLASH)
       if (c == CHAR_BACKSLASH)  
1149          {          {
1150          if (ptr[1] == CHAR_E)          if (ptr[2] == CHAR_E)
1151            ptr++;            ptr+= 2;
1152          else if (strncmp((const char *)ptr+1,          else if (strncmp((const char *)ptr+2,
1153                   STR_Q STR_BACKSLASH STR_E, 3) == 0)                   STR_Q STR_BACKSLASH STR_E, 3) == 0)
1154            ptr += 3;            ptr += 4;
1155          else          else
1156            break;            break;
1157          }          }
1158        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)        else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1159            {
1160          negate_class = TRUE;          negate_class = TRUE;
1161            ptr++;
1162            }
1163        else break;        else break;
1164        }        }
1165    
1166      /* If the next character is ']', it is a data character that must be      /* If the next character is ']', it is a data character that must be
1167      skipped, except in JavaScript compatibility mode. */      skipped, except in JavaScript compatibility mode. */
1168    
1169      if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&      if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1170          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1171        ptr++;        ptr++;
1172    
# Line 1093  for (; *ptr != 0; ptr++) Line 1175  for (; *ptr != 0; ptr++)
1175        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1176        if (*ptr == CHAR_BACKSLASH)        if (*ptr == CHAR_BACKSLASH)
1177          {          {
1178          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1179          if (*ptr == CHAR_Q) for (;;)          if (*ptr == CHAR_Q) for (;;)
1180            {            {
1181            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1182            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1183            if (*(++ptr) == CHAR_E) break;            if (*(++ptr) == CHAR_E) break;
1184            }            }
1185          continue;          continue;
# Line 1111  for (; *ptr != 0; ptr++) Line 1193  for (; *ptr != 0; ptr++)
1193    if (xmode && *ptr == CHAR_NUMBER_SIGN)    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1194      {      {
1195      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1196      if (*ptr == 0) return -1;      if (*ptr == 0) goto FAIL_EXIT;
1197      continue;      continue;
1198      }      }
1199    
1200    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1201    
1202    if (*ptr != CHAR_LEFT_PARENTHESIS) continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)  
1203      {      {
1204      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1205      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1206      continue;      if (*ptr == 0) goto FAIL_EXIT;
1207      }      }
1208    
1209    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1210    if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */      {
1211        if (dup_parens && *count < hwm_count) *count = hwm_count;
1212        *ptrptr = ptr;
1213        return -1;
1214        }
1215    
1216    /* We have to disambiguate (?<! and (?<= from (?<name> */    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1217        {
1218        if (*count > hwm_count) hwm_count = *count;
1219        *count = start_count;
1220        }
1221      }
1222    
1223    if ((*ptr != CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_EXCLAMATION_MARK ||  FAIL_EXIT:
1224        ptr[1] == CHAR_EQUALS_SIGN) && *ptr != CHAR_APOSTROPHE)  *ptrptr = ptr;
1225      continue;  return -1;
1226    }
1227    
1228    
1229    
1230    
1231    /*************************************************
1232    *       Find forward referenced subpattern       *
1233    *************************************************/
1234    
1235    count++;  /* This function scans along a pattern's text looking for capturing
1236    subpatterns, and counting them. If it finds a named pattern that matches the
1237    name it is given, it returns its number. Alternatively, if the name is NULL, it
1238    returns when it reaches a given numbered subpattern. This is used for forward
1239    references to subpatterns. We used to be able to start this scan from the
1240    current compiling point, using the current count value from cd->bracount, and
1241    do it all in a single loop, but the addition of the possibility of duplicate
1242    subpattern numbers means that we have to scan from the very start, in order to
1243    take account of such duplicates, and to use a recursive function to keep track
1244    of the different types of group.
1245    
1246    if (name == NULL && count == lorn) return count;  Arguments:
1247    term = *ptr++;    cd           compile background data
1248    if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;    name         name to seek, or NULL if seeking a numbered subpattern
1249    thisname = ptr;    lorn         name length, or subpattern number if name is NULL
1250    while (*ptr != term) ptr++;    xmode        TRUE if we are in /x mode
1251    if (name != NULL && lorn == ptr - thisname &&  
1252        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  Returns:       the number of the found subpattern, or -1 if not found
1253      return count;  */
1254    
1255    static int
1256    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1257    {
1258    uschar *ptr = (uschar *)cd->start_pattern;
1259    int count = 0;
1260    int rc;
1261    
1262    /* If the pattern does not start with an opening parenthesis, the first call
1263    to find_parens_sub() will scan right to the end (if necessary). However, if it
1264    does start with a parenthesis, find_parens_sub() will return when it hits the
1265    matching closing parens. That is why we have to have a loop. */
1266    
1267    for (;;)
1268      {
1269      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1270      if (rc > 0 || *ptr++ == 0) break;
1271    }    }
1272    
1273  return -1;  return rc;
1274  }  }
1275    
1276    
1277    
1278    
1279  /*************************************************  /*************************************************
1280  *      Find first significant op code            *  *      Find first significant op code            *
1281  *************************************************/  *************************************************/
# Line 1200  for (;;) Line 1325  for (;;)
1325    
1326      case OP_CALLOUT:      case OP_CALLOUT:
1327      case OP_CREF:      case OP_CREF:
1328        case OP_NCREF:
1329      case OP_RREF:      case OP_RREF:
1330        case OP_NRREF:
1331      case OP_DEF:      case OP_DEF:
1332      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1333      break;      break;
# Line 1216  for (;;) Line 1343  for (;;)
1343    
1344    
1345  /*************************************************  /*************************************************
1346  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1347  *************************************************/  *************************************************/
1348    
1349  /* 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,
1350  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.
1351  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
1352    temporarily terminated with OP_END when this function is called.
