/[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 544 by ph10, Tue Jun 15 17:20:55 2010 UTC revision 723 by ph10, Sat Oct 8 15:55:23 2011 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) 1997-2010 University of Cambridge             Copyright (c) 1997-2011 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 393  static const char error_texts[] = Line 393  static const char error_texts[] =
393    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
394    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
395    /* 55 */    /* 55 */
396    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
397    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
398    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
399    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
# Line 408  static const char error_texts[] = Line 408  static const char error_texts[] =
408    "different names for subpatterns of the same number are not allowed\0"    "different names for subpatterns of the same number are not allowed\0"
409    "(*MARK) must have an argument\0"    "(*MARK) must have an argument\0"
410    "this version of PCRE is not compiled with PCRE_UCP support\0"    "this version of PCRE is not compiled with PCRE_UCP support\0"
411      "\\c must be followed by an ASCII character\0"
412      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
413    ;    ;
414    
415  /* 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 544  static const unsigned char ebcdic_charta Line 546  static const unsigned char ebcdic_charta
546  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
547    
548  static BOOL  static BOOL
549    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,
550      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
551    
552    
# Line 576  return s; Line 578  return s;
578    
579    
580  /*************************************************  /*************************************************
581    *            Check for counted repeat            *
582    *************************************************/
583    
584    /* This function is called when a '{' is encountered in a place where it might
585    start a quantifier. It looks ahead to see if it really is a quantifier or not.
586    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
587    where the ddds are digits.
588    
589    Arguments:
590      p         pointer to the first char after '{'
591    
592    Returns:    TRUE or FALSE
593    */
594    
595    static BOOL
596    is_counted_repeat(const uschar *p)
597    {
598    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
599    while ((digitab[*p] & ctype_digit) != 0) p++;
600    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
601    
602    if (*p++ != CHAR_COMMA) return FALSE;
603    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
604    
605    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
606    while ((digitab[*p] & ctype_digit) != 0) p++;
607    
608    return (*p == CHAR_RIGHT_CURLY_BRACKET);
609    }
610    
611    
612    
613    /*************************************************
614  *            Handle escapes                      *  *            Handle escapes                      *
615  *************************************************/  *************************************************/
616    
# Line 646  else Line 681  else
681      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
682      break;      break;
683    
684      /* \g must be followed by one of a number of specific things:      /* In a character class, \g is just a literal "g". Outside a character
685        class, \g must be followed by one of a number of specific things:
686    
687      (1) A number, either plain or braced. If positive, it is an absolute      (1) A number, either plain or braced. If positive, it is an absolute
688      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
# Line 663  else Line 699  else
699      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
700    
701      case CHAR_g:      case CHAR_g:
702        if (isclass) break;
703      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
704        {        {
705        c = -ESC_g;        c = -ESC_g;
# Line 841  else Line 878  else
878      break;      break;
879    
880      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
881      This coding is ASCII-specific, but then the whole concept of \cx is      An error is given if the byte following \c is not an ASCII character. This
882        coding is ASCII-specific, but then the whole concept of \cx is
883      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
884    
885      case CHAR_c:      case CHAR_c:
# Line 851  else Line 889  else
889        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
890        break;        break;
891        }        }
892    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
893  #ifndef EBCDIC  /* ASCII/UTF-8 coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
894          {
895          *errorcodeptr = ERR68;
896          break;
897          }
898      if (c >= CHAR_a && c <= CHAR_z) c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
899      c ^= 0x40;      c ^= 0x40;
900  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
901      if (c >= CHAR_a && c <= CHAR_z) c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
902      c ^= 0xC0;      c ^= 0xC0;
903  #endif  #endif
# Line 879  else Line 921  else
921    }    }
922    
923  /* Perl supports \N{name} for character names, as well as plain \N for "not  /* Perl supports \N{name} for character names, as well as plain \N for "not
924  newline". PCRE does not support \N{name}. */  newline". PCRE does not support \N{name}. However, it does support
925    quantification such as \N{2,3}. */
926    
927  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
928         !is_counted_repeat(ptr+2))
929    *errorcodeptr = ERR37;    *errorcodeptr = ERR37;
930    
931  /* If PCRE_UCP is set, we change the values for \d etc. */  /* If PCRE_UCP is set, we change the values for \d etc. */
# Line 991  return -1; Line 1035  return -1;
1035    
1036    
1037  /*************************************************  /*************************************************
 *            Check for counted repeat            *  
 *************************************************/  
   
 /* This function is called when a '{' is encountered in a place where it might  
 start a quantifier. It looks ahead to see if it really is a quantifier or not.  
 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  
 where the ddds are digits.  
   
 Arguments:  
   p         pointer to the first char after '{'  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 is_counted_repeat(const uschar *p)  
 {  
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if (*p++ != CHAR_COMMA) return FALSE;  
 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == CHAR_RIGHT_CURLY_BRACKET);  
 }  
   
   
   
