/[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 574 by ph10, Sat Nov 20 17:47:27 2010 UTC revision 640 by ph10, Mon Jul 25 10:50:28 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"    "\\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 545  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 *, branch_chain *, compile_data *, int *);
551    
552    
553    
# Line 577  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 647  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 664  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 857  else Line 893  else
893      if (c > 127)  /* Excludes all non-ASCII in either mode */      if (c > 127)  /* Excludes all non-ASCII in either mode */
894        {        {
895        *errorcodeptr = ERR68;        *errorcodeptr = ERR68;
896        break;        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 */
# Line 885  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 997  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 1105  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 1116  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    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 1209  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 1285  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      ptr++;      ptr++;
1307      while (*ptr != 0)      while (*ptr != 0)
1308        {        {
1309        if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }        if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1310        ptr++;        ptr++;
1311  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1312        if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;        if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1313  #endif  #endif
1314        }        }
# Line 1348  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    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  */  */
# Line 1384  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 1461  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 1472  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 1489  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_COND:      case OP_COND:
1510      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1511      if (d < 0) return d;      if (d < 0) return d;
1512      branchlength += d;      branchlength += d;
1513      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1502  for (;;) Line 1516  for (;;)
1516    
1517      /* 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
1518      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
1519      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.
1520        Note that we must not include the OP_KETRxxx opcodes here, because they
1521        all imply an unlimited repeat. */
1522    
1523      case OP_ALT:      case OP_ALT:
1524      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1525      case OP_END:      case OP_END:
1526      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1527        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
# Line 1525  for (;;) Line 1539  for (;;)
1539      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1540      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1541      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                /* Recursion */
1542      d = find_fixedlength(cs + 2, options, atend, cd);      d = find_fixedlength(cs + 2, utf8, atend, cd);
1543      if (d < 0) return d;      if (d < 0) return d;
1544      branchlength += d;      branchlength += d;
1545      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1548  for (;;) Line 1562  for (;;)
1562      case OP_RREF:      case OP_RREF:
1563      case OP_NRREF:      case OP_NRREF:
1564      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
1565      case OP_CALLOUT:      case OP_CALLOUT:
1566      case OP_SOD:      case OP_SOD:
1567      case OP_SOM:      case OP_SOM:
# Line 1556  for (;;) Line 1569  for (;;)
1569      case OP_EOD:      case OP_EOD:
1570      case OP_EODN:      case OP_EODN:
1571      case OP_CIRC:      case OP_CIRC:
1572        case OP_CIRCM:
1573      case OP_DOLL:      case OP_DOLL:
1574        case OP_DOLLM:
1575      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1576      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1577      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 1565  for (;;) Line 1580  for (;;)
1580      /* Handle literal characters */      /* Handle literal characters */
1581    
1582      case OP_CHAR:      case OP_CHAR:
1583      case OP_CHARNC:      case OP_CHARI:
1584      case OP_NOT:      case OP_NOT:
1585        case OP_NOTI:
1586      branchlength++;      branchlength++;
1587      cc += 2;      cc += 2;
1588  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1589      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];  
1590  #endif  #endif
1591      break;      break;
1592    
# Line 1582  for (;;) Line 1597  for (;;)
1597      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1598      cc += 4;      cc += 4;
1599  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1600      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];  
1601  #endif  #endif
1602      break;      break;
1603    
# Line 1685  _pcre_find_bracket(const uschar *code, B Line 1699  _pcre_find_bracket(const uschar *code, B
1699  for (;;)  for (;;)
1700    {    {
1701    register int c = *code;    register int c = *code;
1702    
1703    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1704    
1705    /* 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 1703  for (;;) Line 1718  for (;;)
1718    
1719    /* Handle capturing bracket */    /* Handle capturing bracket */
1720    
1721    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1722               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1723      {      {
1724      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1725      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1761  for (;;) Line 1777  for (;;)
1777      if (utf8) switch(c)      if (utf8) switch(c)
1778        {        {
1779        case OP_CHAR:        case OP_CHAR:
1780        case OP_CHARNC:        case OP_CHARI:
1781        case OP_EXACT:        case OP_EXACT:
1782          case OP_EXACTI:
1783        case OP_UPTO:        case OP_UPTO:
1784          case OP_UPTOI:
1785        case OP_MINUPTO:        case OP_MINUPTO:
1786          case OP_MINUPTOI:
1787        case OP_POSUPTO:        case OP_POSUPTO:
1788          case OP_POSUPTOI:
1789        case OP_STAR:        case OP_STAR:
1790          case OP_STARI:
1791        case OP_MINSTAR:        case OP_MINSTAR:
1792          case OP_MINSTARI:
1793        case OP_POSSTAR:        case OP_POSSTAR:
1794          case OP_POSSTARI:
1795        case OP_PLUS:        case OP_PLUS:
1796          case OP_PLUSI:
1797        case OP_MINPLUS:        case OP_MINPLUS:
1798          case OP_MINPLUSI:
1799        case OP_POSPLUS:        case OP_POSPLUS:
1800          case OP_POSPLUSI:
1801        case OP_QUERY:        case OP_QUERY:
1802          case OP_QUERYI:
1803        case OP_MINQUERY:        case OP_MINQUERY:
1804          case OP_MINQUERYI:
1805        case OP_POSQUERY:        case OP_POSQUERY:
1806          case OP_POSQUERYI:
1807        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1808        break;        break;
1809        }        }
# Line 1867  for (;;) Line 1896  for (;;)
1896      if (utf8) switch(c)      if (utf8) switch(c)
1897        {        {
1898        case OP_CHAR:        case OP_CHAR:
1899        case OP_CHARNC:        case OP_CHARI:
1900        case OP_EXACT:        case OP_EXACT:
1901          case OP_EXACTI:
1902        case OP_UPTO:        case OP_UPTO:
1903          case OP_UPTOI:
1904        case OP_MINUPTO:        case OP_MINUPTO:
1905          case OP_MINUPTOI:
1906        case OP_POSUPTO:        case OP_POSUPTO:
1907          case OP_POSUPTOI:
1908        case OP_STAR:        case OP_STAR:
1909          case OP_STARI:
1910        case OP_MINSTAR:        case OP_MINSTAR:
1911          case OP_MINSTARI:
1912        case OP_POSSTAR:        case OP_POSSTAR:
1913          case OP_POSSTARI:
1914        case OP_PLUS:        case OP_PLUS:
1915          case OP_PLUSI:
1916        case OP_MINPLUS:        case OP_MINPLUS:
1917          case OP_MINPLUSI:
1918        case OP_POSPLUS:        case OP_POSPLUS:
1919          case OP_POSPLUSI:
1920        case OP_QUERY:        case OP_QUERY:
1921          case OP_QUERYI:
1922        case OP_MINQUERY:        case OP_MINQUERY:
1923          case OP_MINQUERYI:
1924        case OP_POSQUERY:        case OP_POSQUERY:
1925          case OP_POSQUERYI:
1926        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1927        break;        break;
1928        }        }
# Line 1919  could_be_empty_branch(const uschar *code Line 1961  could_be_empty_branch(const uschar *code
1961    compile_data *cd)    compile_data *cd)
1962  {  {
1963  register int c;  register int c;
1964  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
1965       code < endcode;       code < endcode;
1966       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
1967    {    {
1968    const uschar *ccode;    const uschar *ccode;
1969    
# Line 1937  for (code = first_significant_code(code Line 1979  for (code = first_significant_code(code
1979      continue;      continue;
1980      }      }
1981    
   /* 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;  
     }  
   
1982    /* For a recursion/subroutine call, if its end has been reached, which    /* For a recursion/subroutine call, if its end has been reached, which
1983    implies a subroutine call, we can scan it. */    implies a backward reference subroutine call, we can scan it. If it's a
1984      forward reference subroutine call, we can't. To detect forward reference
1985      we have to scan up the list that is kept in the workspace. This function is
1986      called only when doing the real compile, not during the pre-compile that
1987      measures the size of the compiled pattern. */
1988    
1989    if (c == OP_RECURSE)    if (c == OP_RECURSE)
1990      {      {
1991      BOOL empty_branch = FALSE;      const uschar *scode;
1992      const uschar *scode = cd->start_code + GET(code, 1);      BOOL empty_branch;
1993    
1994        /* Test for forward reference */
1995    
1996        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
1997          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
1998    
1999        /* Not a forward reference, test for completed backward reference */
2000    
2001        empty_branch = FALSE;
2002        scode = cd->start_code + GET(code, 1);
2003      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2004    
2005        /* Completed backwards reference */
2006    
2007      do      do
2008        {        {
2009        if (could_be_empty_branch(scode, endcode, utf8, cd))        if (could_be_empty_branch(scode, endcode, utf8, cd))
# Line 1965  for (code = first_significant_code(code Line 2014  for (code = first_significant_code(code
2014        scode += GET(scode, 1);        scode += GET(scode, 1);
2015        }        }
2016      while (*scode == OP_ALT);      while (*scode == OP_ALT);
2017    
2018      if (!empty_branch) return FALSE;  /* All branches are non-empty */      if (!empty_branch) return FALSE;  /* All branches are non-empty */
2019      continue;      continue;
2020      }      }
2021    
2022      /* Groups with zero repeats can of course be empty; skip them. */
2023    
2024      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2025          c == OP_BRAPOSZERO)
2026        {
2027        code += _pcre_OP_lengths[c];
2028        do code += GET(code, 1); while (*code == OP_ALT);
2029        c = *code;
2030        continue;
2031        }
2032    
2033      /* A nested group that is already marked as "could be empty" can just be
2034      skipped. */
2035    
2036      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2037          c == OP_SCBRA || c == OP_SCBRAPOS)
2038        {
2039        do code += GET(code, 1); while (*code == OP_ALT);
2040        c = *code;
2041        continue;
2042        }
2043    
2044    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2045    
2046    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2047          c == OP_CBRA || c == OP_CBRAPOS ||
2048          c == OP_ONCE || c == OP_COND)
2049      {      {
2050      BOOL empty_branch;      BOOL empty_branch;
2051      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 2058  for (code = first_significant_code(code Line 2132  for (code = first_significant_code(code
2132      case OP_ALLANY:      case OP_ALLANY:
2133      case OP_ANYBYTE:      case OP_ANYBYTE:
2134      case OP_CHAR:      case OP_CHAR:
2135      case OP_CHARNC:      case OP_CHARI:
2136      case OP_NOT:      case OP_NOT:
2137        case OP_NOTI:
2138      case OP_PLUS:      case OP_PLUS:
2139      case OP_MINPLUS:      case OP_MINPLUS:
2140      case OP_POSPLUS:      case OP_POSPLUS:
# Line 2099  for (code = first_significant_code(code Line 2174  for (code = first_significant_code(code
2174      case OP_KET:      case OP_KET:
2175      case OP_KETRMAX:      case OP_KETRMAX:
2176      case OP_KETRMIN:      case OP_KETRMIN:
2177        case OP_KETRPOS:
2178      case OP_ALT:      case OP_ALT:
2179      return TRUE;      return TRUE;
2180    
# Line 2107  for (code = first_significant_code(code Line 2183  for (code = first_significant_code(code
2183    
2184  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2185      case OP_STAR:      case OP_STAR:
2186        case OP_STARI:
2187      case OP_MINSTAR:      case OP_MINSTAR:
2188        case OP_MINSTARI:
2189      case OP_POSSTAR:      case OP_POSSTAR:
2190        case OP_POSSTARI:
2191      case OP_QUERY:      case OP_QUERY:
2192        case OP_QUERYI:
2193      case OP_MINQUERY:      case OP_MINQUERY:
2194        case OP_MINQUERYI:
2195      case OP_POSQUERY:      case OP_POSQUERY:
2196        case OP_POSQUERYI:
2197      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2198      break;      break;
2199    
2200      case OP_UPTO:      case OP_UPTO:
2201        case OP_UPTOI:
2202      case OP_MINUPTO:      case OP_MINUPTO:
2203        case OP_MINUPTOI:
2204      case OP_POSUPTO:      case OP_POSUPTO:
2205        case OP_POSUPTOI:
2206      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2207      break;      break;
2208  #endif  #endif
# Line 2155  return TRUE; Line 2240  return TRUE;
2240  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
2241  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,
2242  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.
2243    This function is called only during the real compile, not during the
2244    pre-compile.
2245    
2246  Arguments:  Arguments:
2247    code        points to start of the recursion    code        points to start of the recursion
# Line 2205  where Perl recognizes it as the POSIX cl Line 2292  where Perl recognizes it as the POSIX cl
2292  "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,
2293  I think.  I think.
2294    
2295    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2296    It seems that the appearance of a nested POSIX class supersedes an apparent
2297    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2298    a digit. Also, unescaped square brackets may also appear as part of class
2299    names. For example, [:a[:abc]b:] gives unknown class "[:abc]b:]"in Perl.