1353    
1354    This function is called when a backward assertion is encountered, so that if it
1355    fails, the error message can point to the correct place in the pattern.
1356    However, we cannot do this when the assertion contains subroutine calls,
1357    because they can be forward references. We solve this by remembering this case
1358    and doing the check at the end; a flag specifies which mode we are running in.
1359    
1360  Arguments:  Arguments:
1361    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1362    options  the compiling options    options  the compiling options
1363      atend    TRUE if called when the pattern is complete
1364      cd       the "compile data" structure
1365    
1366  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1367                 or -1 if there is no fixed length,
1368               or -2 if \C was encountered               or -2 if \C was encountered
1369                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1370  */  */
1371    
1372  static int  static int
1373  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1374  {  {
1375  int length = -1;  int length = -1;
1376    
# Line 1245  branch, check the length against that of Line 1383  branch, check the length against that of
1383  for (;;)  for (;;)
1384    {    {
1385    int d;    int d;
1386      uschar *ce, *cs;
1387    register int op = *cc;    register int op = *cc;
1388    switch (op)    switch (op)
1389      {      {
# Line 1252  for (;;) Line 1391  for (;;)
1391      case OP_BRA:      case OP_BRA:
1392      case OP_ONCE:      case OP_ONCE:
1393      case OP_COND:      case OP_COND:
1394      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1395      if (d < 0) return d;      if (d < 0) return d;
1396      branchlength += d;      branchlength += d;
1397      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1275  for (;;) Line 1414  for (;;)
1414      branchlength = 0;      branchlength = 0;
1415      break;      break;
1416    
1417        /* A true recursion implies not fixed length, but a subroutine call may
1418        be OK. If the subroutine is a forward reference, we can't deal with
1419        it until the end of the pattern, so return -3. */
1420    
1421        case OP_RECURSE:
1422        if (!atend) return -3;
1423        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1424        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1425        if (cc > cs && cc < ce) return -1;                /* Recursion */
1426        d = find_fixedlength(cs + 2, options, atend, cd);
1427        if (d < 0) return d;
1428        branchlength += d;
1429        cc += 1 + LINK_SIZE;
1430        break;
1431    
1432      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1433    
1434      case OP_ASSERT:      case OP_ASSERT:
# Line 1288  for (;;) Line 1442  for (;;)
1442    
1443      case OP_REVERSE:      case OP_REVERSE:
1444      case OP_CREF:      case OP_CREF:
1445        case OP_NCREF:
1446      case OP_RREF:      case OP_RREF:
1447        case OP_NRREF:
1448      case OP_DEF:      case OP_DEF:
1449      case OP_OPT:      case OP_OPT:
1450      case OP_CALLOUT:      case OP_CALLOUT:
1451      case OP_SOD:      case OP_SOD:
1452      case OP_SOM:      case OP_SOM:
1453        case OP_SET_SOM:
1454      case OP_EOD:      case OP_EOD:
1455      case OP_EODN:      case OP_EODN:
1456      case OP_CIRC:      case OP_CIRC:
# Line 1311  for (;;) Line 1468  for (;;)
1468      branchlength++;      branchlength++;
1469      cc += 2;      cc += 2;
1470  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1471      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1472        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1473  #endif  #endif
1474      break;      break;
1475    
# Line 1325  for (;;) Line 1480  for (;;)
1480      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1481      cc += 4;      cc += 4;
1482  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1483      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1484        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1485  #endif  #endif
1486      break;      break;
1487    
# Line 1407  for (;;) Line 1560  for (;;)
1560    
1561    
1562  /*************************************************  /*************************************************
1563  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1564  *************************************************/  *************************************************/
1565    
1566  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1567  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1568    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1569    so that it can be called from pcre_study() when finding the minimum matching
1570    length.
1571    
1572  Arguments:  Arguments:
1573    code        points to start of expression    code        points to start of expression
1574    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1575    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1576    
1577  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
1578  */  */
1579    
1580  static const uschar *  const uschar *
1581  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1582  {  {
1583  for (;;)  for (;;)
1584    {    {
# Line 1435  for (;;) Line 1591  for (;;)
1591    
1592    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1593    
1594      /* Handle recursion */
1595    
1596      else if (c == OP_REVERSE)
1597        {
1598        if (number < 0) return (uschar *)code;
1599        code += _pcre_OP_lengths[c];
1600        }
1601    
1602    /* Handle capturing bracket */    /* Handle capturing bracket */
1603    
1604    else if (c == OP_CBRA)    else if (c == OP_CBRA)
# Line 1621  Arguments: Line 1785  Arguments:
1785    code        points to start of search    code        points to start of search
1786    endcode     points to where to stop    endcode     points to where to stop
1787    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1788      cd          contains pointers to tables etc.
1789    
1790  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1791  */  */
1792    
1793  static BOOL  static BOOL
1794  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1795      compile_data *cd)
1796  {  {
1797  register int c;  register int c;
1798  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 1636  for (code = first_significant_code(code Line 1802  for (code = first_significant_code(code
1802    const uschar *ccode;    const uschar *ccode;
1803    
1804    c = *code;    c = *code;
1805    
1806    /* Skip over forward assertions; the other assertions are skipped by    /* Skip over forward assertions; the other assertions are skipped by
1807    first_significant_code() with a TRUE final argument. */    first_significant_code() with a TRUE final argument. */
1808    
# Line 1656  for (code = first_significant_code(code Line 1822  for (code = first_significant_code(code
1822      c = *code;      c = *code;
1823      continue;      continue;
1824      }      }
1825    
1826      /* For a recursion/subroutine call, if its end has been reached, which
1827      implies a subroutine call, we can scan it. */
1828    
1829      if (c == OP_RECURSE)
1830        {
1831        const uschar *scode = cd->start_code + GET(code, 1);
1832        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1833        do
1834          {
1835          if (!could_be_empty_branch(scode, endcode, utf8, cd)) return FALSE;
1836          scode += GET(scode, 1);
1837          }
1838        while (*scode == OP_ALT);
1839        continue;
1840        }
1841    
1842    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
1843    
# Line 1664  for (code = first_significant_code(code Line 1846  for (code = first_significant_code(code
1846      BOOL empty_branch;      BOOL empty_branch;
1847      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1848    
1849      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1850        empty branch, so just skip over the conditional, because it could be empty.
