/[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 556 by ph10, Tue Oct 26 11:06:44 2010 UTC revision 643 by ph10, Fri Jul 29 15:56:39 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2010 University of Cambridge             Copyright (c) 1997-2011 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 393  static const char error_texts[] = Line 393  static const char error_texts[] =
393    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
394    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
395    /* 55 */    /* 55 */
396    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
397    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
398    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
399    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
# Line 408  static const char error_texts[] = Line 408  static const char error_texts[] =
408    "different names for subpatterns of the same number are not allowed\0"    "different names for subpatterns of the same number are not allowed\0"
409    "(*MARK) must have an argument\0"    "(*MARK) must have an argument\0"
410    "this version of PCRE is not compiled with PCRE_UCP support\0"    "this version of PCRE is not compiled with PCRE_UCP support\0"
411      "\\c must be followed by an ASCII character\0"
412      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
413    ;    ;
414    
415  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 544  static const unsigned char ebcdic_charta Line 546  static const unsigned char ebcdic_charta
546  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
547    
548  static BOOL  static BOOL
549    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,
550      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
551    
552    
# Line 576  return s; Line 578  return s;
578    
579    
580  /*************************************************  /*************************************************
581    *            Check for counted repeat            *
582    *************************************************/
583    
584    /* This function is called when a '{' is encountered in a place where it might
585    start a quantifier. It looks ahead to see if it really is a quantifier or not.
586    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
587    where the ddds are digits.
588    
589    Arguments:
590      p         pointer to the first char after '{'
591    
592    Returns:    TRUE or FALSE
593    */
594    
595    static BOOL
596    is_counted_repeat(const uschar *p)
597    {
598    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
599    while ((digitab[*p] & ctype_digit) != 0) p++;
600    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
601    
602    if (*p++ != CHAR_COMMA) return FALSE;
603    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
604    
605    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
606    while ((digitab[*p] & ctype_digit) != 0) p++;
607    
608    return (*p == CHAR_RIGHT_CURLY_BRACKET);
609    }
610    
611    
612    
613    /*************************************************
614  *            Handle escapes                      *  *            Handle escapes                      *
615  *************************************************/  *************************************************/
616    
# Line 646  else Line 681  else
681      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
682      break;      break;
683    
684      /* \g must be followed by one of a number of specific things:      /* In a character class, \g is just a literal "g". Outside a character
685        class, \g must be followed by one of a number of specific things:
686    
687      (1) A number, either plain or braced. If positive, it is an absolute      (1) A number, either plain or braced. If positive, it is an absolute
688      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
# Line 663  else Line 699  else
699      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
700    
701      case CHAR_g:      case CHAR_g:
702        if (isclass) break;
703      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
704        {        {
705        c = -ESC_g;        c = -ESC_g;
# Line 841  else Line 878  else
878      break;      break;
879    
880      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
881      This coding is ASCII-specific, but then the whole concept of \cx is      An error is given if the byte following \c is not an ASCII character. This
882        coding is ASCII-specific, but then the whole concept of \cx is
883      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
884    
885      case CHAR_c:      case CHAR_c:
# Line 851  else Line 889  else
889        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
890        break;        break;
891        }        }
892    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
893  #ifndef EBCDIC  /* ASCII/UTF-8 coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
894          {
895          *errorcodeptr = ERR68;
896          break;
897          }
898      if (c >= CHAR_a && c <= CHAR_z) c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
899      c ^= 0x40;      c ^= 0x40;
900  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
901      if (c >= CHAR_a && c <= CHAR_z) c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
902      c ^= 0xC0;      c ^= 0xC0;
903  #endif  #endif
# Line 879  else Line 921  else
921    }    }
922    
923  /* Perl supports \N{name} for character names, as well as plain \N for "not  /* Perl supports \N{name} for character names, as well as plain \N for "not
924  newline". PCRE does not support \N{name}. */  newline". PCRE does not support \N{name}. However, it does support
925    quantification such as \N{2,3}. */
926    
927  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
928         !is_counted_repeat(ptr+2))
929    *errorcodeptr = ERR37;    *errorcodeptr = ERR37;
930    
931  /* If PCRE_UCP is set, we change the values for \d etc. */  /* If PCRE_UCP is set, we change the values for \d etc. */
# Line 991  return -1; Line 1035  return -1;
1035    
1036    
1037  /*************************************************  /*************************************************
 *            Check for counted repeat            *  
 *************************************************/  
   
 /* This function is called when a '{' is encountered in a place where it might  
 start a quantifier. It looks ahead to see if it really is a quantifier or not.  
 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  
 where the ddds are digits.  
   