 /*************************************************  
1038  *         Read repeat counts                     *  *         Read repeat counts                     *
1039  *************************************************/  *************************************************/
1040    
# Line 1099  top-level call starts at the beginning o Line 1110  top-level call starts at the beginning o
1110  start at a parenthesis. It scans along a pattern's text looking for capturing  start at a parenthesis. It scans along a pattern's text looking for capturing
1111  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
1112  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
1113  returns when it reaches a given numbered subpattern. We know that if (?P< is  returns when it reaches a given numbered subpattern. Recursion is used to keep
1114  encountered, the name will be terminated by '>' because that is checked in the  track of subpatterns that reset the capturing group numbers - the (?| feature.
1115  first pass. Recursion is used to keep track of subpatterns that reset the  
1116  capturing group numbers - the (?| feature.  This function was originally called only from the second pass, in which we know
1117    that if (?< or (?' or (?P< is encountered, the name will be correctly
1118    terminated because that is checked in the first pass. There is now one call to
1119    this function in the first pass, to check for a recursive back reference by
1120    name (so that we can make the whole group atomic). In this case, we need check
1121    only up to the current position in the pattern, and that is still OK because
1122    and previous occurrences will have been checked. To make this work, the test
1123    for "end of pattern" is a check against cd->end_pattern in the main loop,
1124    instead of looking for a binary zero. This means that the special first-pass
1125    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1126    processing items within the loop are OK, because afterwards the main loop will
1127    terminate.)
1128    
1129  Arguments:  Arguments:
1130    ptrptr       address of the current character pointer (updated)    ptrptr       address of the current character pointer (updated)
# Line 1110  Arguments: Line 1132  Arguments:
1132    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1133    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1134    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1135      utf8         TRUE if we are in UTF-8 mode
1136    count        pointer to the current capturing subpattern number (updated)    count        pointer to the current capturing subpattern number (updated)
1137    
1138  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
# Line 1117  Returns: the number of the named s Line 1140  Returns: the number of the named s
1140    
1141  static int  static int
1142  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1143    BOOL xmode, int *count)    BOOL xmode, BOOL utf8, int *count)
1144  {  {
1145  uschar *ptr = *ptrptr;  uschar *ptr = *ptrptr;
1146  int start_count = *count;  int start_count = *count;
# Line 1130  dealing with. The very first call may no Line 1153  dealing with. The very first call may no
1153  if (ptr[0] == CHAR_LEFT_PARENTHESIS)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1154    {    {
1155    /* Handle specials such as (*SKIP) or (*UTF8) etc. */    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1156    
1157    if (ptr[1] == CHAR_ASTERISK) ptr += 2;    if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1158    
1159    /* Handle a normal, unnamed capturing parenthesis. */    /* Handle a normal, unnamed capturing parenthesis. */
1160    
1161    else if (ptr[1] != CHAR_QUESTION_MARK)    else if (ptr[1] != CHAR_QUESTION_MARK)
# Line 1150  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1173  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1173      ptr += 3;      ptr += 3;
1174      dup_parens = TRUE;      dup_parens = TRUE;
1175      }      }
1176    
1177    /* Handle comments; all characters are allowed until a ket is reached. */    /* Handle comments; all characters are allowed until a ket is reached. */
1178    
1179    else if (ptr[2] == CHAR_NUMBER_SIGN)    else if (ptr[2] == CHAR_NUMBER_SIGN)
1180      {      {
1181      for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;      for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1182      goto FAIL_EXIT;      goto FAIL_EXIT;
1183      }      }
1184    
1185    /* Handle a condition. If it is an assertion, just carry on so that it    /* Handle a condition. If it is an assertion, just carry on so that it
1186    is processed as normal. If not, skip to the closing parenthesis of the    is processed as normal. If not, skip to the closing parenthesis of the
# Line 1202  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1225  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1225    }    }
1226    
1227  /* Past any initial parenthesis handling, scan for parentheses or vertical  /* Past any initial parenthesis handling, scan for parentheses or vertical
1228  bars. */  bars. Stop if we get to cd->end_pattern. Note that this is important for the
1229    first-pass call when this value is temporarily adjusted to stop at the current
1230    position. So DO NOT change this to a test for binary zero. */
1231    
1232  for (; *ptr != 0; ptr++)  for (; ptr < cd->end_pattern; ptr++)
1233    {    {
1234    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1235    
# Line 1278  for (; *ptr != 0; ptr++) Line 1303  for (; *ptr != 0; ptr++)
1303    
1304    if (xmode && *ptr == CHAR_NUMBER_SIGN)    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1305      {      {
1306      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};      ptr++;
1307        while (*ptr != 0)
1308          {
1309          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1310          ptr++;
1311    #ifdef SUPPORT_UTF8
1312          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1313    #endif
1314          }
1315      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1316      continue;      continue;
1317      }      }
# Line 1287  for (; *ptr != 0; ptr++) Line 1320  for (; *ptr != 0; ptr++)
1320    
1321    if (*ptr == CHAR_LEFT_PARENTHESIS)    if (*ptr == CHAR_LEFT_PARENTHESIS)
1322      {      {
1323      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1324      if (rc > 0) return rc;      if (rc > 0) return rc;
1325      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1326      }      }
# Line 1295  for (; *ptr != 0; ptr++) Line 1328  for (; *ptr != 0; ptr++)
1328    else if (*ptr == CHAR_RIGHT_PARENTHESIS)    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1329      {      {
1330      if (dup_parens && *count < hwm_count) *count = hwm_count;      if (dup_parens && *count < hwm_count) *count = hwm_count;
1331      goto FAIL_EXIT;      goto FAIL_EXIT;
1332      }      }
1333    
1334    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
# Line 1333  Arguments: Line 1366  Arguments:
1366    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1367    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1368    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1369      utf8         TRUE if we are in UTF-8 mode
1370    
1371  Returns:       the number of the found subpattern, or -1 if not found  Returns:       the number of the found subpattern, or -1 if not found
1372  */  */
1373    
1374  static int  static int
1375  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1376      BOOL utf8)
1377  {  {
1378  uschar *ptr = (uschar *)cd->start_pattern;  uschar *ptr = (uschar *)cd->start_pattern;
1379  int count = 0;  int count = 0;
# Line 1351  matching closing parens. That is why we Line 1386  matching closing parens. That is why we
1386    
1387  for (;;)  for (;;)
1388    {    {
1389    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1390    if (rc > 0 || *ptr++ == 0) break;    if (rc > 0 || *ptr++ == 0) break;
1391    }    }
1392    
# Line 1367  return rc; Line 1402  return rc;
1402    
1403  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1404  for a fixed first character, or an anchoring op code etc. It skips over things  for a fixed first character, or an anchoring op code etc. It skips over things
1405  that do not influence this. For some calls, a change of option is important.  that do not influence this. For some calls, it makes sense to skip negative
1406  For some calls, it makes sense to skip negative forward and all backward  forward and all backward assertions, and also the \b assertion; for others it
1407  assertions, and also the \b assertion; for others it does not.  does not.
1408    
1409  Arguments:  Arguments:
1410    code         pointer to the start of the group    code         pointer to the start of the group
   options      pointer to external options  
   optbit       the option bit whose changing is significant, or  
                  zero if none are  
1411    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1412    
1413  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1414  */  */
1415    
1416  static const uschar*  static const uschar*
1417  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1418  {  {
1419  for (;;)  for (;;)
1420    {    {
1421    switch ((int)*code)    switch ((int)*code)
1422      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1423      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1424      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1425      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 1444  and doing the check at the end; a flag s Line 1469  and doing the check at the end; a flag s
1469    
1470  Arguments:  Arguments:
1471    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1472    options  the compiling options    utf8     TRUE in UTF-8 mode
1473    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1474    cd       the "compile data" structure    cd       the "compile data" structure
1475    
# Line 1455  Returns: the fixed length, Line 1480  Returns: the fixed length,
1480  */  */
1481    
1482  static int  static int
1483  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1484  {  {
1485  int length = -1;  int length = -1;
1486    
# Line 1472  for (;;) Line 1497  for (;;)
1497    register int op = *cc;    register int op = *cc;
1498    switch (op)    switch (op)
1499      {      {
1500        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1501        OP_BRA (normal non-capturing bracket) because the other variants of these
1502        opcodes are all concerned with unlimited repeated groups, which of course
1503        are not of fixed length. They will cause a -1 response from the default
1504        case of this switch. */
1505    
1506      case OP_CBRA:      case OP_CBRA:
1507      case OP_BRA:      case OP_BRA:
1508      case OP_ONCE:      case OP_ONCE:
1509        case OP_ONCE_NC:
1510      case OP_COND:      case OP_COND:
1511      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1512      if (d < 0) return d;      if (d < 0) return d;
1513      branchlength += d;      branchlength += d;
1514      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1485  for (;;) Line 1517  for (;;)
1517    
1518      /* Reached end of a branch; if it's a ket it is the end of a nested      /* Reached end of a branch; if it's a ket it is the end of a nested
1519      call. If it's ALT it is an alternation in a nested call. If it is      call. If it's ALT it is an alternation in a nested call. If it is
1520      END it's the end of the outer call. All can be handled by the same code. */      END it's the end of the outer call. All can be handled by the same code.
1521        Note that we must not include the OP_KETRxxx opcodes here, because they
1522        all imply an unlimited repeat. */
1523    
1524      case OP_ALT:      case OP_ALT:
1525      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1526      case OP_END:      case OP_END:
1527      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1528        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
# Line 1508  for (;;) Line 1540  for (;;)
1540      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1541      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1542      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                /* Recursion */
1543      d = find_fixedlength(cs + 2, options, atend, cd);      d = find_fixedlength(cs + 2, utf8, atend, cd);
1544      if (d < 0) return d;      if (d < 0) return d;
1545      branchlength += d;      branchlength += d;
1546      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1531  for (;;) Line 1563  for (;;)
1563      case OP_RREF:      case OP_RREF:
1564      case OP_NRREF:      case OP_NRREF:
1565      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
1566      case OP_CALLOUT:      case OP_CALLOUT:
1567      case OP_SOD:      case OP_SOD:
1568      case OP_SOM:      case OP_SOM:
# Line 1539  for (;;) Line 1570  for (;;)
1570      case OP_EOD:      case OP_EOD:
1571      case OP_EODN:      case OP_EODN:
1572      case OP_CIRC:      case OP_CIRC:
1573        case OP_CIRCM:
1574      case OP_DOLL:      case OP_DOLL:
1575        case OP_DOLLM:
1576      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1577      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1578      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 1548  for (;;) Line 1581  for (;;)
1581      /* Handle literal characters */      /* Handle literal characters */
1582    
1583      case OP_CHAR:      case OP_CHAR:
1584      case OP_CHARNC:      case OP_CHARI:
1585      case OP_NOT:      case OP_NOT:
1586        case OP_NOTI:
1587      branchlength++;      branchlength++;
1588      cc += 2;      cc += 2;
1589  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1590      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
1591  #endif  #endif
1592      break;      break;
1593    
# Line 1565  for (;;) Line 1598  for (;;)
1598      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1599      cc += 4;      cc += 4;
1600  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1601      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
1602  #endif  #endif
1603      break;      break;
1604    
# Line 1668  _pcre_find_bracket(const uschar *code, B Line 1700  _pcre_find_bracket(const uschar *code, B
1700  for (;;)  for (;;)
1701    {    {
1702    register int c = *code;    register int c = *code;
1703    
1704    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1705    
1706    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1686  for (;;) Line 1719  for (;;)
1719    
1720    /* Handle capturing bracket */    /* Handle capturing bracket */
1721    
1722    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1723               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1724      {      {
1725      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1726      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1724  for (;;) Line 1758  for (;;)
1758        case OP_MARK:        case OP_MARK:
1759        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
1760        case OP_SKIP_ARG:        case OP_SKIP_ARG:
1761          code += code[1];
1762          break;
1763    
1764        case OP_THEN_ARG:        case OP_THEN_ARG:
1765        code += code[1];        code += code[1];
1766        break;        break;
# Line 1741  for (;;) Line 1778  for (;;)
1778      if (utf8) switch(c)      if (utf8) switch(c)
1779        {        {
1780        case OP_CHAR:        case OP_CHAR:
1781        case OP_CHARNC:        case OP_CHARI:
1782        case OP_EXACT:        case OP_EXACT:
1783          case OP_EXACTI:
1784        case OP_UPTO:        case OP_UPTO:
1785          case OP_UPTOI:
1786        case OP_MINUPTO:        case OP_MINUPTO:
1787          case OP_MINUPTOI:
1788        case OP_POSUPTO:        case OP_POSUPTO:
1789          case OP_POSUPTOI:
1790        case OP_STAR:        case OP_STAR:
1791          case OP_STARI:
1792        case OP_MINSTAR:        case OP_MINSTAR:
1793          case OP_MINSTARI:
1794        case OP_POSSTAR:        case OP_POSSTAR:
1795          case OP_POSSTARI:
1796        case OP_PLUS:        case OP_PLUS:
1797          case OP_PLUSI:
1798        case OP_MINPLUS:        case OP_MINPLUS:
1799          case OP_MINPLUSI:
1800        case OP_POSPLUS:        case OP_POSPLUS:
1801          case OP_POSPLUSI:
1802        case OP_QUERY:        case OP_QUERY:
1803          case OP_QUERYI:
1804        case OP_MINQUERY:        case OP_MINQUERY:
1805          case OP_MINQUERYI:
1806        case OP_POSQUERY:        case OP_POSQUERY:
1807          case OP_POSQUERYI:
1808        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1809        break;        break;
1810        }        }
# Line 1827  for (;;) Line 1877  for (;;)
1877        case OP_MARK:        case OP_MARK:
1878        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
1879        case OP_SKIP_ARG:        case OP_SKIP_ARG:
1880          code += code[1];
1881          break;
1882    
1883        case OP_THEN_ARG:        case OP_THEN_ARG:
1884        code += code[1];        code += code[1];
1885        break;        break;
# Line 1844  for (;;) Line 1897  for (;;)
1897      if (utf8) switch(c)      if (utf8) switch(c)
1898        {        {
1899        case OP_CHAR:        case OP_CHAR:
1900        case OP_CHARNC:        case OP_CHARI:
1901        case OP_EXACT:        case OP_EXACT:
1902          case OP_EXACTI:
1903        case OP_UPTO:        case OP_UPTO:
1904          case OP_UPTOI:
1905        case OP_MINUPTO:        case OP_MINUPTO:
1906          case OP_MINUPTOI:
1907        case OP_POSUPTO:        case OP_POSUPTO:
1908          case OP_POSUPTOI:
1909        case OP_STAR:        case OP_STAR:
1910          case OP_STARI:
1911        case OP_MINSTAR:        case OP_MINSTAR:
1912          case OP_MINSTARI:
1913        case OP_POSSTAR:        case OP_POSSTAR:
1914          case OP_POSSTARI:
1915        case OP_PLUS:        case OP_PLUS:
1916          case OP_PLUSI:
1917        case OP_MINPLUS:        case OP_MINPLUS:
1918          case OP_MINPLUSI:
1919        case OP_POSPLUS:        case OP_POSPLUS:
1920          case OP_POSPLUSI:
1921        case OP_QUERY:        case OP_QUERY:
1922          case OP_QUERYI:
1923        case OP_MINQUERY:        case OP_MINQUERY:
1924          case OP_MINQUERYI:
1925        case OP_POSQUERY:        case OP_POSQUERY:
1926          case OP_POSQUERYI:
1927        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1928        break;        break;
1929        }        }
# Line 1896  could_be_empty_branch(const uschar *code Line 1962  could_be_empty_branch(const uschar *code
1962    compile_data *cd)    compile_data *cd)
1963  {  {
1964  register int c;  register int c;
1965  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
1966       code < endcode;       code < endcode;
1967       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
1968    {    {
1969    const uschar *ccode;    const uschar *ccode;
1970    
# Line 1914  for (code = first_significant_code(code Line 1980  for (code = first_significant_code(code
1980      continue;      continue;
1981      }      }
1982    
   /* Groups with zero repeats can of course be empty; skip them. */  
   
   if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)  
     {  
     code += _pcre_OP_lengths[c];  
     do code += GET(code, 1); while (*code == OP_ALT);  
     c = *code;  
     continue;  
     }  
   