1851        Otherwise, scan the individual branches of the group. */
1852    
1853      empty_branch = FALSE;      if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
     do  
       {  
       if (!empty_branch && could_be_empty_branch(code, endcode, utf8))  
         empty_branch = TRUE;  
1854        code += GET(code, 1);        code += GET(code, 1);
1855        else
1856          {
1857          empty_branch = FALSE;
1858          do
1859            {
1860            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1861              empty_branch = TRUE;
1862            code += GET(code, 1);
1863            }
1864          while (*code == OP_ALT);
1865          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1866        }        }
1867      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
1868      c = *code;      c = *code;
1869      continue;      continue;
1870      }      }
# Line 1792  for (code = first_significant_code(code Line 1982  for (code = first_significant_code(code
1982      case OP_QUERY:      case OP_QUERY:
1983      case OP_MINQUERY:      case OP_MINQUERY:
1984      case OP_POSQUERY:      case OP_POSQUERY:
1985        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
1986        break;
1987    
1988      case OP_UPTO:      case OP_UPTO:
1989      case OP_MINUPTO:      case OP_MINUPTO:
1990      case OP_POSUPTO:      case OP_POSUPTO:
1991      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
1992      break;      break;
1993  #endif  #endif
1994    
1995        /* None of the remaining opcodes are required to match a character. */
1996    
1997        default:
1998        break;
1999      }      }
2000    }    }
2001    
# Line 1820  Arguments: Line 2018  Arguments:
2018    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2019    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2020    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2021      cd          pointers to tables etc
2022    
2023  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2024  */  */
2025    
2026  static BOOL  static BOOL
2027  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2028    BOOL utf8)    BOOL utf8, compile_data *cd)
2029  {  {
2030  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2031    {    {
2032    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2033        return FALSE;
2034    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2035    }    }
2036  return TRUE;  return TRUE;
# Line 2173  if ((options & PCRE_EXTENDED) != 0) Line 2373  if ((options & PCRE_EXTENDED) != 0)
2373    
2374  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2375    
2376  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2377    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)
2378      return FALSE;      return FALSE;
2379    
# Line 2492  BOOL utf8 = FALSE; Line 2692  BOOL utf8 = FALSE;
2692  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2693  #endif  #endif
2694    
2695  #ifdef DEBUG  #ifdef PCRE_DEBUG
2696  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2697  #endif  #endif
2698    
# Line 2551  for (;; ptr++) Line 2751  for (;; ptr++)
2751    
2752    if (lengthptr != NULL)    if (lengthptr != NULL)
2753      {      {
2754  #ifdef DEBUG  #ifdef PCRE_DEBUG
2755      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2756  #endif  #endif
2757      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */
# Line 2639  for (;; ptr++) Line 2839  for (;; ptr++)
2839    /* Fill in length of a previous callout, except when the next thing is    /* Fill in length of a previous callout, except when the next thing is
2840    a quantifier. */    a quantifier. */
2841    
2842    is_quantifier =    is_quantifier =
2843      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2844      (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));      (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2845    
# Line 2759  for (;; ptr++) Line 2959  for (;; ptr++)
2959      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
2960      they are encountered at the top level, so we'll do that too. */      they are encountered at the top level, so we'll do that too. */
2961    
2962      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2963           ptr[1] == CHAR_EQUALS_SIGN) &&           ptr[1] == CHAR_EQUALS_SIGN) &&
2964          check_posix_syntax(ptr, &tempptr))          check_posix_syntax(ptr, &tempptr))
2965        {        {
# Line 2777  for (;; ptr++) Line 2977  for (;; ptr++)
2977        c = *(++ptr);        c = *(++ptr);
2978        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
2979          {          {
2980          if (ptr[1] == CHAR_E)          if (ptr[1] == CHAR_E)
2981            ptr++;            ptr++;
2982          else if (strncmp((const char *)ptr+1,          else if (strncmp((const char *)ptr+1,
2983                            STR_Q STR_BACKSLASH STR_E, 3) == 0)                            STR_Q STR_BACKSLASH STR_E, 3) == 0)
2984            ptr += 3;            ptr += 3;
2985          else          else
2986            break;            break;
2987          }          }
2988        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
# Line 2795  for (;; ptr++) Line 2995  for (;; ptr++)
2995      that. In JS mode, [] must always fail, so generate OP_FAIL, whereas      that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
2996      [^] must match any character, so generate OP_ALLANY. */      [^] must match any character, so generate OP_ALLANY. */
2997    
2998      if (c == CHAR_RIGHT_SQUARE_BRACKET &&      if (c == CHAR_RIGHT_SQUARE_BRACKET &&
2999          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3000        {        {
3001        *code++ = negate_class? OP_ALLANY : OP_FAIL;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
# Line 2877  for (;; ptr++) Line 3077  for (;; ptr++)
3077        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3078    
3079        if (c == CHAR_LEFT_SQUARE_BRACKET &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3080            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3081             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3082          {          {
3083          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
# Line 3227  for (;; ptr++) Line 3427  for (;; ptr++)
3427          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3428            {            {
3429            ptr += 2;            ptr += 2;
3430            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3431              { ptr += 2; continue; }              { ptr += 2; continue; }
3432            inescq = TRUE;            inescq = TRUE;
3433            break;            break;
# Line 3749  we set the flag only if there is a liter Line 3949  we set the flag only if there is a liter
3949    
3950        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
3951    
3952          /*--------------------------------------------------------------------*/
3953          /* This code is obsolete from release 8.00; the restriction was finally
3954          removed: */
3955    
3956        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3957        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3958    
3959        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
3960          /*--------------------------------------------------------------------*/
3961    
3962        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3963    
# Line 3899  we set the flag only if there is a liter Line 4104  we set the flag only if there is a liter
4104          goto END_REPEAT;          goto END_REPEAT;
4105          }          }
4106    
4107          /*--------------------------------------------------------------------*/
4108          /* This code is obsolete from release 8.00; the restriction was finally
4109          removed: */
4110    
4111        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4112        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4113    
4114        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4115          /*--------------------------------------------------------------------*/
4116    
4117        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4118          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 4037  we set the flag only if there is a liter Line 4247  we set the flag only if there is a liter
4247            {            {
4248            /* 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
4249            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
4250            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4251              integer type when available, otherwise double. */
4252    
4253            if (lengthptr != NULL)            if (lengthptr != NULL)