2300    
2301  Arguments:  Arguments:
2302    ptr      pointer to the initial [    ptr      pointer to the initial [
2303    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 2219  int terminator;          /* Don't combin Line 2312  int terminator;          /* Don't combin
2312  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2313  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2314    {    {
2315    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2316        ptr++;
2317      else
2318      {      {
     if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
2319      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2320        {        {
2321        *endptr = ptr;        *endptr = ptr;
2322        return TRUE;        return TRUE;
2323        }        }
2324        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2325             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2326              ptr[1] == CHAR_EQUALS_SIGN) &&
2327            check_posix_syntax(ptr, endptr))
2328          return FALSE;
2329      }      }
2330    }    }
2331  return FALSE;  return FALSE;
# Line 2532  if ((options & PCRE_EXTENDED) != 0) Line 2631  if ((options & PCRE_EXTENDED) != 0)
2631      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2632      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2633        {        {
2634        ptr++;        ptr++;
2635        while (*ptr != 0)        while (*ptr != 0)
2636          {          {
2637          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2638          ptr++;          ptr++;
2639  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2640          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2641  #endif  #endif
2642          }          }
# Line 2576  if ((options & PCRE_EXTENDED) != 0) Line 2675  if ((options & PCRE_EXTENDED) != 0)
2675      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2676      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2677        {        {
2678        ptr++;        ptr++;
2679        while (*ptr != 0)        while (*ptr != 0)
2680          {          {
2681          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2682          ptr++;          ptr++;
2683  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2684          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2685  #endif  #endif
2686          }          }
# Line 2609  if (next >= 0) switch(op_code) Line 2708  if (next >= 0) switch(op_code)
2708  #endif  #endif
2709    return c != next;    return c != next;
2710    
2711    /* 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
2712    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
2713    high-valued characters. */    high-valued characters. */
2714    
2715    case OP_CHARNC:    case OP_CHARI:
2716  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2717    GETCHARTEST(c, previous);    GETCHARTEST(c, previous);
2718  #else  #else
# Line 2636  if (next >= 0) switch(op_code) Line 2735  if (next >= 0) switch(op_code)
2735  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2736    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2737    
2738    /* 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
2739      opcodes are not used for multi-byte characters, because they are coded using
2740      an XCLASS instead. */
2741    
2742    case OP_NOT:    case OP_NOT:
2743      return (c = *previous) == next;
2744    
2745      case OP_NOTI:
2746    if ((c = *previous) == next) return TRUE;    if ((c = *previous) == next) return TRUE;
   if ((options & PCRE_CASELESS) == 0) return FALSE;  
2747  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2748    if (utf8)    if (utf8)
2749      {      {
# Line 2745  replaced by OP_PROP codes when PCRE_UCP Line 2848  replaced by OP_PROP codes when PCRE_UCP
2848  switch(op_code)  switch(op_code)
2849    {    {
2850    case OP_CHAR:    case OP_CHAR:
2851    case OP_CHARNC:    case OP_CHARI:
2852  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2853    GETCHARTEST(c, previous);    GETCHARTEST(c, previous);
2854  #else  #else
# Line 2952  int greedy_default, greedy_non_default; Line 3055  int greedy_default, greedy_non_default;
3055  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3056  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3057  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3058  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3059  int after_manual_callout = 0;  int after_manual_callout = 0;
3060  int length_prevgroup = 0;  int length_prevgroup = 0;
3061  register int c;  register int c;
# Line 2970  uschar *previous_callout = NULL; Line 3073  uschar *previous_callout = NULL;
3073  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3074  uschar classbits[32];  uschar classbits[32];
3075    
3076    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3077    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3078    dynamically as we process the pattern. */
3079    
3080  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3081  BOOL class_utf8;  BOOL class_utf8;
3082  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
# Line 3150  for (;; ptr++) Line 3257  for (;; ptr++)
3257      previous_callout = NULL;      previous_callout = NULL;
3258      }      }
3259    
3260    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3261    
3262    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3263      {      {
3264      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3265      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
3266        {        {
3267        ptr++;        ptr++;
3268        while (*ptr != 0)        while (*ptr != 0)
3269          {          {
3270          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3271          ptr++;          ptr++;
3272  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3273          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3274  #endif  #endif
3275          }          }
# Line 3209  for (;; ptr++) Line 3316  for (;; ptr++)
3316      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3317    
3318      case CHAR_CIRCUMFLEX_ACCENT:      case CHAR_CIRCUMFLEX_ACCENT:
3319        previous = NULL;
3320      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3321        {        {
3322        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3323          *code++ = OP_CIRCM;
3324        }        }
3325      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3326      break;      break;
3327    
3328      case CHAR_DOLLAR_SIGN:      case CHAR_DOLLAR_SIGN:
3329      previous = NULL;      previous = NULL;
3330      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3331      break;      break;
3332    
3333      /* 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 3540  for (;; ptr++) Line 3648  for (;; ptr++)
3648              continue;              continue;
3649    
3650              /* Perl 5.004 onwards omits VT from \s, but we must preserve it              /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3651              if it was previously set by something earlier in the character              if it was previously set by something earlier in the character
3652              class. */              class. */
3653    
3654              case ESC_s:              case ESC_s:
3655              classbits[0] |= cbits[cbit_space];              classbits[0] |= cbits[cbit_space];
3656              classbits[1] |= cbits[cbit_space+1] & ~0x08;              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3657              for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3658              continue;              continue;
3659    
# Line 3965  for (;; ptr++) Line 4073  for (;; ptr++)
4073    
4074      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
4075      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4076      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4077      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4078    
4079      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
4080      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.