1983    /* For a recursion/subroutine call, if its end has been reached, which    /* For a recursion/subroutine call, if its end has been reached, which
1984    implies a subroutine call, we can scan it. */    implies a backward reference subroutine call, we can scan it. If it's a
1985      forward reference subroutine call, we can't. To detect forward reference
1986      we have to scan up the list that is kept in the workspace. This function is
1987      called only when doing the real compile, not during the pre-compile that
1988      measures the size of the compiled pattern. */
1989    
1990    if (c == OP_RECURSE)    if (c == OP_RECURSE)
1991      {      {
1992      BOOL empty_branch = FALSE;      const uschar *scode;
1993      const uschar *scode = cd->start_code + GET(code, 1);      BOOL empty_branch;
1994    
1995        /* Test for forward reference */
1996    
1997        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
1998          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
1999    
2000        /* Not a forward reference, test for completed backward reference */
2001    
2002        empty_branch = FALSE;
2003        scode = cd->start_code + GET(code, 1);
2004      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2005    
2006        /* Completed backwards reference */
2007    
2008      do      do
2009        {        {
2010        if (could_be_empty_branch(scode, endcode, utf8, cd))        if (could_be_empty_branch(scode, endcode, utf8, cd))
# Line 1942  for (code = first_significant_code(code Line 2015  for (code = first_significant_code(code
2015        scode += GET(scode, 1);        scode += GET(scode, 1);
2016        }        }
2017      while (*scode == OP_ALT);      while (*scode == OP_ALT);
2018    
2019      if (!empty_branch) return FALSE;  /* All branches are non-empty */      if (!empty_branch) return FALSE;  /* All branches are non-empty */
2020      continue;      continue;
2021      }      }
2022    
2023      /* Groups with zero repeats can of course be empty; skip them. */
2024    
2025      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2026          c == OP_BRAPOSZERO)
2027        {
2028        code += _pcre_OP_lengths[c];
2029        do code += GET(code, 1); while (*code == OP_ALT);
2030        c = *code;
2031        continue;
2032        }
2033    
2034      /* A nested group that is already marked as "could be empty" can just be
2035      skipped. */
2036    
2037      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2038          c == OP_SCBRA || c == OP_SCBRAPOS)
2039        {
2040        do code += GET(code, 1); while (*code == OP_ALT);
2041        c = *code;
2042        continue;
2043        }
2044    
2045    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2046    
2047    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2048          c == OP_CBRA || c == OP_CBRAPOS ||
2049          c == OP_ONCE || c == OP_ONCE_NC ||
2050          c == OP_COND)
2051      {      {
2052      BOOL empty_branch;      BOOL empty_branch;
2053      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 2035  for (code = first_significant_code(code Line 2134  for (code = first_significant_code(code
2134      case OP_ALLANY:      case OP_ALLANY:
2135      case OP_ANYBYTE:      case OP_ANYBYTE:
2136      case OP_CHAR:      case OP_CHAR:
2137      case OP_CHARNC:      case OP_CHARI:
2138      case OP_NOT:      case OP_NOT:
2139        case OP_NOTI:
2140      case OP_PLUS:      case OP_PLUS:
2141      case OP_MINPLUS:      case OP_MINPLUS:
2142      case OP_POSPLUS:      case OP_POSPLUS:
# Line 2076  for (code = first_significant_code(code Line 2176  for (code = first_significant_code(code
2176      case OP_KET:      case OP_KET:
2177      case OP_KETRMAX:      case OP_KETRMAX:
2178      case OP_KETRMIN:      case OP_KETRMIN:
2179        case OP_KETRPOS:
2180      case OP_ALT:      case OP_ALT:
2181      return TRUE;      return TRUE;
2182    
# Line 2084  for (code = first_significant_code(code Line 2185  for (code = first_significant_code(code
2185    
2186  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2187      case OP_STAR:      case OP_STAR:
2188        case OP_STARI:
2189      case OP_MINSTAR:      case OP_MINSTAR:
2190        case OP_MINSTARI:
2191      case OP_POSSTAR:      case OP_POSSTAR:
2192        case OP_POSSTARI:
2193      case OP_QUERY:      case OP_QUERY:
2194        case OP_QUERYI:
2195      case OP_MINQUERY:      case OP_MINQUERY:
2196        case OP_MINQUERYI:
2197      case OP_POSQUERY:      case OP_POSQUERY:
2198        case OP_POSQUERYI:
2199      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2200      break;      break;
2201    
2202      case OP_UPTO:      case OP_UPTO:
2203        case OP_UPTOI:
2204      case OP_MINUPTO:      case OP_MINUPTO:
2205        case OP_MINUPTOI:
2206      case OP_POSUPTO:      case OP_POSUPTO:
2207        case OP_POSUPTOI:
2208      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2209      break;      break;
2210  #endif  #endif
# Line 2105  for (code = first_significant_code(code Line 2215  for (code = first_significant_code(code
2215      case OP_MARK:      case OP_MARK:
2216      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
2217      case OP_SKIP_ARG:      case OP_SKIP_ARG:
2218        code += code[1];
2219        break;
2220    
2221      case OP_THEN_ARG:      case OP_THEN_ARG:
2222      code += code[1];      code += code[1];
2223      break;      break;
# Line 2129  return TRUE; Line 2242  return TRUE;
2242  the current branch of the current pattern to see if it could match the empty  the current branch of the current pattern to see if it could match the empty
2243  string. If it could, we must look outwards for branches at other levels,  string. If it could, we must look outwards for branches at other levels,
2244  stopping when we pass beyond the bracket which is the subject of the recursion.  stopping when we pass beyond the bracket which is the subject of the recursion.
2245    This function is called only during the real compile, not during the
2246    pre-compile.
2247    
2248  Arguments:  Arguments:
2249    code        points to start of the recursion    code        points to start of the recursion
# Line 2179  where Perl recognizes it as the POSIX cl Line 2294  where Perl recognizes it as the POSIX cl
2294  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2295  I think.  I think.
2296    
2297    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2298    It seems that the appearance of a nested POSIX class supersedes an apparent
2299    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2300    a digit.
2301    
2302    In Perl, unescaped square brackets may also appear as part of class names. For
2303    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2304    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2305    seem right at all. PCRE does not allow closing square brackets in POSIX class
2306    names.
2307    
2308  Arguments:  Arguments:
2309    ptr      pointer to the initial [    ptr      pointer to the initial [
2310    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 2193  int terminator; /* Don't combin Line 2319  int terminator; /* Don't combin
2319  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2320  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2321    {    {
2322    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2323        ptr++;
2324      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2325      else
2326      {      {
     if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
2327      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2328        {        {
2329        *endptr = ptr;        *endptr = ptr;
2330        return TRUE;        return TRUE;
2331        }        }
2332        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2333             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2334              ptr[1] == CHAR_EQUALS_SIGN) &&
2335            check_posix_syntax(ptr, endptr))
2336          return FALSE;
2337      }      }
2338    }    }
2339  return FALSE;  return FALSE;
# Line 2506  if ((options & PCRE_EXTENDED) != 0) Line 2639  if ((options & PCRE_EXTENDED) != 0)
2639      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2640      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2641        {        {
2642        while (*(++ptr) != 0)        ptr++;
2643          while (*ptr != 0)
2644            {
2645          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2646            ptr++;
2647    #ifdef SUPPORT_UTF8
2648            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2649    #endif
2650            }
2651        }        }
2652      else break;      else break;
2653      }      }
# Line 2543  if ((options & PCRE_EXTENDED) != 0) Line 2683  if ((options & PCRE_EXTENDED) != 0)
2683      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2684      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2685        {        {
2686        while (*(++ptr) != 0)        ptr++;
2687          while (*ptr != 0)
2688            {
2689          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2690            ptr++;
2691    #ifdef SUPPORT_UTF8
2692            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2693    #endif
2694            }
2695        }        }
2696      else break;      else break;
2697      }      }
# Line 2569  if (next >= 0) switch(op_code) Line 2716  if (next >= 0) switch(op_code)
2716  #endif  #endif
2717    return c != next;    return c != next;
2718    
2719    /* For CHARNC (caseless character) we must check the other case. If we have    /* For CHARI (caseless character) we must check the other case. If we have
2720    Unicode property support, we can use it to test the other case of    Unicode property support, we can use it to test the other case of
2721    high-valued characters. */    high-valued characters. */
2722    
2723    case OP_CHARNC:    case OP_CHARI:
2724  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2725    GETCHARTEST(c, previous);    GETCHARTEST(c, previous);
2726  #else  #else
# Line 2596  if (next >= 0) switch(op_code) Line 2743  if (next >= 0) switch(op_code)
2743  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2744    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2745    
2746    /* For OP_NOT, its data is always a single-byte character. */    /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These
2747      opcodes are not used for multi-byte characters, because they are coded using
2748      an XCLASS instead. */
2749    
2750    case OP_NOT:    case OP_NOT:
2751      return (c = *previous) == next;
2752    
2753      case OP_NOTI:
2754    if ((c = *previous) == next) return TRUE;    if ((c = *previous) == next) return TRUE;
   if ((options & PCRE_CASELESS) == 0) return FALSE;  
2755  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2756    if (utf8)    if (utf8)
2757      {      {
# Line 2705  replaced by OP_PROP codes when PCRE_UCP Line 2856  replaced by OP_PROP codes when PCRE_UCP
2856  switch(op_code)  switch(op_code)
2857    {    {
2858    case OP_CHAR:    case OP_CHAR:
2859    case OP_CHARNC:    case OP_CHARI:
2860  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2861    GETCHARTEST(c, previous);    GETCHARTEST(c, previous);
2862  #else  #else
# Line 2892  Arguments: Line 3043  Arguments:
3043    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3044    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3045    bcptr          points to current branch chain    bcptr          points to current branch chain
3046      cond_depth     conditional nesting depth
3047    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3048    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3049                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2903  Returns: TRUE on success Line 3055  Returns: TRUE on success
3055  static BOOL  static BOOL
3056  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3057    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3058    compile_data *cd, int *lengthptr)    int cond_depth, compile_data *cd, int *lengthptr)
3059  {  {
3060  int repeat_type, op_type;  int repeat_type, op_type;
3061  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 2912  int greedy_default, greedy_non_default; Line 3064  int greedy_default, greedy_non_default;
3064  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3065  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3066  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3067  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3068  int after_manual_callout = 0;  int after_manual_callout = 0;