4254              {              {
4255              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4256              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4257                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4258                        (INT64_OR_DOUBLE)INT_MAX ||
4259                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4260                {                {
4261                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 4089  we set the flag only if there is a liter Line 4301  we set the flag only if there is a liter
4301          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
4302          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
4303          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
4304          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4305            a 64-bit integer type when available, otherwise double. */
4306    
4307          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4308            {            {
4309            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4310                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4311            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4312                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4313                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4314                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4315              {              {
4316              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 4174  we set the flag only if there is a liter Line 4387  we set the flag only if there is a liter
4387            uschar *scode = bracode;            uschar *scode = bracode;
4388            do            do
4389              {              {
4390              if (could_be_empty_branch(scode, ketcode, utf8))              if (could_be_empty_branch(scode, ketcode, utf8, cd))
4391                {                {
4392                *bracode += OP_SBRA - OP_BRA;                *bracode += OP_SBRA - OP_BRA;
4393                break;                break;
# Line 4217  we set the flag only if there is a liter Line 4430  we set the flag only if there is a liter
4430      if (possessive_quantifier)      if (possessive_quantifier)
4431        {        {
4432        int len;        int len;
4433        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4434            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4435          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += _pcre_OP_lengths[*tempcode] +
4436            ((*tempcode == OP_TYPEEXACT &&            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4437               (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);  
4438          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4439            {
4440            tempcode += _pcre_OP_lengths[*tempcode];
4441    #ifdef SUPPORT_UTF8
4442            if (utf8 && tempcode[-1] >= 0xc0)
4443              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4444    #endif
4445            }
4446    
4447        len = code - tempcode;        len = code - tempcode;
4448        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4449          {          {
# Line 4240  we set the flag only if there is a liter Line 4462  we set the flag only if there is a liter
4462          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4463          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4464    
4465            /* Because we are moving code along, we must ensure that any
4466            pending recursive references are updated. */
4467    
4468          default:          default:
4469            *code = OP_END;
4470            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4471          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4472          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4473          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4299  we set the flag only if there is a liter Line 4526  we set the flag only if there is a liter
4526          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4527              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4528            {            {
4529            *code = verbs[i].op;            /* Check for open captures before ACCEPT */
4530            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;  
4531              if (verbs[i].op == OP_ACCEPT)
4532                {
4533                open_capitem *oc;
4534                cd->had_accept = TRUE;
4535                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4536                  {
4537                  *code++ = OP_CLOSE;
4538                  PUT2INC(code, 0, oc->number);
4539                  }
4540                }
4541              *code++ = verbs[i].op;
4542            break;            break;
4543            }            }
4544          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
# Line 4427  we set the flag only if there is a liter Line 4665  we set the flag only if there is a liter
4665            }            }
4666          namelen = ptr - name;          namelen = ptr - name;
4667    
4668          if ((terminator > 0 && *ptr++ != terminator) ||          if ((terminator > 0 && *ptr++ != terminator) ||
4669              *ptr++ != CHAR_RIGHT_PARENTHESIS)              *ptr++ != CHAR_RIGHT_PARENTHESIS)
4670            {            {
4671            ptr--;      /* Error offset */            ptr--;      /* Error offset */
# Line 4462  we set the flag only if there is a liter Line 4700  we set the flag only if there is a liter
4700            }            }
4701    
4702          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
4703          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
4704            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
4705            except they record that the reference was originally to a name. The
4706            information is used to check duplicate names. */
4707    
4708          slot = cd->name_table;          slot = cd->name_table;
4709          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 4477  we set the flag only if there is a liter Line 4718  we set the flag only if there is a liter
4718            {            {
4719            recno = GET2(slot, 0);            recno = GET2(slot, 0);
4720            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4721              code[1+LINK_SIZE]++;
4722            }            }
4723    
4724          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4725    
4726          else if ((i = find_parens(ptr, cd, name, namelen,          else if ((i = find_parens(cd, name, namelen,
4727                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4728            {            {
4729            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
4730              code[1+LINK_SIZE]++;
4731            }            }
4732    
4733          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 4626  we set the flag only if there is a liter Line 4869  we set the flag only if there is a liter
4869    
4870          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4871          case CHAR_P:              /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
4872          if (*(++ptr) == CHAR_EQUALS_SIGN ||          if (*(++ptr) == CHAR_EQUALS_SIGN ||
4873              *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */              *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
4874            {            {
4875            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
# Line 4645  we set the flag only if there is a liter Line 4888  we set the flag only if there is a liter
4888          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
4889          case CHAR_APOSTROPHE:          case CHAR_APOSTROPHE:
4890            {            {
4891            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
4892              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
4893            name = ++ptr;            name = ++ptr;
4894    
# Line 4677  we set the flag only if there is a liter Line 4920  we set the flag only if there is a liter
4920                }                }
4921              }              }
4922    
4923            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
4924              alphabetical order. Duplicate names for different numbers are
4925              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
4926              number are always OK. (An existing number can be re-used if (?|
4927              appears in the pattern.) In either event, a duplicate name results in
4928              a duplicate entry in the table, even if the number is the same. This