4081      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
4082      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
4083      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
4084      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4085    
4086  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4087      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
# Line 3984  for (;; ptr++) Line 4092  for (;; ptr++)
4092        {        {
4093        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4094    
4095        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4096    
4097        if (negate_class)        if (negate_class)
4098          {          {
4099          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4100          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4101          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4102          *code++ = class_lastchar;          *code++ = class_lastchar;
4103          break;          break;
4104          }          }
# Line 4118  for (;; ptr++) Line 4226  for (;; ptr++)
4226      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4227      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4228    
4229      /* 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
4230      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4231    
4232      tempcode = previous;      tempcode = previous;
4233    
# Line 4141  for (;; ptr++) Line 4249  for (;; ptr++)
4249        ptr++;        ptr++;
4250        }        }
4251      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4252    
4253        /* If previous was a recursion call, wrap it in atomic brackets so that
4254        previous becomes the atomic group. All recursions were so wrapped in the
4255        past, but it no longer happens for non-repeated recursions. In fact, the
4256        repeated ones could be re-implemented independently so as not to need this,
4257        but for the moment we rely on the code for repeating groups. */
4258    
4259        if (*previous == OP_RECURSE)
4260          {
4261          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4262          *previous = OP_ONCE;
4263          PUT(previous, 1, 2 + 2*LINK_SIZE);
4264          previous[2 + 2*LINK_SIZE] = OP_KET;
4265          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4266          code += 2 + 2 * LINK_SIZE;
4267          length_prevgroup = 3 + 3*LINK_SIZE;
4268    
4269          /* When actually compiling, we need to check whether this was a forward
4270          reference, and if so, adjust the offset. */
4271    
4272          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4273            {
4274            int offset = GET(cd->hwm, -LINK_SIZE);
4275            if (offset == previous + 1 - cd->start_code)
4276              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4277            }
4278          }
4279    
4280        /* Now handle repetition for the different types of item. */
4281    
4282      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4283      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
# Line 4148  for (;; ptr++) Line 4285  for (;; ptr++)
4285      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
4286      instead.  */      instead.  */
4287    
4288      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4289        {        {
4290          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4291    
4292        /* 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
4293        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
4294        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 4194  for (;; ptr++) Line 4333  for (;; ptr++)
4333      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4334      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-
4335      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4336      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
4337      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4338    
4339      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4340        {        {
4341        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4342        c = previous[1];        c = previous[1];
4343        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4344            repeat_max < 0 &&            repeat_max < 0 &&
# Line 4396  for (;; ptr++) Line 4535  for (;; ptr++)
4535  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4536               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4537  #endif  #endif
4538               *previous == OP_REF)               *previous == OP_REF ||
4539                 *previous == OP_REFI)
4540        {        {
4541        if (repeat_max == 0)        if (repeat_max == 0)
4542          {          {
# Line 4430  for (;; ptr++) Line 4570  for (;; ptr++)
4570        }        }
4571    
4572      /* 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
4573      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
4574        opcodes such as BRA and CBRA, as this is the place where they get converted
4575        into the more special varieties such as BRAPOS and SBRA. A test for >=
4576        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4577        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4578        repetition of assertions, but now it does, for Perl compatibility. */
4579    
4580      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4581        {        {
4582        register int i;        register int i;
       int ketoffset = 0;  
4583        int len = (int)(code - previous);        int len = (int)(code - previous);
4584        uschar *bralink = NULL;        uschar *bralink = NULL;
4585          uschar *brazeroptr = NULL;
4586        /* Repeating a DEFINE group is pointless */  
4587          /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4588          we just ignore the repeat. */
4589    
4590        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4591          {          goto END_REPEAT;
         *errorcodeptr = ERR55;  
         goto FAILED;  
         }  
4592    
4593        /* If the maximum repeat count is unlimited, find the end of the bracket        /* There is no sense in actually repeating assertions. The only potential
4594        by scanning through from the start, and compute the offset back to it        use of repetition is in cases when the assertion is optional. Therefore,
4595        from the current code pointer. There may be an OP_OPT setting following        if the minimum is greater than zero, just ignore the repeat. If the
4596        the final KET, so we can't find the end just by going back from the code        maximum is not not zero or one, set it to 1. */
4597        pointer. */  
4598          if (*previous < OP_ONCE)    /* Assertion */
4599        if (repeat_max == -1)          {
4600          {          if (repeat_min > 0) goto END_REPEAT;
4601          register uschar *ket = previous;          if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
4602          do ket += GET(ket, 1); while (*ket != OP_KET);          }
         ketoffset = (int)(code - ket);  
         }  
4603    
4604        /* 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
4605        OP_BRAZERO in front of it, and because the group appears once in the        OP_BRAZERO in front of it, and because the group appears once in the
# Line 4479  for (;; ptr++) Line 4619  for (;; ptr++)
4619          **   goto END_REPEAT;          **   goto END_REPEAT;
4620          **   }          **   }
4621    
4622          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
4623          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
4624          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
4625          groups are referenced, we cannot do this selectively.          don't have a list of which groups are referenced, we cannot do this
4626            selectively.
4627    
4628          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
4629          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 4502  for (;; ptr++) Line 4643  for (;; ptr++)
4643              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
4644              goto END_REPEAT;              goto END_REPEAT;
4645              }              }
4646              brazeroptr = previous;    /* Save for possessive optimizing */
4647            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4648            }            }
4649    
# Line 4666  for (;; ptr++) Line 4808  for (;; ptr++)
4808            }            }
4809          }          }
4810    
4811        /* 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
4812        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
4813        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
4814        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
4815          deal with possessive ONCEs specially.
4816    
4817          Otherwise, if the quantifier was possessive, we convert the BRA code to
4818          the POS form, and the KET code to KETRPOS. (It turns out to be convenient
4819          at runtime to detect this kind of subpattern at both the start and at the
4820          end.) The use of special opcodes makes it possible to reduce greatly the
4821          stack usage in pcre_exec(). If the group is preceded by OP_BRAZERO,
4822          convert this to OP_BRAPOSZERO. Then cancel the possessive flag so that
4823          the default action below, of wrapping everything inside atomic brackets,
4824          does not happen.
4825    
4826        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
4827        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
4828        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
4829        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
4830        atomic groups at runtime, but in a different way.] */        runtime, but in a different way.] */
4831    
4832        else        else
4833          {          {
4834          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
4835          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
4836          *ketcode = OP_KETRMAX + repeat_type;  
4837          if (lengthptr == NULL && *bracode != OP_ONCE)          if (*bracode == OP_ONCE && possessive_quantifier) *bracode = OP_BRA;
4838            if (*bracode == OP_ONCE)
4839              *ketcode = OP_KETRMAX + repeat_type;
4840            else
4841            {            {
4842            uschar *scode = bracode;            if (possessive_quantifier)
4843            do              {
4844                *bracode += 1;                   /* Switch to xxxPOS opcodes */
4845                *ketcode = OP_KETRPOS;
4846                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
4847                possessive_quantifier = FALSE;
4848                }
4849              else *ketcode = OP_KETRMAX + repeat_type;
4850    
4851              if (lengthptr == NULL)
4852              {              {
4853              if (could_be_empty_branch(scode, ketcode, utf8, cd))              uschar *scode = bracode;
4854                do
4855                {                {
4856                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
4857                break;                  {
4858                    *bracode += OP_SBRA - OP_BRA;
4859                    break;
4860                    }
4861                  scode += GET(scode, 1);
4862                }                }
4863              scode += GET(scode, 1);              while (*scode == OP_ALT);
4864              }              }
           while (*scode == OP_ALT);  
4865            }            }
4866          }          }
4867        }        }
# Line 4715  for (;; ptr++) Line 4882  for (;; ptr++)
4882        }        }
4883    
4884      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
4885      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
4886      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
4887      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
4888      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
4889      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
4890      tempcode, not at previous, which might be the first part of a string whose  
4891      (former) last char we repeated.      Possessively repeated subpatterns have already been handled in the code
4892        just above, so possessive_quantifier is always FALSE for them at this
4893        stage.