3069  int length_prevgroup = 0;  int length_prevgroup = 0;
3070  register int c;  register int c;
# Line 2930  uschar *previous_callout = NULL; Line 3082  uschar *previous_callout = NULL;
3082  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3083  uschar classbits[32];  uschar classbits[32];
3084    
3085    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3086    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3087    dynamically as we process the pattern. */
3088    
3089  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3090  BOOL class_utf8;  BOOL class_utf8;
3091  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
# Line 2938  uschar *class_utf8data_base; Line 3094  uschar *class_utf8data_base;
3094  uschar utf8_char[6];  uschar utf8_char[6];
3095  #else  #else
3096  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3097  #endif  #endif
3098    
3099  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
# Line 2989  for (;; ptr++) Line 3144  for (;; ptr++)
3144    int subfirstbyte;    int subfirstbyte;
3145    int terminator;    int terminator;
3146    int mclength;    int mclength;
3147      int tempbracount;
3148    uschar mcbuffer[8];    uschar mcbuffer[8];
3149    
3150    /* Get next byte in the pattern */    /* Get next byte in the pattern */
# Line 3110  for (;; ptr++) Line 3266  for (;; ptr++)
3266      previous_callout = NULL;      previous_callout = NULL;
3267      }      }
3268    
3269    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3270    
3271    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3272      {      {
3273      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3274      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
3275        {        {
3276        while (*(++ptr) != 0)        ptr++;
3277          while (*ptr != 0)
3278          {          {
3279          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3280            ptr++;
3281    #ifdef SUPPORT_UTF8
3282            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3283    #endif
3284          }          }
3285        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3286    
# Line 3164  for (;; ptr++) Line 3325  for (;; ptr++)
3325      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3326    
3327      case CHAR_CIRCUMFLEX_ACCENT:      case CHAR_CIRCUMFLEX_ACCENT:
3328        previous = NULL;
3329      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3330        {        {
3331        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3332          *code++ = OP_CIRCM;
3333        }        }
3334      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3335      break;      break;
3336    
3337      case CHAR_DOLLAR_SIGN:      case CHAR_DOLLAR_SIGN:
3338      previous = NULL;      previous = NULL;
3339      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3340      break;      break;
3341    
3342      /* There can never be a first char if '.' is first, whatever happens about      /* There can never be a first char if '.' is first, whatever happens about
# Line 3494  for (;; ptr++) Line 3656  for (;; ptr++)
3656              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3657              continue;              continue;
3658    
3659                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3660                if it was previously set by something earlier in the character
3661                class. */
3662    
3663              case ESC_s:              case ESC_s:
3664              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3665              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3666                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3667              continue;              continue;
3668    
3669              case ESC_S:              case ESC_S:
# Line 3915  for (;; ptr++) Line 4082  for (;; ptr++)
4082    
4083      In UTF-8 mode, we can optimize the negative case only if there were no      In UTF-8 mode, we can optimize the negative case only if there were no
4084      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4085      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4086      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4087    
4088      The optimization throws away the bit map. We turn the item into a      The optimization throws away the bit map. We turn the item into a
4089      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[I] if it's positive, or OP_NOT[I] if it's negative.
4090      that OP_NOT does not support multibyte characters. In the positive case, it      Note that OP_NOT[I] does not support multibyte characters. In the positive
4091      can cause firstbyte to be set. Otherwise, there can be no first char if      case, it can cause firstbyte to be set. Otherwise, there can be no first
4092      this item is first, whatever repeat count may follow. In the case of      char if this item is first, whatever repeat count may follow. In the case
4093      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4094    
4095  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4096      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
# Line 3934  for (;; ptr++) Line 4101  for (;; ptr++)
4101        {        {
4102        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4103    
4104        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4105    
4106        if (negate_class)        if (negate_class)
4107          {          {
4108          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4109          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4110          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4111          *code++ = class_lastchar;          *code++ = class_lastchar;
4112          break;          break;
4113          }          }
# Line 4068  for (;; ptr++) Line 4235  for (;; ptr++)
4235      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4236      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4237    
4238      /* Save start of previous item, in case we have to move it up to make space      /* Save start of previous item, in case we have to move it up in order to
4239      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4240    
4241      tempcode = previous;      tempcode = previous;
4242    
# Line 4092  for (;; ptr++) Line 4259  for (;; ptr++)
4259        }        }
4260      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4261    
4262        /* If previous was a recursion call, wrap it in atomic brackets so that
4263        previous becomes the atomic group. All recursions were so wrapped in the
4264        past, but it no longer happens for non-repeated recursions. In fact, the
4265        repeated ones could be re-implemented independently so as not to need this,
4266        but for the moment we rely on the code for repeating groups. */
4267    
4268        if (*previous == OP_RECURSE)
4269          {
4270          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4271          *previous = OP_ONCE;
4272          PUT(previous, 1, 2 + 2*LINK_SIZE);
4273          previous[2 + 2*LINK_SIZE] = OP_KET;
4274          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4275          code += 2 + 2 * LINK_SIZE;
4276          length_prevgroup = 3 + 3*LINK_SIZE;
4277    
4278          /* When actually compiling, we need to check whether this was a forward
4279          reference, and if so, adjust the offset. */
4280    
4281          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4282            {
4283            int offset = GET(cd->hwm, -LINK_SIZE);
4284            if (offset == previous + 1 - cd->start_code)
4285              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4286            }
4287          }
4288    
4289        /* Now handle repetition for the different types of item. */
4290    
4291      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4292      repeat item instead. If a char item has a minumum of more than one, ensure      repeat item instead. If a char item has a minumum of more than one, ensure
4293      that it is set in reqbyte - it might not be if a sequence such as x{3} is      that it is set in reqbyte - it might not be if a sequence such as x{3} is
4294      the first thing in a branch because the x will have gone into firstbyte      the first thing in a branch because the x will have gone into firstbyte
4295      instead.  */      instead.  */
4296    
4297      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4298        {        {
4299          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4300    
4301        /* Deal with UTF-8 characters that take up more than one byte. It's        /* Deal with UTF-8 characters that take up more than one byte. It's
4302        easier to write this out separately than try to macrify it. Use c to        easier to write this out separately than try to macrify it. Use c to
4303        hold the length of the character in bytes, plus 0x80 to flag that it's a        hold the length of the character in bytes, plus 0x80 to flag that it's a
# Line 4144  for (;; ptr++) Line 4342  for (;; ptr++)
4342      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4343      one of the special opcodes, replacing it. The code is shared with single-      one of the special opcodes, replacing it. The code is shared with single-
4344      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4345      repeat_type. We can also test for auto-possessification. OP_NOT is      repeat_type. We can also test for auto-possessification. OP_NOT and OP_NOTI
4346      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4347    
4348      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4349        {        {
4350        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4351        c = previous[1];        c = previous[1];
4352        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4353            repeat_max < 0 &&            repeat_max < 0 &&
# Line 4346  for (;; ptr++) Line 4544  for (;; ptr++)
4544  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4545               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4546  #endif  #endif
4547               *previous == OP_REF)               *previous == OP_REF ||
4548                 *previous == OP_REFI)
4549        {        {
4550        if (repeat_max == 0)        if (repeat_max == 0)
4551          {          {
# Line 4380  for (;; ptr++) Line 4579  for (;; ptr++)
4579        }        }
4580    
4581      /* If previous was a bracket group, we may have to replicate it in certain      /* If previous was a bracket group, we may have to replicate it in certain
4582      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
4583        opcodes such as BRA and CBRA, as this is the place where they get converted
4584        into the more special varieties such as BRAPOS and SBRA. A test for >=
4585        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4586        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4587        repetition of assertions, but now it does, for Perl compatibility. */
4588    
4589      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4590        {        {
4591        register int i;        register int i;
       int ketoffset = 0;  
4592        int len = (int)(code - previous);        int len = (int)(code - previous);
4593        uschar *bralink = NULL;        uschar *bralink = NULL;
4594          uschar *brazeroptr = NULL;
4595    
4596        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4597          we just ignore the repeat. */
4598    
4599        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4600          {          goto END_REPEAT;
4601          *errorcodeptr = ERR55;  
4602          goto FAILED;        /* There is no sense in actually repeating assertions. The only potential
4603          }        use of repetition is in cases when the assertion is optional. Therefore,
4604          if the minimum is greater than zero, just ignore the repeat. If the
4605          maximum is not not zero or one, set it to 1. */
4606    
4607        /* If the maximum repeat count is unlimited, find the end of the bracket        if (*previous < OP_ONCE)    /* Assertion */
4608        by scanning through from the start, and compute the offset back to it          {
4609        from the current code pointer. There may be an OP_OPT setting following          if (repeat_min > 0) goto END_REPEAT;
4610        the final KET, so we can't find the end just by going back from the code          if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
       pointer. */  
   