4929              is because the number of names, and hence the table size, is computed
4930              in the pre-compile, and it affects various numbers and pointers which
4931              would all have to be modified, and the compiled code moved down, if
4932              duplicates with the same number were omitted from the table. This
4933              doesn't seem worth the hassle. However, *different* names for the
4934              same number are not permitted. */
4935    
4936            else            else
4937              {              {
4938                BOOL dupname = FALSE;
4939              slot = cd->name_table;              slot = cd->name_table;
4940    
4941              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
4942                {                {
4943                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 4689  we set the flag only if there is a liter Line 4945  we set the flag only if there is a liter
4945                  {                  {
4946                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
4947                    {                    {
4948                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
4949                          (options & PCRE_DUPNAMES) == 0)
4950                      {                      {
4951                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
4952                      goto FAILED;                      goto FAILED;
4953                      }                      }
4954                      else dupname = TRUE;
4955                    }                    }
4956                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
4957                  }                  }
4958    
4959                  /* Make space in the table and break the loop for an earlier
4960                  name. For a duplicate or later name, carry on. We do this for
4961                  duplicates so that in the simple case (when ?(| is not used) they
4962                  are in order of their numbers. */
4963    
4964                if (crc < 0)                if (crc < 0)
4965                  {                  {
4966                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
4967                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
4968                  break;                  break;
4969                  }                  }
4970    
4971                  /* Continue the loop for a later or duplicate name */
4972    
4973                slot += cd->name_entry_size;                slot += cd->name_entry_size;
4974                }                }
4975    
4976                /* For non-duplicate names, check for a duplicate number before
4977                adding the new name. */
4978    
4979                if (!dupname)
4980                  {
4981                  uschar *cslot = cd->name_table;
4982                  for (i = 0; i < cd->names_found; i++)
4983                    {
4984                    if (cslot != slot)
4985                      {
4986                      if (GET2(cslot, 0) == cd->bracount + 1)
4987                        {
4988                        *errorcodeptr = ERR65;
4989                        goto FAILED;
4990                        }
4991                      }
4992                    else i--;
4993                    cslot += cd->name_entry_size;
4994                    }
4995                  }
4996    
4997              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
4998              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
4999              slot[2+namelen] = 0;              slot[2+namelen] = 0;
5000              }              }
5001            }            }
5002    
5003          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
5004            encountered. */
5005    
         ptr++;                    /* Move past > or ' */  
5006          cd->names_found++;          cd->names_found++;
5007            ptr++;                    /* Move past > or ' */
5008          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
5009    
5010    
# Line 4780  we set the flag only if there is a liter Line 5069  we set the flag only if there is a liter
5069              recno = GET2(slot, 0);              recno = GET2(slot, 0);
5070              }              }
5071            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5072                      find_parens(ptr, cd, name, namelen,                      find_parens(cd, name, namelen,
5073                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
5074              {              {
5075              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 4884  we set the flag only if there is a liter Line 5173  we set the flag only if there is a liter
5173            if (lengthptr == NULL)            if (lengthptr == NULL)
5174              {              {
5175              *code = OP_END;              *code = OP_END;
5176              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);              if (recno != 0)
5177                  called = _pcre_find_bracket(cd->start_code, utf8, recno);
5178    
5179              /* Forward reference */              /* Forward reference */
5180    
5181              if (called == NULL)              if (called == NULL)
5182                {                {
5183                if (find_parens(ptr, cd, NULL, recno,                if (find_parens(cd, NULL, recno,
5184                      (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
5185                  {                  {
5186                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5187                  goto FAILED;                  goto FAILED;
5188                  }                  }
5189    
5190                  /* Fudge the value of "called" so that when it is inserted as an
5191                  offset below, what it actually inserted is the reference number
5192                  of the group. */
5193    
5194                called = cd->start_code + recno;                called = cd->start_code + recno;
5195                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);
5196                }                }
# Line 4905  we set the flag only if there is a liter Line 5200  we set the flag only if there is a liter
5200              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. */
5201    
5202              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 &&
5203                       could_be_empty(called, code, bcptr, utf8))                       could_be_empty(called, code, bcptr, utf8, cd))
5204                {                {
5205                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
5206                goto FAILED;                goto FAILED;
# Line 5000  we set the flag only if there is a liter Line 5295  we set the flag only if there is a liter
5295              {              {
5296              cd->external_options = newoptions;              cd->external_options = newoptions;
5297              }              }
5298           else            else
5299              {              {
5300              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))
5301                {                {
# Line 5240  we set the flag only if there is a liter Line 5535  we set the flag only if there is a liter
5535        {        {
5536        if (-c == ESC_Q)            /* Handle start of quoted string */        if (-c == ESC_Q)            /* Handle start of quoted string */
5537          {          {
5538          if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)          if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5539            ptr += 2;               /* avoid empty string */            ptr += 2;               /* avoid empty string */
5540              else inescq = TRUE;              else inescq = TRUE;
5541          continue;          continue;
# Line 5270  we set the flag only if there is a liter Line 5565  we set the flag only if there is a liter
5565          {          {
5566          const uschar *p;          const uschar *p;
5567          save_hwm = cd->hwm;   /* Normally this is set when '(' is read */          save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5568          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5569            CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;            CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5570    
5571          /* These two statements stop the compiler for warning about possibly          /* These two statements stop the compiler for warning about possibly
# Line 5321  we set the flag only if there is a liter Line 5616  we set the flag only if there is a liter
5616        /* \k<name> or \k'name' is a back reference by name (Perl syntax).        /* \k<name> or \k'name' is a back reference by name (Perl syntax).