4894    
4895        Note that the repeated item starts at tempcode, not at previous, which
4896        might be the first part of a string whose (former) last char we repeated.
4897    
4898      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
4899      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 4752  for (;; ptr++) Line 4924  for (;; ptr++)
4924          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
4925          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4926    
4927          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
4928          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
4929          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
4930          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
4931    
4932          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4933          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4934          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4935          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4936    
4937            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
4938            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
4939            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
4940            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
4941    
4942            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
4943            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
4944            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
4945            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
4946    
4947          /* Because we are moving code along, we must ensure that any          /* Because we are moving code along, we must ensure that any
4948          pending recursive references are updated. */          pending recursive references are updated. */
4949    
# Line 4836  for (;; ptr++) Line 5018  for (;; ptr++)
5018          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
5019              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
5020            {            {
5021            /* Check for open captures before ACCEPT */            /* Check for open captures before ACCEPT and convert it to
5022              ASSERT_ACCEPT if in an assertion. */
5023    
5024            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
5025              {              {
5026              open_capitem *oc;              open_capitem *oc;
5027                if (arglen != 0)
5028                  {
5029                  *errorcodeptr = ERR59;
5030                  goto FAILED;
5031                  }
5032              cd->had_accept = TRUE;              cd->had_accept = TRUE;
5033              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5034                {                {
5035                *code++ = OP_CLOSE;                *code++ = OP_CLOSE;
5036                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
5037                }                }
5038                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5039              }              }
5040    
5041            /* Handle the cases with/without an argument */            /* Handle other cases with/without an argument */
5042    
5043            if (arglen == 0)            else if (arglen == 0)
5044              {              {
5045              if (verbs[i].op < 0)   /* Argument is mandatory */              if (verbs[i].op < 0)   /* Argument is mandatory */
5046                {                {
# Line 4862  for (;; ptr++) Line 5051  for (;; ptr++)
5051              if (*code++ == OP_THEN)              if (*code++ == OP_THEN)
5052                {                {
5053                PUT(code, 0, code - bcptr->current_branch - 1);                PUT(code, 0, code - bcptr->current_branch - 1);
5054                code += LINK_SIZE;                code += LINK_SIZE;
5055                }                }
5056              }              }
5057    
5058            else            else
# Line 4877  for (;; ptr++) Line 5066  for (;; ptr++)
5066              if (*code++ == OP_THEN_ARG)              if (*code++ == OP_THEN_ARG)
5067                {                {
5068                PUT(code, 0, code - bcptr->current_branch - 1);                PUT(code, 0, code - bcptr->current_branch - 1);
5069                code += LINK_SIZE;                code += LINK_SIZE;
5070                }                }
5071              *code++ = arglen;              *code++ = arglen;
5072              memcpy(code, arg, arglen);              memcpy(code, arg, arglen);
5073              code += arglen;              code += arglen;
# Line 5140  for (;; ptr++) Line 5329  for (;; ptr++)
5329          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5330          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5331          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5332            cd->assert_depth += 1;
5333          ptr++;          ptr++;
5334          break;          break;
5335    
# Line 5154  for (;; ptr++) Line 5344  for (;; ptr++)
5344            continue;            continue;
5345            }            }
5346          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5347            cd->assert_depth += 1;
5348          break;          break;
5349    
5350    
# Line 5163  for (;; ptr++) Line 5354  for (;; ptr++)
5354            {            {
5355            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5356            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5357              cd->assert_depth += 1;
5358            ptr += 2;            ptr += 2;
5359            break;            break;
5360    
5361            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5362            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5363              cd->assert_depth += 1;
5364            ptr += 2;            ptr += 2;
5365            break;            break;
5366    
# Line 5373  for (;; ptr++) Line 5566  for (;; ptr++)
5566          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5567          namelen = (int)(ptr - name);          namelen = (int)(ptr - name);
5568    
5569          /* 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
5570          reference number. */          a dummy reference number, because it was not used in the first pass.
5571            However, with the change of recursive back references to be atomic,
5572            we have to look for the number so that this state can be identified, as
5573            otherwise the incorrect length is computed. If it's not a backwards
5574            reference, the dummy number will do. */
5575    
5576          if (lengthptr != NULL)          if (lengthptr != NULL)
5577            {            {
5578              const uschar *temp;
5579    
5580            if (namelen == 0)            if (namelen == 0)
5581              {              {
5582              *errorcodeptr = ERR62;              *errorcodeptr = ERR62;
# Line 5393  for (;; ptr++) Line 5592  for (;; ptr++)
5592              *errorcodeptr = ERR48;              *errorcodeptr = ERR48;
5593              goto FAILED;              goto FAILED;
5594              }              }
5595            recno = 0;  
5596              /* The name table does not exist in the first pass, so we cannot
5597              do a simple search as in the code below. Instead, we have to scan the
5598              pattern to find the number. It is important that we scan it only as
5599              far as we have got because the syntax of named subpatterns has not
5600              been checked for the rest of the pattern, and find_parens() assumes
5601              correct syntax. In any case, it's a waste of resources to scan
5602              further. We stop the scan at the current point by temporarily
5603              adjusting the value of cd->endpattern. */
5604    
5605              temp = cd->end_pattern;
5606              cd->end_pattern = ptr;
5607              recno = find_parens(cd, name, namelen,
5608                (options & PCRE_EXTENDED) != 0, utf8);
5609              cd->end_pattern = temp;
5610              if (recno < 0) recno = 0;    /* Forward ref; set dummy number */
5611            }            }
5612    
5613          /* 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 5537  for (;; ptr++) Line 5751  for (;; ptr++)
5751    
5752                /* 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
5753                offset below, what it actually inserted is the reference number                offset below, what it actually inserted is the reference number
5754                of the group. */                of the group. Then remember the forward reference. */
5755    
5756                called = cd->start_code + recno;                called = cd->start_code + recno;
5757                PUTINC(cd->hwm, 0, (int)(code + 2 + LINK_SIZE - cd->start_code));                PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
5758                }                }
5759    
5760              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
# Line 5555  for (;; ptr++) Line 5769  for (;; ptr++)
5769                }                }
5770              }              }
5771    
5772            /* Insert the recursion/subroutine item, automatically wrapped inside            /* Insert the recursion/subroutine item. */
5773            "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;  
   
5774            *code = OP_RECURSE;            *code = OP_RECURSE;
5775            PUT(code, 1, (int)(called - cd->start_code));            PUT(code, 1, (int)(called - cd->start_code));
5776            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;  
5777            }            }
5778    
5779          /* Can't determine a first byte now */          /* Can't determine a first byte now */
# Line 5632  for (;; ptr++) Line 5834  for (;; ptr++)
5834          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
5835          both phases.          both phases.