       if (repeat_max == -1)  
         {  
         register uschar *ket = previous;  
         do ket += GET(ket, 1); while (*ket != OP_KET);  
         ketoffset = (int)(code - ket);  
4611          }          }
4612    
4613        /* The case of a zero minimum is special because of the need to stick        /* The case of a zero minimum is special because of the need to stick
# Line 4429  for (;; ptr++) Line 4628  for (;; ptr++)
4628          **   goto END_REPEAT;          **   goto END_REPEAT;
4629          **   }          **   }
4630    
4631          However, that fails when a group is referenced as a subroutine from          However, that fails when a group or a subgroup within it is referenced
4632          elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it          as a subroutine from elsewhere in the pattern, so now we stick in
4633          so that it is skipped on execution. As we don't have a list of which          OP_SKIPZERO in front of it so that it is skipped on execution. As we
4634          groups are referenced, we cannot do this selectively.          don't have a list of which groups are referenced, we cannot do this
4635            selectively.
4636    
4637          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4638          and do no more at this point. However, we do need to adjust any          and do no more at this point. However, we do need to adjust any
# Line 4452  for (;; ptr++) Line 4652  for (;; ptr++)
4652              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
4653              goto END_REPEAT;              goto END_REPEAT;
4654              }              }
4655              brazeroptr = previous;    /* Save for possessive optimizing */
4656            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4657            }            }
4658    
# Line 4616  for (;; ptr++) Line 4817  for (;; ptr++)
4817            }            }
4818          }          }
4819    
4820        /* If the maximum is unlimited, set a repeater in the final copy. We        /* If the maximum is unlimited, set a repeater in the final copy. For
4821        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
4822        don't know if there's been an options resetting after the ket. The        ONCE brackets can be converted into non-capturing brackets, as the
4823        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
4824          deal with possessive ONCEs specially.
4825    
4826          Otherwise, if the quantifier was possessive, we convert the BRA code to
4827          the POS form, and the KET code to KETRPOS. (It turns out to be convenient
4828          at runtime to detect this kind of subpattern at both the start and at the
4829          end.) The use of special opcodes makes it possible to reduce greatly the
4830          stack usage in pcre_exec(). If the group is preceded by OP_BRAZERO,
4831          convert this to OP_BRAPOSZERO. Then cancel the possessive flag so that
4832          the default action below, of wrapping everything inside atomic brackets,
4833          does not happen.
4834    
4835        Then, when we are doing the actual compile phase, check to see whether        Then, when we are doing the actual compile phase, check to see whether
4836        this group is a non-atomic one that could match an empty string. If so,        this group is one that could match an empty string. If so, convert the
4837        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so        initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so that runtime
4838        that runtime checking can be done. [This check is also applied to        checking can be done. [This check is also applied to ONCE groups at
4839        atomic groups at runtime, but in a different way.] */        runtime, but in a different way.] */
4840    
4841        else        else
4842          {          {
4843          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
4844          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
4845          *ketcode = OP_KETRMAX + repeat_type;  
4846          if (lengthptr == NULL && *bracode != OP_ONCE)          if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
4847                possessive_quantifier) *bracode = OP_BRA;
4848    
4849            if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
4850              *ketcode = OP_KETRMAX + repeat_type;
4851            else
4852            {            {
4853            uschar *scode = bracode;            if (possessive_quantifier)
4854            do              {
4855                *bracode += 1;                   /* Switch to xxxPOS opcodes */
4856                *ketcode = OP_KETRPOS;
4857                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
4858                possessive_quantifier = FALSE;
4859                }
4860              else *ketcode = OP_KETRMAX + repeat_type;
4861    
4862              if (lengthptr == NULL)
4863              {              {
4864              if (could_be_empty_branch(scode, ketcode, utf8, cd))              uschar *scode = bracode;
4865                do
4866                {                {
4867                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
4868                break;                  {
4869                    *bracode += OP_SBRA - OP_BRA;
4870                    break;
4871                    }
4872                  scode += GET(scode, 1);
4873                }                }
4874              scode += GET(scode, 1);              while (*scode == OP_ALT);
4875              }              }
           while (*scode == OP_ALT);  
4876            }            }
4877          }          }
4878        }        }
# Line 4665  for (;; ptr++) Line 4893  for (;; ptr++)
4893        }        }
4894    
4895      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
4896      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
4897      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
4898      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
4899      The '+' notation is just syntactic sugar, taken from Sun's Java package,      notation is just syntactic sugar, taken from Sun's Java package, but the
4900      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
4901      tempcode, not at previous, which might be the first part of a string whose  
4902      (former) last char we repeated.      Possessively repeated subpatterns have already been handled in the code
4903        just above, so possessive_quantifier is always FALSE for them at this
4904        stage.
4905    
4906        Note that the repeated item starts at tempcode, not at previous, which
4907        might be the first part of a string whose (former) last char we repeated.
4908    
4909      Possessifying an 'exact' quantifier has no effect, so we can ignore it. But      Possessifying an 'exact' quantifier has no effect, so we can ignore it. But
4910      an 'upto' may follow. We skip over an 'exact' item, and then test the      an 'upto' may follow. We skip over an 'exact' item, and then test the
# Line 4702  for (;; ptr++) Line 4935  for (;; ptr++)
4935          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
4936          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4937    
4938          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
4939          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
4940          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
4941          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
4942    
4943          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4944          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4945          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4946          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4947    
4948            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
4949            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
4950            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
4951            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
4952    
4953            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
4954            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
4955            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
4956            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
4957    
4958          /* Because we are moving code along, we must ensure that any          /* Because we are moving code along, we must ensure that any
4959          pending recursive references are updated. */          pending recursive references are updated. */
4960    
# Line 4765  for (;; ptr++) Line 5008  for (;; ptr++)
5008        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
5009        namelen = (int)(ptr - name);        namelen = (int)(ptr - name);
5010    
5011          /* It appears that Perl allows any characters whatsoever, other than
5012          a closing parenthesis, to appear in arguments, so we no longer insist on
5013          letters, digits, and underscores. */
5014    
5015        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
5016          {          {
5017          arg = ++ptr;          arg = ++ptr;
5018          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
           || *ptr == '_') ptr++;  
5019          arglen = (int)(ptr - arg);          arglen = (int)(ptr - arg);
5020          }          }
5021    
# Line 4786  for (;; ptr++) Line 5032  for (;; ptr++)
5032          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
5033              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
5034            {            {
5035            /* Check for open captures before ACCEPT */            /* Check for open captures before ACCEPT and convert it to
5036              ASSERT_ACCEPT if in an assertion. */
5037    
5038            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
5039              {              {
5040              open_capitem *oc;              open_capitem *oc;
5041                if (arglen != 0)
5042                  {
5043                  *errorcodeptr = ERR59;
5044                  goto FAILED;
5045                  }
5046              cd->had_accept = TRUE;              cd->had_accept = TRUE;
5047              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5048                {                {
5049                *code++ = OP_CLOSE;                *code++ = OP_CLOSE;
5050                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
5051                }                }
5052                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5053    
5054                /* Do not set firstbyte after *ACCEPT */
5055                if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
5056              }              }
5057    
5058            /* Handle the cases with/without an argument */            /* Handle other cases with/without an argument */
5059    
5060            if (arglen == 0)            else if (arglen == 0)
5061              {              {
5062              if (verbs[i].op < 0)   /* Argument is mandatory */              if (verbs[i].op < 0)   /* Argument is mandatory */
5063                {                {
5064                *errorcodeptr = ERR66;                *errorcodeptr = ERR66;
5065                goto FAILED;                goto FAILED;
5066                }                }
5067              *code++ = verbs[i].op;              *code = verbs[i].op;
5068                if (*code++ == OP_THEN) cd->external_flags |= PCRE_HASTHEN;
5069              }              }
5070    
5071            else            else
# Line 4818  for (;; ptr++) Line 5075  for (;; ptr++)
5075                *errorcodeptr = ERR59;                *errorcodeptr = ERR59;
5076                goto FAILED;                goto FAILED;
5077                }                }
5078              *code++ = verbs[i].op_arg;              *code = verbs[i].op_arg;
5079                if (*code++ == OP_THEN_ARG) cd->external_flags |= PCRE_HASTHEN;
5080              *code++ = arglen;              *code++ = arglen;
5081              memcpy(code, arg, arglen);              memcpy(code, arg, arglen);
5082              code += arglen;              code += arglen;
# Line 5012  for (;; ptr++) Line 5270  for (;; ptr++)
5270          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5271    
5272          else if ((i = find_parens(cd, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5273                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0, utf8)) > 0)
5274            {            {
5275            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5276            code[1+LINK_SIZE]++;            code[1+LINK_SIZE]++;
# Line 5080  for (;; ptr++) Line 5338  for (;; ptr++)
5338          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5339          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5340          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5341            cd->assert_depth += 1;
5342          ptr++;          ptr++;
5343          break;          break;
5344    
# Line 5094  for (;; ptr++) Line 5353  for (;; ptr++)
5353            continue;            continue;
5354            }            }
5355          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5356            cd->assert_depth += 1;
5357          break;          break;
5358    
5359    
# Line 5103  for (;; ptr++) Line 5363  for (;; ptr++)
5363            {            {
5364            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5365            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5366              cd->assert_depth += 1;
5367            ptr += 2;            ptr += 2;
5368            break;            break;
5369    
5370            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5371            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5372              cd->assert_depth += 1;
5373            ptr += 2;            ptr += 2;
5374            break;            break;
5375    
# Line 5313  for (;; ptr++) Line 5575  for (;; ptr++)
5575          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5576          namelen = (int)(ptr - name);          namelen = (int)(ptr - name);
5577    
5578          /* In the pre-compile phase, do a syntax check and set a dummy          /* In the pre-compile phase, do a syntax check. We used to just set
5579          reference number. */          a dummy reference number, because it was not used in the first pass.
5580            However, with the change of recursive back references to be atomic,
5581            we have to look for the number so that this state can be identified, as
5582            otherwise the incorrect length is computed. If it's not a backwards
5583            reference, the dummy number will do. */
5584    
5585          if (lengthptr != NULL)          if (lengthptr != NULL)
5586            {            {
5587              const uschar *temp;
5588    
5589            if (namelen == 0)            if (namelen == 0)
5590              {              {
5591              *errorcodeptr = ERR62;              *errorcodeptr = ERR62;
# Line 5333  for (;; ptr++) Line 5601  for (;; ptr++)
5601              *errorcodeptr = ERR48;              *errorcodeptr = ERR48;
5602              goto FAILED;              goto FAILED;
5603              }              }
5604            recno = 0;  
5605              /* The name table does not exist in the first pass, so we cannot
5606              do a simple search as in the code below. Instead, we have to scan the
5607              pattern to find the number. It is important that we scan it only as
5608              far as we have got because the syntax of named subpatterns has not
5609              been checked for the rest of the pattern, and find_parens() assumes
5610              correct syntax. In any case, it's a waste of resources to scan
5611              further. We stop the scan at the current point by temporarily
5612              adjusting the value of cd->endpattern. */
5613    
5614              temp = cd->end_pattern;
5615              cd->end_pattern = ptr;
5616              recno = find_parens(cd, name, namelen,
5617                (options & PCRE_EXTENDED) != 0, utf8);
5618              cd->end_pattern = temp;
5619              if (recno < 0) recno = 0;    /* Forward ref; set dummy number */
5620            }            }
5621    
5622          /* In the real compile, seek the name in the table. We check the name          /* In the real compile, seek the name in the table. We check the name
# Line 5358  for (;; ptr++) Line 5641  for (;; ptr++)
5641              }              }
5642            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5643                      find_parens(cd, name, namelen,                      find_parens(cd, name, namelen,
5644                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0, utf8)) <= 0)
5645              {              {
5646              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
5647              goto FAILED;              goto FAILED;
# Line 5469  for (;; ptr++) Line 5752  for (;; ptr++)
5752              if (called == NULL)              if (called == NULL)
5753                {                {
5754                if (find_parens(cd, NULL, recno,                if (find_parens(cd, NULL, recno,
5755                      (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0, utf8) < 0)
5756                  {                  {
5757                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5758                  goto FAILED;                  goto FAILED;
# Line 5477  for (;; ptr++) Line 5760  for (;; ptr++)
5760    
5761                /* Fudge the value of "called" so that when it is inserted as an                /* Fudge the value of "called" so that when it is inserted as an
5762                offset below, what it actually inserted is the reference number                offset below, what it actually inserted is the reference number
5763                of the group. */                of the group. Then remember the forward reference. */
5764    
5765                called = cd->start_code + recno;                called = cd->start_code + recno;
5766                PUTINC(cd->hwm, 0, (int)(code + 2 + LINK_SIZE - cd->start_code));                PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
5767                }                }
5768    
5769              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
5770              this is a recursive call. We check to see if this is a left              this is a recursive call. We check to see if this is a left
5771              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. We
5772                must not, however, do this check if we are in a conditional
5773                subpattern because the condition might be testing for recursion in
5774                a pattern such as /(?(R)a+|(?R)b)/, which is perfectly valid.
5775                Forever loops are also detected at runtime, so those that occur in
5776                conditional subpatterns will be picked up then. */
5777    
5778              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 && cond_depth <= 0 &&
5779                       could_be_empty(called, code, bcptr, utf8, cd))                       could_be_empty(called, code, bcptr, utf8, cd))
5780                {                {
5781                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
# Line 5495  for (;; ptr++) Line 5783  for (;; ptr++)
5783                }                }
5784              }              }
5785    
5786            /* Insert the recursion/subroutine item, automatically wrapped inside            /* Insert the recursion/subroutine item. */
           "once" brackets. Set up a "previous group" length so that a  
           subsequent quantifier will work. */  
   