5617        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax) */
5618    
5619        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||
5620            ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))            ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))
5621          {          {
5622          is_recurse = FALSE;          is_recurse = FALSE;
5623          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5624            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
5625            CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;            CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
5626          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
5627          }          }
# Line 5337  we set the flag only if there is a liter Line 5632  we set the flag only if there is a liter
5632    
5633        if (-c >= ESC_REF)        if (-c >= ESC_REF)
5634          {          {
5635            open_capitem *oc;
5636          recno = -c - ESC_REF;          recno = -c - ESC_REF;
5637    
5638          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
# Line 5346  we set the flag only if there is a liter Line 5642  we set the flag only if there is a liter
5642          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
5643          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
5644          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
5645    
5646            /* Check to see if this back reference is recursive, that it, it
5647            is inside the group that it references. A flag is set so that the
5648            group can be made atomic. */
5649    
5650            for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5651              {
5652              if (oc->number == recno)
5653                {
5654                oc->flag = TRUE;
5655                break;
5656                }
5657              }
5658          }          }
5659    
5660        /* So are Unicode property matches, if supported. */        /* So are Unicode property matches, if supported. */
# Line 5528  uschar *code = *codeptr; Line 5837  uschar *code = *codeptr;
5837  uschar *last_branch = code;  uschar *last_branch = code;
5838  uschar *start_bracket = code;  uschar *start_bracket = code;
5839  uschar *reverse_count = NULL;  uschar *reverse_count = NULL;
5840    open_capitem capitem;
5841    int capnumber = 0;
5842  int firstbyte, reqbyte;  int firstbyte, reqbyte;
5843  int branchfirstbyte, branchreqbyte;  int branchfirstbyte, branchreqbyte;
5844  int length;  int length;
5845  int orig_bracount;  int orig_bracount;
5846  int max_bracount;  int max_bracount;
5847    int old_external_options = cd->external_options;
5848  branch_chain bc;  branch_chain bc;
5849    
5850  bc.outer = bcptr;  bc.outer = bcptr;
5851  bc.current = code;  bc.current_branch = code;
5852    
5853  firstbyte = reqbyte = REQ_UNSET;  firstbyte = reqbyte = REQ_UNSET;
5854    
# Line 5554  the code that abstracts option settings Line 5866  the code that abstracts option settings
5866  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
5867  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. */
5868    
5869    /* If this is a capturing subpattern, add to the chain of open capturing items
5870    so that we can detect them if (*ACCEPT) is encountered. This is also used to
5871    detect groups that contain recursive back references to themselves. */
5872    
5873    if (*code == OP_CBRA)
5874      {
5875      capnumber = GET2(code, 1 + LINK_SIZE);
5876      capitem.number = capnumber;
5877      capitem.next = cd->open_caps;
5878      capitem.flag = FALSE;
5879      cd->open_caps = &capitem;
5880      }
5881    
5882  /* Offset is set zero to mark that this bracket is still open */  /* Offset is set zero to mark that this bracket is still open */
5883    
5884  PUT(code, 1, 0);  PUT(code, 1, 0);
# Line 5598  for (;;) Line 5923  for (;;)
5923      return FALSE;      return FALSE;
5924      }      }
5925    
5926      /* If the external options have changed during this branch, it means that we
5927      are at the top level, and a leading option setting has been encountered. We
5928      need to re-set the original option values to take account of this so that,
5929      during the pre-compile phase, we know to allow for a re-set at the start of
5930      subsequent branches. */
5931    
5932      if (old_external_options != cd->external_options)
5933        oldims = cd->external_options & PCRE_IMS;
5934    
5935    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
5936    has fewer than the rest. */    has fewer than the rest. */
5937    
# Line 5648  for (;;) Line 5982  for (;;)
5982    
5983      /* If lookbehind, check that this branch matches a fixed-length string, and      /* If lookbehind, check that this branch matches a fixed-length string, and
5984      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
5985      branch with OP_END. */      branch with OP_END. If the branch contains OP_RECURSE, the result is -3
5986        because there may be forward references that we can't check here. Set a
5987        flag to cause another lookbehind check at the end. Why not do it all at the
5988        end? Because common, erroneous checks are picked up here and the offset of
5989        the problem can be shown. */
5990    
5991      if (lookbehind)      if (lookbehind)
5992        {        {
5993        int fixed_length;        int fixed_length;
5994        *code = OP_END;        *code = OP_END;
5995        fixed_length = find_fixedlength(last_branch, options);        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);
5996        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
5997        if (fixed_length < 0)        if (fixed_length == -3)
5998            {
5999            cd->check_lookbehind = TRUE;
6000            }
6001          else if (fixed_length < 0)
6002          {          {
6003          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;
6004          *ptrptr = ptr;          *ptrptr = ptr;
6005          return FALSE;          return FALSE;
6006          }          }
6007        PUT(reverse_count, 0, fixed_length);        else { PUT(reverse_count, 0, fixed_length); }
6008        }        }
6009      }      }
6010    
# Line 5696  for (;;) Line 6038  for (;;)
6038      PUT(code, 1, code - start_bracket);      PUT(code, 1, code - start_bracket);
6039      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6040    
6041      /* Resetting option if needed */      /* If it was a capturing subpattern, check to see if it contained any
6042        recursive back references. If so, we must wrap it in atomic brackets.