5836    
5837          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
5838          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. */  
5839    
5840          if (*ptr == CHAR_RIGHT_PARENTHESIS)          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5841            {            {
# Line 5645  for (;; ptr++) Line 5846  for (;; ptr++)
5846              }              }
5847            else            else
5848              {              {
             if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))  
               {  
               *code++ = OP_OPT;  
               *code++ = newoptions & PCRE_IMS;  
               }  
5849              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5850              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5851              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5852              }              }
5853    
5854            /* Change options at this level, and pass them back for use            /* Change options at this level, and pass them back for use
5855            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). */  
5856    
5857            *optionsptr = options = newoptions;            *optionsptr = options = newoptions;
5858            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
# Line 5694  for (;; ptr++) Line 5888  for (;; ptr++)
5888        skipbytes = 2;        skipbytes = 2;
5889        }        }
5890    
5891      /* Process nested bracketed regex. Assertions may not be repeated, but      /* Process nested bracketed regex. Assertions used not to be repeatable,
5892      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
5893      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
5894      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. */  
5895    
5896      previous = (bravalue >= OP_ONCE)? code : NULL;      previous = code;                   /* For handling repetition */
5897      *code = bravalue;      *code = bravalue;
5898      tempcode = code;      tempcode = code;
5899      tempreqvary = cd->req_varyopt;     /* Save value before bracket */      tempreqvary = cd->req_varyopt;     /* Save value before bracket */
# Line 5708  for (;; ptr++) Line 5901  for (;; ptr++)
5901    
5902      if (!compile_regex(      if (!compile_regex(
5903           newoptions,                   /* The complete new option state */           newoptions,                   /* The complete new option state */
          options & PCRE_IMS,           /* The previous ims option state */  
5904           &tempcode,                    /* Where to put code (updated) */           &tempcode,                    /* Where to put code (updated) */
5905           &ptr,                         /* Input pointer (updated) */           &ptr,                         /* Input pointer (updated) */
5906           errorcodeptr,                 /* Where to put an error message */           errorcodeptr,                 /* Where to put an error message */
# Line 5724  for (;; ptr++) Line 5916  for (;; ptr++)
5916             &length_prevgroup           /* Pre-compile phase */             &length_prevgroup           /* Pre-compile phase */
5917           ))           ))
5918        goto FAILED;        goto FAILED;
5919    
5920        if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
5921          cd->assert_depth -= 1;
5922    
5923      /* 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
5924      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
# Line 5795  for (;; ptr++) Line 5990  for (;; ptr++)
5990          goto FAILED;          goto FAILED;
5991          }          }
5992        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
5993        *code++ = OP_BRA;        code++;   /* This already contains bravalue */
5994        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5995        *code++ = OP_KET;        *code++ = OP_KET;
5996        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
# Line 5963  for (;; ptr++) Line 6158  for (;; ptr++)
6158          }          }
6159    
6160        /* \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).
6161        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax).  */
6162    
6163        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||        if (-c == ESC_k)
           ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))  
6164          {          {
6165            if ((ptr[1] != CHAR_LESS_THAN_SIGN &&
6166              ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))
6167              {
6168              *errorcodeptr = ERR69;
6169              break;
6170              }
6171          is_recurse = FALSE;          is_recurse = FALSE;
6172          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
6173            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
6174            CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;            CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
6175          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
6176          }          }
6177    
6178        /* Back references are handled specially; must disable firstbyte if        /* Back references are handled specially; must disable firstbyte if
6179        not set to cope with cases like (?=(\w+))\1: which would otherwise set        not set to cope with cases like (?=(\w+))\1: which would otherwise set
# Line 5987  for (;; ptr++) Line 6187  for (;; ptr++)
6187          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
6188          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
6189          previous = code;          previous = code;
6190          *code++ = OP_REF;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_REFI : OP_REF;
6191          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
6192          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
6193          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
# Line 6095  for (;; ptr++) Line 6295  for (;; ptr++)
6295    
6296      ONE_CHAR:      ONE_CHAR:
6297      previous = code;      previous = code;
6298      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARI : OP_CHAR;
6299      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
6300    
6301      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
# Line 6159  return FALSE; Line 6359  return FALSE;
6359  /* 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
6360  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
6361  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.  
   
6362  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
6363  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
6364  value of lengthptr distinguishes the two phases.  value of lengthptr distinguishes the two phases.
6365    
6366  Arguments:  Arguments:
6367    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  
6368    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
6369    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
6370    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
# Line 6188  Returns:         TRUE on success Line 6382  Returns:         TRUE on success
6382  */  */
6383    
6384  static BOOL  static BOOL
6385  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,
6386    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
6387    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,
6388    int *lengthptr)    int *lengthptr)
# Line 6205  int branchfirstbyte, branchreqbyte; Line 6399  int branchfirstbyte, branchreqbyte;
6399  int length;  int length;
6400  int orig_bracount;  int orig_bracount;
6401  int max_bracount;  int max_bracount;
 int old_external_options = cd->external_options;  
6402  branch_chain bc;  branch_chain bc;
6403    
6404  bc.outer = bcptr;  bc.outer = bcptr;
# Line 6229  pre-compile phase to find out whether an Line 6422  pre-compile phase to find out whether an
6422    
6423  /* 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
6424  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
6425  detect groups that contain recursive back references to themselves. */  detect groups that contain recursive back references to themselves. Note that
6426    only OP_CBRA need be tested here; changing this opcode to one of its variants,
6427    e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
6428    
6429  if (*code == OP_CBRA)  if (*code == OP_CBRA)
6430    {    {
# Line 6255  for (;;) Line 6450  for (;;)
6450    
6451    if (reset_bracount) cd->bracount = orig_bracount;    if (reset_bracount) cd->bracount = orig_bracount;
6452    
   /* 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;  
     }  
   
6453    /* Set up dummy OP_REVERSE if lookbehind assertion */    /* Set up dummy OP_REVERSE if lookbehind assertion */
6454    
6455    if (lookbehind)    if (lookbehind)
# Line 6284  for (;;) Line 6470  for (;;)
6470      return FALSE;      return FALSE;
6471      }      }
6472    
   /* 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;  
   
6473    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6474    has fewer than the rest. */    has fewer than the rest. */
6475    
# Line 6353  for (;;) Line 6530  for (;;)
6530        {        {
6531        int fixed_length;        int fixed_length;
6532        *code = OP_END;        *code = OP_END;
6533        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);        fixed_length = find_fixedlength(last_branch,  (options & PCRE_UTF8) != 0,
6534            FALSE, cd);
6535        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6536        if (fixed_length == -3)        if (fixed_length == -3)
6537          {          {
# Line 6374  for (;;) Line 6552  for (;;)
6552    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
6553    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
6554    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
6555    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. */  
6556    
6557    if (*ptr != CHAR_VERTICAL_LINE)    if (*ptr != CHAR_VERTICAL_LINE)
6558      {      {
# Line 6420  for (;;) Line 6596  for (;;)
6596        cd->open_caps = cd->open_caps->next;        cd->open_caps = cd->open_caps->next;
6597        }        }
6598    
     /* Reset options if needed. */  
   
     if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)  
       {  
       *code++ = OP_OPT;  
       *code++ = oldims;  
       length += 2;  
       }  
   
6599      /* Retain the highest bracket number, in case resetting was used. */      /* Retain the highest bracket number, in case resetting was used. */
6600    
6601      cd->bracount = max_bracount;      cd->bracount = max_bracount;
# Line 6488  for (;;) Line 6655  for (;;)
6655  /* 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
6656  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
6657  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
6658  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
6659  counts, since OP_CIRC can match in the middle.  be found, because ^ generates OP_CIRCM in that mode.