           *code = OP_ONCE;  
           PUT(code, 1, 2 + 2*LINK_SIZE);  
           code += 1 + LINK_SIZE;  
5787    
5788            *code = OP_RECURSE;            *code = OP_RECURSE;
5789            PUT(code, 1, (int)(called - cd->start_code));            PUT(code, 1, (int)(called - cd->start_code));
5790            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
   
           *code = OP_KET;  
           PUT(code, 1, 2 + 2*LINK_SIZE);  
           code += 1 + LINK_SIZE;  
   
           length_prevgroup = 3 + 3*LINK_SIZE;  
5791            }            }
5792    
5793          /* Can't determine a first byte now */          /* Can't determine a first byte now */
# Line 5572  for (;; ptr++) Line 5848  for (;; ptr++)
5848          is necessary to ensure we correctly detect the start of the pattern in          is necessary to ensure we correctly detect the start of the pattern in
5849          both phases.          both phases.
5850    
5851          If we are not at the pattern start, compile code to change the ims          If we are not at the pattern start, reset the greedy defaults and the
5852          options if this setting actually changes any of them, and reset the          case value for firstbyte and reqbyte. */
         greedy defaults and the case value for firstbyte and reqbyte. */  
5853    
5854          if (*ptr == CHAR_RIGHT_PARENTHESIS)          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5855            {            {
# Line 5585  for (;; ptr++) Line 5860  for (;; ptr++)
5860              }              }
5861            else            else
5862              {              {
             if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))  
               {  
               *code++ = OP_OPT;  
               *code++ = newoptions & PCRE_IMS;  
               }  
5863              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5864              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5865              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5866              }              }
5867    
5868            /* Change options at this level, and pass them back for use            /* Change options at this level, and pass them back for use
5869            in subsequent branches. When not at the start of the pattern, this            in subsequent branches. */
           information is also necessary so that a resetting item can be  
           compiled at the end of a group (if we are in a group). */  
5870    
5871            *optionsptr = options = newoptions;            *optionsptr = options = newoptions;
5872            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
# Line 5634  for (;; ptr++) Line 5902  for (;; ptr++)
5902        skipbytes = 2;        skipbytes = 2;
5903        }        }
5904    
5905      /* Process nested bracketed regex. Assertions may not be repeated, but      /* Process nested bracketed regex. Assertions used not to be repeatable,
5906      other kinds can be. All their opcodes are >= OP_ONCE. We copy code into a      but this was changed for Perl compatibility, so all kinds can now be
5907      non-register variable in order to be able to pass its address because some      repeated. We copy code into a non-register variable (tempcode) in order to
5908      compilers complain otherwise. Pass in a new setting for the ims options if      be able to pass its address because some compilers complain otherwise. */
     they have changed. */  
5909    
5910      previous = (bravalue >= OP_ONCE)? code : NULL;      previous = code;                      /* For handling repetition */
5911      *code = bravalue;      *code = bravalue;
5912      tempcode = code;      tempcode = code;
5913      tempreqvary = cd->req_varyopt;     /* Save value before bracket */      tempreqvary = cd->req_varyopt;        /* Save value before bracket */
5914      length_prevgroup = 0;              /* Initialize for pre-compile phase */      tempbracount = cd->bracount;          /* Save value before bracket */
5915        length_prevgroup = 0;                 /* Initialize for pre-compile phase */
5916    
5917      if (!compile_regex(      if (!compile_regex(
5918           newoptions,                   /* The complete new option state */           newoptions,                      /* The complete new option state */
5919           options & PCRE_IMS,           /* The previous ims option state */           &tempcode,                       /* Where to put code (updated) */
5920           &tempcode,                    /* Where to put code (updated) */           &ptr,                            /* Input pointer (updated) */
5921           &ptr,                         /* Input pointer (updated) */           errorcodeptr,                    /* Where to put an error message */
          errorcodeptr,                 /* Where to put an error message */  
5922           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
5923            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
5924           reset_bracount,               /* True if (?| group */           reset_bracount,                  /* True if (?| group */
5925           skipbytes,                    /* Skip over bracket number */           skipbytes,                       /* Skip over bracket number */
5926           &subfirstbyte,                /* For possible first char */           cond_depth +
5927           &subreqbyte,                  /* For possible last char */             ((bravalue == OP_COND)?1:0),   /* Depth of condition subpatterns */
5928           bcptr,                        /* Current branch chain */           &subfirstbyte,                   /* For possible first char */
5929           cd,                           /* Tables block */           &subreqbyte,                     /* For possible last char */
5930           (lengthptr == NULL)? NULL :   /* Actual compile phase */           bcptr,                           /* Current branch chain */
5931             &length_prevgroup           /* Pre-compile phase */           cd,                              /* Tables block */
5932             (lengthptr == NULL)? NULL :      /* Actual compile phase */
5933               &length_prevgroup              /* Pre-compile phase */
5934           ))           ))
5935        goto FAILED;        goto FAILED;
5936    
5937        /* If this was an atomic group and there are no capturing groups within it,
5938        generate OP_ONCE_NC instead of OP_ONCE. */
5939    
5940        if (bravalue == OP_ONCE && cd->bracount <= tempbracount)
5941          *code = OP_ONCE_NC;
5942    
5943        if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
5944          cd->assert_depth -= 1;
5945    
5946      /* At the end of compiling, code is still pointing to the start of the      /* At the end of compiling, code is still pointing to the start of the
5947      group, while tempcode has been updated to point past the end of the group      group, while tempcode has been updated to point past the end of the group.
5948      and any option resetting that may follow it. The pattern pointer (ptr)      The pattern pointer (ptr) is on the bracket.
     is on the bracket. */  
5949    
5950      /* If this is a conditional bracket, check that there are no more than      If this is a conditional bracket, check that there are no more than
5951      two branches in the group, or just one if it's a DEFINE group. We do this      two branches in the group, or just one if it's a DEFINE group. We do this
5952      in the real compile phase, not in the pre-pass, where the whole group may      in the real compile phase, not in the pre-pass, where the whole group may
5953      not be available. */      not be available. */
# Line 5735  for (;; ptr++) Line 6012  for (;; ptr++)
6012          goto FAILED;          goto FAILED;
6013          }          }
6014        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
6015        *code++ = OP_BRA;        code++;   /* This already contains bravalue */
6016        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
6017        *code++ = OP_KET;        *code++ = OP_KET;
6018        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
# Line 5903  for (;; ptr++) Line 6180  for (;; ptr++)
6180          }          }
6181    
6182        /* \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).
6183        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax).  */
6184    
6185        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||        if (-c == ESC_k)
           ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))  
6186          {          {
6187            if ((ptr[1] != CHAR_LESS_THAN_SIGN &&
6188              ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))
6189              {
6190              *errorcodeptr = ERR69;
6191              break;
6192              }
6193          is_recurse = FALSE;          is_recurse = FALSE;
6194          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
6195            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
# Line 5927  for (;; ptr++) Line 6209  for (;; ptr++)
6209          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
6210          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
6211          previous = code;          previous = code;
6212          *code++ = OP_REF;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_REFI : OP_REF;
6213          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
6214          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
6215          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
# Line 6035  for (;; ptr++) Line 6317  for (;; ptr++)
6317    
6318      ONE_CHAR:      ONE_CHAR:
6319      previous = code;      previous = code;
6320      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARI : OP_CHAR;
6321      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
6322    
6323      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
# Line 6064  for (;; ptr++) Line 6346  for (;; ptr++)
6346        else firstbyte = reqbyte = REQ_NONE;        else firstbyte = reqbyte = REQ_NONE;
6347        }        }
6348    
6349      /* firstbyte was previously set; we can set reqbyte only the length is      /* firstbyte was previously set; we can set reqbyte only if the length is
6350      1 or the matching is caseful. */      1 or the matching is caseful. */
6351    
6352      else      else
# Line 6099  return FALSE; Line 6381  return FALSE;
6381  /* On entry, ptr is pointing past the bracket character, but on return it  /* On entry, ptr is pointing past the bracket character, but on return it
6382  points to the closing bracket, or vertical bar, or end of string. The code  points to the closing bracket, or vertical bar, or end of string. The code
6383  variable is pointing at the byte into which the BRA operator has been stored.  variable is pointing at the byte into which the BRA operator has been stored.
 If the ims options are changed at the start (for a (?ims: group) or during any  
 branch, we need to insert an OP_OPT item at the start of every following branch  
 to ensure they get set correctly at run time, and also pass the new options  
 into every subsequent branch compile.  
   