6043        In any event, remove the block from the chain. */
6044    
6045        if (capnumber > 0)
6046          {
6047          if (cd->open_caps->flag)
6048            {
6049            memmove(start_bracket + 1 + LINK_SIZE, start_bracket,
6050              code - start_bracket);
6051            *start_bracket = OP_ONCE;
6052            code += 1 + LINK_SIZE;
6053            PUT(start_bracket, 1, code - start_bracket);
6054            *code = OP_KET;
6055            PUT(code, 1, code - start_bracket);
6056            code += 1 + LINK_SIZE;
6057            length += 2 + 2*LINK_SIZE;
6058            }
6059          cd->open_caps = cd->open_caps->next;
6060          }
6061    
6062        /* Reset options if needed. */
6063    
6064      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)
6065        {        {
# Line 5745  for (;;) Line 6108  for (;;)
6108      {      {
6109      *code = OP_ALT;      *code = OP_ALT;
6110      PUT(code, 1, code - last_branch);      PUT(code, 1, code - last_branch);
6111      bc.current = last_branch = code;      bc.current_branch = last_branch = code;
6112      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6113      }      }
6114    
# Line 5879  do { Line 6242  do {
6242     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6243       NULL, 0, FALSE);       NULL, 0, FALSE);
6244     register int op = *scode;     register int op = *scode;
6245    
6246     /* If we are at the start of a conditional assertion group, *both* the     /* If we are at the start of a conditional assertion group, *both* the
6247     conditional assertion *and* what follows the condition must satisfy the test     conditional assertion *and* what follows the condition must satisfy the test
6248     for start of line. Other kinds of condition fail. Note that there may be an     for start of line. Other kinds of condition fail. Note that there may be an
# Line 5887  do { Line 6250  do {
6250    
6251     if (op == OP_COND)     if (op == OP_COND)
6252       {       {
6253       scode += 1 + LINK_SIZE;       scode += 1 + LINK_SIZE;
6254       if (*scode == OP_CALLOUT) scode += _pcre_OP_lengths[OP_CALLOUT];       if (*scode == OP_CALLOUT) scode += _pcre_OP_lengths[OP_CALLOUT];
6255       switch (*scode)       switch (*scode)
6256         {         {
6257         case OP_CREF:         case OP_CREF:
6258           case OP_NCREF:
6259         case OP_RREF:         case OP_RREF:
6260           case OP_NRREF:
6261         case OP_DEF:         case OP_DEF:
6262         return FALSE;         return FALSE;
6263    
6264         default:     /* Assertion */         default:     /* Assertion */
6265         if (!is_startline(scode, bracket_map, backref_map)) return FALSE;         if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6266         do scode += GET(scode, 1); while (*scode == OP_ALT);         do scode += GET(scode, 1); while (*scode == OP_ALT);
6267         scode += 1 + LINK_SIZE;         scode += 1 + LINK_SIZE;
6268         break;         break;
6269         }         }
6270       scode = first_significant_code(scode, NULL, 0, FALSE);       scode = first_significant_code(scode, NULL, 0, FALSE);
6271       op = *scode;       op = *scode;
6272       }       }
6273    
6274     /* Non-capturing brackets */     /* Non-capturing brackets */
6275    
# Line 5925  do { Line 6290  do {
6290     /* Other brackets */     /* Other brackets */
6291    
6292     else if (op == OP_ASSERT || op == OP_ONCE)     else if (op == OP_ASSERT || op == OP_ONCE)
6293       {       {
6294       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6295       }       }
6296    
6297     /* .* means "start at start or after \n" if it isn't in brackets that     /* .* means "start at start or after \n" if it isn't in brackets that
# Line 6061  int length = 1;  /* For final END opcode Line 6426  int length = 1;  /* For final END opcode
6426  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
6427  int errorcode = 0;  int errorcode = 0;
6428  int skipatstart = 0;  int skipatstart = 0;
6429  #ifdef SUPPORT_UTF8  BOOL utf8 = (options & PCRE_UTF8) != 0;
 BOOL utf8;  
 #endif  
6430  size_t size;  size_t size;
6431  uschar *code;  uschar *code;
6432  const uschar *codestart;  const uschar *codestart;
# Line 6106  if (erroroffset == NULL) Line 6469  if (erroroffset == NULL)
6469    
6470  *erroroffset = 0;  *erroroffset = 0;
6471    
 /* 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;  
   }  
   
6472  /* Set up pointers to the individual character tables */  /* Set up pointers to the individual character tables */
6473    
6474  if (tables == NULL) tables = _pcre_default_tables;  if (tables == NULL) tables = _pcre_default_tables;
# Line 6138  cd->fcc = tables + fcc_offset; Line 6477  cd->fcc = tables + fcc_offset;
6477  cd->cbits = tables + cbits_offset;  cd->cbits = tables + cbits_offset;
6478  cd->ctypes = tables + ctypes_offset;  cd->ctypes = tables + ctypes_offset;
6479    
6480    /* Check that all undefined public option bits are zero */
6481    
6482    if ((options & ~PUBLIC_COMPILE_OPTIONS) != 0)
6483      {
6484      errorcode = ERR17;
6485      goto PCRE_EARLY_ERROR_RETURN;
6486      }
6487    
6488  /* 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
6489  the offset for later. */  the offset for later. */
6490    
6491  while (ptr[skipatstart] == CHAR_LEFT_PARENTHESIS &&  while (ptr[skipatstart] == CHAR_LEFT_PARENTHESIS &&
6492         ptr[skipatstart+1] == CHAR_ASTERISK)         ptr[skipatstart+1] == CHAR_ASTERISK)
6493    {    {
6494    int newnl = 0;    int newnl = 0;
6495    int newbsr = 0;    int newbsr = 0;
6496    
6497      if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)
6498        { skipatstart += 7; options |= PCRE_UTF8; continue; }
6499    
6500    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
6501      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
6502    else if (strncmp((char *)(ptr+skipatstart+2), STRING_LF_RIGHTPAR, 3)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_LF_RIGHTPAR, 3)  == 0)