6660    
6661  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.
6662  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 6510  of the more common cases more precisely. Line 6677  of the more common cases more precisely.
6677    
6678  Arguments:  Arguments:
6679    code           points to start of expression (the bracket)    code           points to start of expression (the bracket)
   options        points to the options setting  
6680    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
6681                    handles up to substring 31; after that we just have to take                    handles up to substring 31; after that we just have to take
6682                    the less precise approach                    the less precise approach
# Line 6520  Returns:     TRUE or FALSE Line 6686  Returns:     TRUE or FALSE
6686  */  */
6687    
6688  static BOOL  static BOOL
6689  is_anchored(register const uschar *code, int *options, unsigned int bracket_map,  is_anchored(register const uschar *code, unsigned int bracket_map,
6690    unsigned int backref_map)    unsigned int backref_map)
6691  {  {
6692  do {  do {
6693     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6694       options, PCRE_MULTILINE, FALSE);       FALSE);
6695     register int op = *scode;     register int op = *scode;
6696    
6697     /* Non-capturing brackets */     /* Non-capturing brackets */
6698    
6699     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6700           op == OP_SBRA || op == OP_SBRAPOS)
6701       {       {
6702       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6703       }       }
6704    
6705     /* Capturing brackets */     /* Capturing brackets */
6706    
6707     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6708                op == OP_SCBRA || op == OP_SCBRAPOS)
6709       {       {
6710       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6711       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
6712       if (!is_anchored(scode, options, new_map, backref_map)) return FALSE;       if (!is_anchored(scode, new_map, backref_map)) return FALSE;
6713       }       }
6714    
6715     /* Other brackets */     /* Other brackets */
6716    
6717     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)
6718       {       {
6719       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6720       }       }
6721    
6722     /* .* 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 6563  do { Line 6731  do {
6731    
6732     /* Check for explicit anchoring */     /* Check for explicit anchoring */
6733    
6734     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;  
6735     code += GET(code, 1);     code += GET(code, 1);
6736     }     }
6737  while (*code == OP_ALT);   /* Loop for each alternative */  while (*code == OP_ALT);   /* Loop for each alternative */
# Line 6601  is_startline(const uschar *code, unsigne Line 6767  is_startline(const uschar *code, unsigne
6767  {  {
6768  do {  do {
6769     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6770       NULL, 0, FALSE);       FALSE);
6771     register int op = *scode;     register int op = *scode;
6772    
6773     /* 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 6628  do { Line 6794  do {
6794         scode += 1 + LINK_SIZE;         scode += 1 + LINK_SIZE;
6795         break;         break;
6796         }         }
6797       scode = first_significant_code(scode, NULL, 0, FALSE);       scode = first_significant_code(scode, FALSE);
6798       op = *scode;       op = *scode;
6799       }       }
6800    
6801     /* Non-capturing brackets */     /* Non-capturing brackets */
6802    
6803     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6804           op == OP_SBRA || op == OP_SBRAPOS)
6805       {       {
6806       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6807       }       }
6808    
6809     /* Capturing brackets */     /* Capturing brackets */
6810    
6811     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6812                op == OP_SCBRA || op == OP_SCBRAPOS)
6813       {       {
6814       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6815       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
# Line 6665  do { Line 6833  do {
6833    
6834     /* Check for explicit circumflex */     /* Check for explicit circumflex */
6835    
6836     else if (op != OP_CIRC) return FALSE;     else if (op != OP_CIRC && op != OP_CIRCM) return FALSE;
6837    
6838     /* Move on to the next alternative */     /* Move on to the next alternative */