6384  This function is used during the pre-compile phase when we are trying to find  This function is used during the pre-compile phase when we are trying to find
6385  out the amount of memory needed, as well as during the real compile phase. The  out the amount of memory needed, as well as during the real compile phase. The
6386  value of lengthptr distinguishes the two phases.  value of lengthptr distinguishes the two phases.
6387    
6388  Arguments:  Arguments:
6389    options        option bits, including any changes for this subpattern    options        option bits, including any changes for this subpattern
   oldims         previous settings of ims option bits  
6390    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
6391    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
6392    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
6393    lookbehind     TRUE if this is a lookbehind assertion    lookbehind     TRUE if this is a lookbehind assertion
6394    reset_bracount TRUE to reset the count for each branch    reset_bracount TRUE to reset the count for each branch
6395    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes      skip this many bytes at start (for brackets and OP_COND)
6396      cond_depth     depth of nesting for conditional subpatterns
6397    firstbyteptr   place to put the first required character, or a negative number    firstbyteptr   place to put the first required character, or a negative number
6398    reqbyteptr     place to put the last required character, or a negative number    reqbyteptr     place to put the last required character, or a negative number
6399    bcptr          pointer to the chain of currently open branches    bcptr          pointer to the chain of currently open branches
# Line 6128  Returns: TRUE on success Line 6405  Returns: TRUE on success
6405  */  */
6406    
6407  static BOOL  static BOOL
6408  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,
6409    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
6410    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,    int cond_depth, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
6411    int *lengthptr)    compile_data *cd, int *lengthptr)
6412  {  {
6413  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
6414  uschar *code = *codeptr;  uschar *code = *codeptr;
# Line 6145  int branchfirstbyte, branchreqbyte; Line 6422  int branchfirstbyte, branchreqbyte;
6422  int length;  int length;
6423  int orig_bracount;  int orig_bracount;
6424  int max_bracount;  int max_bracount;
 int old_external_options = cd->external_options;  
6425  branch_chain bc;  branch_chain bc;
6426    
6427  bc.outer = bcptr;  bc.outer = bcptr;
# Line 6169  pre-compile phase to find out whether an Line 6445  pre-compile phase to find out whether an
6445    
6446  /* If this is a capturing subpattern, add to the chain of open capturing items  /* If this is a capturing subpattern, add to the chain of open capturing items
6447  so that we can detect them if (*ACCEPT) is encountered. This is also used to  so that we can detect them if (*ACCEPT) is encountered. This is also used to
6448  detect groups that contain recursive back references to themselves. */  detect groups that contain recursive back references to themselves. Note that
6449    only OP_CBRA need be tested here; changing this opcode to one of its variants,
6450    e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
6451    
6452  if (*code == OP_CBRA)  if (*code == OP_CBRA)
6453    {    {
# Line 6195  for (;;) Line 6473  for (;;)
6473    
6474    if (reset_bracount) cd->bracount = orig_bracount;    if (reset_bracount) cd->bracount = orig_bracount;
6475    
   /* Handle a change of ims options at the start of the branch */  
   
   if ((options & PCRE_IMS) != oldims)  
     {  
     *code++ = OP_OPT;  
     *code++ = options & PCRE_IMS;  
     length += 2;  
     }  
   
6476    /* Set up dummy OP_REVERSE if lookbehind assertion */    /* Set up dummy OP_REVERSE if lookbehind assertion */
6477    
6478    if (lookbehind)    if (lookbehind)
# Line 6218  for (;;) Line 6487  for (;;)
6487    into the length. */    into the length. */
6488    
6489    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,
6490          &branchreqbyte, &bc, cd, (lengthptr == NULL)? NULL : &length))          &branchreqbyte, &bc, cond_depth, cd,
6491            (lengthptr == NULL)? NULL : &length))
6492      {      {
6493      *ptrptr = ptr;      *ptrptr = ptr;
6494      return FALSE;      return FALSE;
6495      }      }
6496    
   /* If the external options have changed during this branch, it means that we  
   are at the top level, and a leading option setting has been encountered. We  
   need to re-set the original option values to take account of this so that,  
   during the pre-compile phase, we know to allow for a re-set at the start of  
   subsequent branches. */  
   
   if (old_external_options != cd->external_options)  
     oldims = cd->external_options & PCRE_IMS;  
   
6497    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6498    has fewer than the rest. */    has fewer than the rest. */
6499    
# Line 6293  for (;;) Line 6554  for (;;)
6554        {        {
6555        int fixed_length;        int fixed_length;
6556        *code = OP_END;        *code = OP_END;
6557        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);        fixed_length = find_fixedlength(last_branch,  (options & PCRE_UTF8) != 0,
6558            FALSE, cd);
6559        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6560        if (fixed_length == -3)        if (fixed_length == -3)
6561          {          {
# Line 6314  for (;;) Line 6576  for (;;)
6576    of offsets, with the field in the BRA item now becoming an offset to the    of offsets, with the field in the BRA item now becoming an offset to the
6577    first alternative. If there are no alternatives, it points to the end of the    first alternative. If there are no alternatives, it points to the end of the
6578    group. The length in the terminating ket is always the length of the whole    group. The length in the terminating ket is always the length of the whole
6579    bracketed item. If any of the ims options were changed inside the group,    bracketed item. Return leaving the pointer at the terminating char. */
   compile a resetting op-code following, except at the very end of the pattern.  
   Return leaving the pointer at the terminating char. */  
6580    
6581    if (*ptr != CHAR_VERTICAL_LINE)    if (*ptr != CHAR_VERTICAL_LINE)
6582      {      {
# Line 6360  for (;;) Line 6620  for (;;)
6620        cd->open_caps = cd->open_caps->next;        cd->open_caps = cd->open_caps->next;
6621        }        }
6622    
     /* Reset options if needed. */  
   
     if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)  
       {  
       *code++ = OP_OPT;  
       *code++ = oldims;  
       length += 2;  
       }  
   