# Line 6170  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 6520  while (ptr[skipatstart] == CHAR_LEFT_PAR
6520    else break;    else break;
6521    }    }
6522    
6523    /* Can't support UTF8 unless PCRE has been compiled to include the code. */
6524    
6525    #ifdef SUPPORT_UTF8
6526    if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
6527         (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)
6528      {
6529      errorcode = ERR44;
6530      goto PCRE_EARLY_ERROR_RETURN2;
6531      }
6532    #else
6533    if (utf8)
6534      {
6535      errorcode = ERR32;
6536      goto PCRE_EARLY_ERROR_RETURN;
6537      }
6538    #endif
6539    
6540  /* Check validity of \R options. */  /* Check validity of \R options. */
6541    
6542  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
# Line 6252  cd->end_pattern = (const uschar *)(patte Line 6619  cd->end_pattern = (const uschar *)(patte
6619  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6620  cd->external_options = options;  cd->external_options = options;
6621  cd->external_flags = 0;  cd->external_flags = 0;
6622    cd->open_caps = NULL;
6623    
6624  /* 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
6625  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 6326  cd->start_code = codestart; Line 6694  cd->start_code = codestart;
6694  cd->hwm = cworkspace;  cd->hwm = cworkspace;
6695  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6696  cd->had_accept = FALSE;  cd->had_accept = FALSE;
6697    cd->check_lookbehind = FALSE;
6698    cd->open_caps = NULL;
6699    
6700  /* Set up a starting, non-extracting bracket, then compile the expression. On  /* Set up a starting, non-extracting bracket, then compile the expression. On
6701  error, errorcode will be set non-zero, so we don't need to look at the result  error, errorcode will be set non-zero, so we don't need to look at the result
# Line 6351  if debugging, leave the test till after Line 6721  if debugging, leave the test till after
6721    
6722  *code++ = OP_END;  *code++ = OP_END;
6723    
6724  #ifndef DEBUG  #ifndef PCRE_DEBUG
6725  if (code - codestart > length) errorcode = ERR23;  if (code - codestart > length) errorcode = ERR23;
6726  #endif  #endif
6727    
# Line 6364  while (errorcode == 0 && cd->hwm > cwork Line 6734  while (errorcode == 0 && cd->hwm > cwork
6734    cd->hwm -= LINK_SIZE;    cd->hwm -= LINK_SIZE;
6735    offset = GET(cd->hwm, 0);    offset = GET(cd->hwm, 0);
6736    recno = GET(codestart, offset);    recno = GET(codestart, offset);
6737    groupptr = find_bracket(codestart, (re->options & PCRE_UTF8) != 0, recno);    groupptr = _pcre_find_bracket(codestart, utf8, recno);
6738    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
6739      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, groupptr - codestart);
6740    }    }
# Line 6374  subpattern. */ Line 6744  subpattern. */
6744    
6745  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;
6746    
6747    /* If there were any lookbehind assertions that contained OP_RECURSE
6748    (recursions or subroutine calls), a flag is set for them to be checked here,
6749    because they may contain forward references. Actual recursions can't be fixed
6750    length, but subroutine calls can. It is done like this so that those without
6751    OP_RECURSE that are not fixed length get a diagnosic with a useful offset. The
6752    exceptional ones forgo this. We scan the pattern to check that they are fixed
6753    length, and set their lengths. */
6754    
6755    if (cd->check_lookbehind)
6756      {
6757      uschar *cc = (uschar *)codestart;
6758    
6759      /* Loop, searching for OP_REVERSE items, and process those that do not have
6760      their length set. (Actually, it will also re-process any that have a length
6761      of zero, but that is a pathological case, and it does no harm.) When we find
6762      one, we temporarily terminate the branch it is in while we scan it. */
6763    
6764      for (cc = (uschar *)_pcre_find_bracket(codestart, utf8, -1);
6765           cc != NULL;
6766           cc = (uschar *)_pcre_find_bracket(cc, utf8, -1))
6767        {
6768        if (GET(cc, 1) == 0)
6769          {
6770          int fixed_length;
6771          uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
6772          int end_op = *be;
6773          *be = OP_END;
6774          fixed_length = find_fixedlength(cc, re->options, TRUE, cd);
6775          *be = end_op;
6776          DPRINTF(("fixed length = %d\n", fixed_length));
6777          if (fixed_length < 0)
6778            {
6779            errorcode = (fixed_length == -2)? ERR36 : ERR25;
6780            break;
6781            }
6782          PUT(cc, 1, fixed_length);
6783          }
6784        cc += 1 + LINK_SIZE;
6785        }
6786      }
6787    
6788  /* Failed to compile, or error while post-processing */  /* Failed to compile, or error while post-processing */
6789    
6790  if (errorcode != 0)  if (errorcode != 0)
# Line 6434  if (reqbyte >= 0 && Line 6845  if (reqbyte >= 0 &&
6845  /* 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
6846  case when building a production library. */  case when building a production library. */
6847    
6848  #ifdef DEBUG  #ifdef PCRE_DEBUG
   
6849  printf("Length = %d top_bracket = %d top_backref = %d\n",  printf("Length = %d top_bracket = %d top_backref = %d\n",
6850    length, re->top_bracket, re->top_backref);    length, re->top_bracket, re->top_backref);
6851    
# Line 6472  if (code - codestart > length) Line 6882  if (code - codestart > length)
6882    if (errorcodeptr != NULL) *errorcodeptr = ERR23;    if (errorcodeptr != NULL) *errorcodeptr = ERR23;
6883    return NULL;    return NULL;
6884    }    }
6885  #endif   /* DEBUG */  #endif   /* PCRE_DEBUG */
6886    
6887  return (pcre *)re;  return (pcre *)re;
6888  }  }

Legend:
Removed from v.391  
changed lines
  Added in v.503

  ViewVC Help
Powered by ViewVC 1.1.5