6839    
# Line 6691  we return that char, otherwise -1. Line 6859  we return that char, otherwise -1.
6859    
6860  Arguments:  Arguments:
6861    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)  
6862    inassert   TRUE if in an assertion    inassert   TRUE if in an assertion
6863    
6864  Returns:     -1 or the fixed first char  Returns:     -1 or the fixed first char
6865  */  */
6866    
6867  static int  static int
6868  find_firstassertedchar(const uschar *code, int *options, BOOL inassert)  find_firstassertedchar(const uschar *code, BOOL inassert)
6869  {  {
6870  register int c = -1;  register int c = -1;
6871  do {  do {
6872     int d;     int d;
6873     const uschar *scode =     int xl = (*code == OP_CBRA || *code == OP_SCBRA ||
6874       first_significant_code(code + 1+LINK_SIZE, options, PCRE_CASELESS, TRUE);               *code == OP_CBRAPOS || *code == OP_SCBRAPOS)? 2:0;
6875       const uschar *scode = first_significant_code(code + 1+LINK_SIZE + xl, TRUE);
6876     register int op = *scode;     register int op = *scode;
6877    
6878     switch(op)     switch(op)
# Line 6713  do { Line 6881  do {
6881       return -1;       return -1;
6882    
6883       case OP_BRA:       case OP_BRA:
6884         case OP_BRAPOS:
6885       case OP_CBRA:       case OP_CBRA:
6886         case OP_SCBRA:
6887         case OP_CBRAPOS:
6888         case OP_SCBRAPOS:
6889       case OP_ASSERT:       case OP_ASSERT:
6890       case OP_ONCE:       case OP_ONCE:
6891       case OP_COND:       case OP_COND:
6892       if ((d = find_firstassertedchar(scode, options, op == OP_ASSERT)) < 0)       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)
6893         return -1;         return -1;
6894       if (c < 0) c = d; else if (c != d) return -1;       if (c < 0) c = d; else if (c != d) return -1;
6895       break;       break;
6896    
6897       case OP_EXACT:       /* Fall through */       case OP_EXACT:
6898       scode += 2;       scode += 2;
6899         /* Fall through */
6900    
6901       case OP_CHAR:       case OP_CHAR:
      case OP_CHARNC:  
6902       case OP_PLUS:       case OP_PLUS:
6903       case OP_MINPLUS:       case OP_MINPLUS:
6904       case OP_POSPLUS:       case OP_POSPLUS:
6905       if (!inassert) return -1;       if (!inassert) return -1;
6906       if (c < 0)       if (c < 0) c = scode[1];
6907         {         else if (c != scode[1]) return -1;
6908         c = scode[1];       break;
6909         if ((*options & PCRE_CASELESS) != 0) c |= REQ_CASELESS;  
6910         }       case OP_EXACTI:
6911       else if (c != scode[1]) return -1;       scode += 2;
6912         /* Fall through */
6913    
6914         case OP_CHARI:
6915         case OP_PLUSI:
6916         case OP_MINPLUSI:
6917         case OP_POSPLUSI:
6918         if (!inassert) return -1;
6919         if (c < 0) c = scode[1] | REQ_CASELESS;
6920           else if (c != scode[1]) return -1;
6921       break;       break;
6922       }       }
6923    
# Line 6859  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7040  while (ptr[skipatstart] == CHAR_LEFT_PAR
7040      { skipatstart += 7; options |= PCRE_UTF8; continue; }      { skipatstart += 7; options |= PCRE_UTF8; continue; }
7041    else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)
7042      { skipatstart += 6; options |= PCRE_UCP; continue; }      { skipatstart += 6; options |= PCRE_UCP; continue; }
7043      else if (strncmp((char *)(ptr+skipatstart+2), STRING_NO_START_OPT_RIGHTPAR, 13) == 0)
7044        { skipatstart += 15; options |= PCRE_NO_START_OPTIMIZE; continue; }
7045    
7046    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
7047      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
# Line 6885  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7068  while (ptr[skipatstart] == CHAR_LEFT_PAR
7068    
7069  utf8 = (options & PCRE_UTF8) != 0;  utf8 = (options & PCRE_UTF8) != 0;
7070    
7071  /* 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
7072    return of an error code from _pcre_valid_utf8() is a new feature, introduced in
7073    release 8.13. It is passed back from pcre_[dfa_]exec(), but at the moment is
7074    not used here. */
7075    
7076  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
7077  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
7078       (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)       (errorcode = _pcre_valid_utf8((USPTR)pattern, -1, erroroffset)) != 0)
7079    {    {
7080    errorcode = ERR44;    errorcode = ERR44;
7081    goto PCRE_EARLY_ERROR_RETURN2;    goto PCRE_EARLY_ERROR_RETURN2;
# Line 6914  if ((options & PCRE_UCP) != 0) Line 7100  if ((options & PCRE_UCP) != 0)
7100    
7101  /* Check validity of \R options. */  /* Check validity of \R options. */
7102    
7103  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  if ((options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) ==
7104         (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
7105    {    {
7106    case 0:    errorcode = ERR56;
7107    case PCRE_BSR_ANYCRLF:    goto PCRE_EARLY_ERROR_RETURN;
   case PCRE_BSR_UNICODE:  
   break;  
   default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;  
7108    }    }
7109    
7110  /* 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 7005  outside can help speed up starting point Line 7189  outside can help speed up starting point
7189  ptr += skipatstart;  ptr += skipatstart;
7190  code = cworkspace;  code = cworkspace;
7191  *code = OP_BRA;  *code = OP_BRA;
7192  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
7193    &code, &ptr, &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd,    FALSE, 0, &firstbyte, &reqbyte, NULL, cd, &length);
   &length);  
7194  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
7195    
7196  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,
# Line 7061  field; this time it's used for rememberi Line 7244  field; this time it's used for rememberi
7244  */  */
7245    
7246  cd->final_bracount = cd->bracount;  /* Save for checking forward references */  cd->final_bracount = cd->bracount;  /* Save for checking forward references */
7247    cd->assert_depth = 0;
7248  cd->bracount = 0;  cd->bracount = 0;
7249  cd->names_found = 0;  cd->names_found = 0;
7250  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
# Line 7079  of the function here. */ Line 7263  of the function here. */
7263  ptr = (const uschar *)pattern + skipatstart;  ptr = (const uschar *)pattern + skipatstart;
7264  code = (uschar *)codestart;  code = (uschar *)codestart;
7265  *code = OP_BRA;  *code = OP_BRA;
7266  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,  (void)compile_regex(re->options, &code, &ptr, &errorcode, FALSE, FALSE, 0,
7267    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);    &firstbyte, &reqbyte, NULL, cd, NULL);
7268  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
7269  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
7270  re->flags = cd->external_flags;  re->flags = cd->external_flags;
# Line 7146  if (cd->check_lookbehind) Line 7330  if (cd->check_lookbehind)
7330        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
7331        int end_op = *be;        int end_op = *be;
7332        *be = OP_END;        *be = OP_END;
7333        fixed_length = find_fixedlength(cc, re->options, TRUE, cd);        fixed_length = find_fixedlength(cc, (re->options & PCRE_UTF8) != 0, TRUE,
7334            cd);
7335        *be = end_op;        *be = end_op;
7336        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
7337        if (fixed_length < 0)        if (fixed_length < 0)
# Line 7185  start with ^. and also when all branches Line 7370  start with ^. and also when all branches
7370    
7371  if ((re->options & PCRE_ANCHORED) == 0)  if ((re->options & PCRE_ANCHORED) == 0)
7372    {    {
7373    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))  
7374      re->options |= PCRE_ANCHORED;      re->options |= PCRE_ANCHORED;
7375    else    else
7376      {      {
7377      if (firstbyte < 0)      if (firstbyte < 0)
7378        firstbyte = find_firstassertedchar(codestart, &temp_options, FALSE);        firstbyte = find_firstassertedchar(codestart, FALSE);
7379      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */
7380        {        {
7381        int ch = firstbyte & 255;        int ch = firstbyte & 255;

Legend:
Removed from v.574  
changed lines
  Added in v.640

  ViewVC Help
Powered by ViewVC 1.1.5