6623      /* Retain the highest bracket number, in case resetting was used. */      /* Retain the highest bracket number, in case resetting was used. */
6624    
6625      cd->bracount = max_bracount;      cd->bracount = max_bracount;
# Line 6428  for (;;) Line 6679  for (;;)
6679  /* Try to find out if this is an anchored regular expression. Consider each  /* Try to find out if this is an anchored regular expression. Consider each
6680  alternative branch. If they all start with OP_SOD or OP_CIRC, or with a bracket  alternative branch. If they all start with OP_SOD or OP_CIRC, or with a bracket
6681  all of whose alternatives start with OP_SOD or OP_CIRC (recurse ad lib), then  all of whose alternatives start with OP_SOD or OP_CIRC (recurse ad lib), then
6682  it's anchored. However, if this is a multiline pattern, then only OP_SOD  it's anchored. However, if this is a multiline pattern, then only OP_SOD will
6683  counts, since OP_CIRC can match in the middle.  be found, because ^ generates OP_CIRCM in that mode.
6684    
6685  We can also consider a regex to be anchored if OP_SOM starts all its branches.  We can also consider a regex to be anchored if OP_SOM starts all its branches.
6686  This is the code for \G, which means "match at start of match position, taking  This is the code for \G, which means "match at start of match position, taking
# Line 6450  of the more common cases more precisely. Line 6701  of the more common cases more precisely.
6701    
6702  Arguments:  Arguments:
6703    code           points to start of expression (the bracket)    code           points to start of expression (the bracket)
   options        points to the options setting  
6704    bracket_map    a bitmap of which brackets we are inside while testing; this    bracket_map    a bitmap of which brackets we are inside while testing; this
6705                    handles up to substring 31; after that we just have to take                    handles up to substring 31; after that we just have to take
6706                    the less precise approach                    the less precise approach
# Line 6460  Returns: TRUE or FALSE Line 6710  Returns: TRUE or FALSE
6710  */  */
6711    
6712  static BOOL  static BOOL
6713  is_anchored(register const uschar *code, int *options, unsigned int bracket_map,  is_anchored(register const uschar *code, unsigned int bracket_map,
6714    unsigned int backref_map)    unsigned int backref_map)
6715  {  {
6716  do {  do {
6717     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6718       options, PCRE_MULTILINE, FALSE);       FALSE);
6719     register int op = *scode;     register int op = *scode;
6720    
6721     /* Non-capturing brackets */     /* Non-capturing brackets */
6722    
6723     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6724           op == OP_SBRA || op == OP_SBRAPOS)
6725       {       {
6726       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6727       }       }
6728    
6729     /* Capturing brackets */     /* Capturing brackets */
6730    
6731     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6732                op == OP_SCBRA || op == OP_SCBRAPOS)
6733       {       {
6734       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6735       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
6736       if (!is_anchored(scode, options, new_map, backref_map)) return FALSE;       if (!is_anchored(scode, new_map, backref_map)) return FALSE;
6737       }       }
6738    
6739     /* Other brackets */     /* Other brackets */
6740    
6741     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC ||
6742                op == OP_COND)
6743       {       {
6744       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6745       }       }
6746    
6747     /* .* is not anchored unless DOTALL is set (which generates OP_ALLANY) and     /* .* is not anchored unless DOTALL is set (which generates OP_ALLANY) and
# Line 6503  do { Line 6756  do {
6756    
6757     /* Check for explicit anchoring */     /* Check for explicit anchoring */
6758    
6759     else if (op != OP_SOD && op != OP_SOM &&     else if (op != OP_SOD && op != OP_SOM && op != OP_CIRC) return FALSE;
            ((*options & PCRE_MULTILINE) != 0 || op != OP_CIRC))  
      return FALSE;  
6760     code += GET(code, 1);     code += GET(code, 1);
6761     }     }
6762  while (*code == OP_ALT);   /* Loop for each alternative */  while (*code == OP_ALT);   /* Loop for each alternative */
# Line 6541  is_startline(const uschar *code, unsigne Line 6792  is_startline(const uschar *code, unsigne
6792  {  {
6793  do {  do {
6794     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6795       NULL, 0, FALSE);       FALSE);
6796     register int op = *scode;     register int op = *scode;
6797    
6798     /* 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
# Line 6568  do { Line 6819  do {
6819         scode += 1 + LINK_SIZE;         scode += 1 + LINK_SIZE;
6820         break;         break;
6821         }         }
6822       scode = first_significant_code(scode, NULL, 0, FALSE);       scode = first_significant_code(scode, FALSE);
6823       op = *scode;       op = *scode;
6824       }       }
6825    
6826     /* Non-capturing brackets */     /* Non-capturing brackets */
6827    
6828     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6829           op == OP_SBRA || op == OP_SBRAPOS)
6830       {       {
6831       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6832       }       }
6833    
6834     /* Capturing brackets */     /* Capturing brackets */
6835    
6836     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6837                op == OP_SCBRA || op == OP_SCBRAPOS)
6838       {       {
6839       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6840       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
# Line 6590  do { Line 6843  do {
6843    
6844     /* Other brackets */     /* Other brackets */
6845    
6846     else if (op == OP_ASSERT || op == OP_ONCE)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC)
6847       {       {
6848       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6849       }       }
# Line 6605  do { Line 6858  do {
6858    
6859     /* Check for explicit circumflex */     /* Check for explicit circumflex */
6860    
6861     else if (op != OP_CIRC) return FALSE;     else if (op != OP_CIRC && op != OP_CIRCM) return FALSE;
6862    
6863     /* Move on to the next alternative */     /* Move on to the next alternative */
6864    
# Line 6631  we return that char, otherwise -1. Line 6884  we return that char, otherwise -1.
6884    
6885  Arguments:  Arguments:
6886    code       points to start of expression (the bracket)    code       points to start of expression (the bracket)
   options    pointer to the options (used to check casing changes)  
6887    inassert   TRUE if in an assertion    inassert   TRUE if in an assertion
6888    
6889  Returns:     -1 or the fixed first char  Returns:     -1 or the fixed first char
6890  */  */
6891    
6892  static int  static int
6893  find_firstassertedchar(const uschar *code, int *options, BOOL inassert)  find_firstassertedchar(const uschar *code, BOOL inassert)
6894  {  {
6895  register int c = -1;  register int c = -1;
6896  do {  do {
6897     int d;     int d;
6898     const uschar *scode =     int xl = (*code == OP_CBRA || *code == OP_SCBRA ||
6899       first_significant_code(code + 1+LINK_SIZE, options, PCRE_CASELESS, TRUE);               *code == OP_CBRAPOS || *code == OP_SCBRAPOS)? 2:0;
6900       const uschar *scode = first_significant_code(code + 1+LINK_SIZE + xl, TRUE);
6901     register int op = *scode;     register int op = *scode;
6902    
6903     switch(op)     switch(op)
# Line 6653  do { Line 6906  do {
6906       return -1;       return -1;
6907    
6908       case OP_BRA:       case OP_BRA:
6909         case OP_BRAPOS:
6910       case OP_CBRA:       case OP_CBRA:
6911         case OP_SCBRA:
6912         case OP_CBRAPOS:
6913         case OP_SCBRAPOS:
6914       case OP_ASSERT:       case OP_ASSERT:
6915       case OP_ONCE:       case OP_ONCE:
6916         case OP_ONCE_NC:
6917       case OP_COND:       case OP_COND:
6918       if ((d = find_firstassertedchar(scode, options, op == OP_ASSERT)) < 0)       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)
6919         return -1;         return -1;
6920       if (c < 0) c = d; else if (c != d) return -1;       if (c < 0) c = d; else if (c != d) return -1;
6921       break;       break;
6922    
6923       case OP_EXACT:       /* Fall through */       case OP_EXACT:
6924       scode += 2;       scode += 2;
6925         /* Fall through */
6926    
6927       case OP_CHAR:       case OP_CHAR:
      case OP_CHARNC:  
6928       case OP_PLUS:       case OP_PLUS:
6929       case OP_MINPLUS:       case OP_MINPLUS:
6930       case OP_POSPLUS:       case OP_POSPLUS:
6931       if (!inassert) return -1;       if (!inassert) return -1;
6932       if (c < 0)       if (c < 0) c = scode[1];
6933         {         else if (c != scode[1]) return -1;
6934         c = scode[1];       break;
6935         if ((*options & PCRE_CASELESS) != 0) c |= REQ_CASELESS;  
6936         }       case OP_EXACTI:
6937       else if (c != scode[1]) return -1;       scode += 2;
6938         /* Fall through */
6939    
6940         case OP_CHARI:
6941         case OP_PLUSI:
6942         case OP_MINPLUSI:
6943         case OP_POSPLUSI:
6944         if (!inassert) return -1;
6945         if (c < 0) c = scode[1] | REQ_CASELESS;
6946           else if (c != scode[1]) return -1;
6947       break;       break;
6948       }       }
6949    
# Line 6799  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7066  while (ptr[skipatstart] == CHAR_LEFT_PAR
7066      { skipatstart += 7; options |= PCRE_UTF8; continue; }      { skipatstart += 7; options |= PCRE_UTF8; continue; }
7067    else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)
7068      { skipatstart += 6; options |= PCRE_UCP; continue; }      { skipatstart += 6; options |= PCRE_UCP; continue; }
7069      else if (strncmp((char *)(ptr+skipatstart+2), STRING_NO_START_OPT_RIGHTPAR, 13) == 0)
7070        { skipatstart += 15; options |= PCRE_NO_START_OPTIMIZE; continue; }
7071    
7072    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
7073      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
# Line 6825  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7094  while (ptr[skipatstart] == CHAR_LEFT_PAR
7094    
7095  utf8 = (options & PCRE_UTF8) != 0;  utf8 = (options & PCRE_UTF8) != 0;
7096    
7097  /* Can't support UTF8 unless PCRE has been compiled to include the code. */  /* Can't support UTF8 unless PCRE has been compiled to include the code. The
7098    return of an error code from _pcre_valid_utf8() is a new feature, introduced in
7099    release 8.13. It is passed back from pcre_[dfa_]exec(), but at the moment is
7100    not used here. */
7101    
7102  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
7103  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
7104       (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)       (errorcode = _pcre_valid_utf8((USPTR)pattern, -1, erroroffset)) != 0)
7105    {    {
7106    errorcode = ERR44;    errorcode = ERR44;
7107    goto PCRE_EARLY_ERROR_RETURN2;    goto PCRE_EARLY_ERROR_RETURN2;
# Line 6854  if ((options & PCRE_UCP) != 0) Line 7126  if ((options & PCRE_UCP) != 0)
7126    
7127  /* Check validity of \R options. */  /* Check validity of \R options. */
7128    
7129  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  if ((options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) ==
7130         (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
7131    {    {
7132    case 0:    errorcode = ERR56;
7133    case PCRE_BSR_ANYCRLF:    goto PCRE_EARLY_ERROR_RETURN;
   case PCRE_BSR_UNICODE:  
   break;  
   default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;  
7134    }    }
7135    
7136  /* Handle different types of newline. The three bits give seven cases. The  /* Handle different types of newline. The three bits give seven cases. The
# Line 6945  outside can help speed up starting point Line 7215  outside can help speed up starting point
7215  ptr += skipatstart;  ptr += skipatstart;
7216  code = cworkspace;  code = cworkspace;
7217  *code = OP_BRA;  *code = OP_BRA;
7218  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
7219    &code, &ptr, &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd,    FALSE, 0, 0, &firstbyte, &reqbyte, NULL, cd, &length);
   &length);  
7220  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
7221    
7222  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,
# Line 7001  field; this time it's used for rememberi Line 7270  field; this time it's used for rememberi
7270  */  */
7271    
7272  cd->final_bracount = cd->bracount;  /* Save for checking forward references */  cd->final_bracount = cd->bracount;  /* Save for checking forward references */
7273    cd->assert_depth = 0;
7274  cd->bracount = 0;  cd->bracount = 0;
7275  cd->names_found = 0;  cd->names_found = 0;
7276  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
# Line 7019  of the function here. */ Line 7289  of the function here. */
7289  ptr = (const uschar *)pattern + skipatstart;  ptr = (const uschar *)pattern + skipatstart;
7290  code = (uschar *)codestart;  code = (uschar *)codestart;
7291  *code = OP_BRA;  *code = OP_BRA;
7292  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,  (void)compile_regex(re->options, &code, &ptr, &errorcode, FALSE, FALSE, 0, 0,
7293    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);    &firstbyte, &reqbyte, NULL, cd, NULL);
7294  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
7295  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
7296  re->flags = cd->external_flags;  re->flags = cd->external_flags;
7297    
7298  if (cd->had_accept) reqbyte = -1;   /* Must disable after (*ACCEPT) */  if (cd->had_accept) reqbyte = REQ_NONE;   /* Must disable after (*ACCEPT) */
7299    
7300  /* If not reached end of pattern on success, there's an excess bracket. */  /* If not reached end of pattern on success, there's an excess bracket. */
7301    
# Line 7086  if (cd->check_lookbehind) Line 7356  if (cd->check_lookbehind)
7356        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
7357        int end_op = *be;        int end_op = *be;
7358        *be = OP_END;        *be = OP_END;
7359        fixed_length = find_fixedlength(cc, re->options, TRUE, cd);        fixed_length = find_fixedlength(cc, (re->options & PCRE_UTF8) != 0, TRUE,
7360            cd);
7361        *be = end_op;        *be = end_op;
7362        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
7363        if (fixed_length < 0)        if (fixed_length < 0)
# Line 7125  start with ^. and also when all branches Line 7396  start with ^. and also when all branches
7396    
7397  if ((re->options & PCRE_ANCHORED) == 0)  if ((re->options & PCRE_ANCHORED) == 0)
7398    {    {
7399    int temp_options = re->options;   /* May get changed during these scans */    if (is_anchored(codestart, 0, cd->backref_map))
   if (is_anchored(codestart, &temp_options, 0, cd->backref_map))  
7400      re->options |= PCRE_ANCHORED;      re->options |= PCRE_ANCHORED;
7401    else    else
7402      {      {
7403      if (firstbyte < 0)      if (firstbyte < 0)
7404        firstbyte = find_firstassertedchar(codestart, &temp_options, FALSE);        firstbyte = find_firstassertedchar(codestart, FALSE);
7405      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */
7406        {        {
7407        int ch = firstbyte & 255;        int ch = firstbyte & 255;

Legend:
Removed from v.544  
changed lines
  Added in v.723

webmaster@exim.org
ViewVC Help
Powered by ViewVC 1.1.12