 Arguments:  
   p         pointer to the first char after '{'  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 is_counted_repeat(const uschar *p)  
 {  
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if (*p++ != CHAR_COMMA) return FALSE;  
 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == CHAR_RIGHT_CURLY_BRACKET);  
 }  
   
   
   
 /*************************************************  
1038  *         Read repeat counts                     *  *         Read repeat counts                     *
1039  *************************************************/  *************************************************/
1040    
# Line 1099  top-level call starts at the beginning o Line 1110  top-level call starts at the beginning o
1110  start at a parenthesis. It scans along a pattern's text looking for capturing  start at a parenthesis. It scans along a pattern's text looking for capturing
1111  subpatterns, and counting them. If it finds a named pattern that matches the  subpatterns, and counting them. If it finds a named pattern that matches the
1112  name it is given, it returns its number. Alternatively, if the name is NULL, it  name it is given, it returns its number. Alternatively, if the name is NULL, it
1113  returns when it reaches a given numbered subpattern. We know that if (?P< is  returns when it reaches a given numbered subpattern. Recursion is used to keep
1114  encountered, the name will be terminated by '>' because that is checked in the  track of subpatterns that reset the capturing group numbers - the (?| feature.
1115  first pass. Recursion is used to keep track of subpatterns that reset the  
1116  capturing group numbers - the (?| feature.  This function was originally called only from the second pass, in which we know
1117    that if (?< or (?' or (?P< is encountered, the name will be correctly
1118    terminated because that is checked in the first pass. There is now one call to
1119    this function in the first pass, to check for a recursive back reference by
1120    name (so that we can make the whole group atomic). In this case, we need check
1121    only up to the current position in the pattern, and that is still OK because
1122    and previous occurrences will have been checked. To make this work, the test
1123    for "end of pattern" is a check against cd->end_pattern in the main loop,
1124    instead of looking for a binary zero. This means that the special first-pass
1125    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1126    processing items within the loop are OK, because afterwards the main loop will
1127    terminate.)
1128    
1129  Arguments:  Arguments:
1130    ptrptr       address of the current character pointer (updated)    ptrptr       address of the current character pointer (updated)
# Line 1110  Arguments: Line 1132  Arguments:
1132    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1133    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1134    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1135    utf8         TRUE if we are in UTF-8 mode    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 1203  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 1279  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 1342  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 1378  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 1455  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 1466  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 1483  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 1496  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 1519  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 1542  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 1550  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 1559  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 1576  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 1679  _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 1697  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 1755  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 1861  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 1913  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 1931  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 1959  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 2052  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 2093  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 2101  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 2149  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 2199  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 2213  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 2526  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 2570  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 2603  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 2630  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 2739  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 2926  Arguments: Line 3035  Arguments:
3035    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3036    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3037    bcptr          points to current branch chain    bcptr          points to current branch chain
3038      cond_depth     conditional nesting depth
3039    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3040    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3041                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2937  Returns: TRUE on success Line 3047  Returns: TRUE on success
3047  static BOOL  static BOOL
3048  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3049    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3050    compile_data *cd, int *lengthptr)    int cond_depth, compile_data *cd, int *lengthptr)
3051  {  {
3052  int repeat_type, op_type;  int repeat_type, op_type;
3053  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 2946  int greedy_default, greedy_non_default; Line 3056  int greedy_default, greedy_non_default;
3056  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3057  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3058  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3059  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3060  int after_manual_callout = 0;  int after_manual_callout = 0;
3061  int length_prevgroup = 0;  int length_prevgroup = 0;
3062  register int c;  register int c;
# Line 2964  uschar *previous_callout = NULL; Line 3074  uschar *previous_callout = NULL;
3074  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3075  uschar classbits[32];  uschar classbits[32];
3076    
3077    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3078    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3079    dynamically as we process the pattern. */
3080    
3081  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3082  BOOL class_utf8;  BOOL class_utf8;
3083  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
# Line 3144  for (;; ptr++) Line 3258  for (;; ptr++)
3258      previous_callout = NULL;      previous_callout = NULL;
3259      }      }
3260    
3261    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3262    
3263    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3264      {      {
3265      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3266      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
3267        {        {
3268        ptr++;        ptr++;
3269        while (*ptr != 0)        while (*ptr != 0)
3270          {          {
3271          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3272          ptr++;          ptr++;
3273  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3274          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3275  #endif  #endif
3276          }          }
# Line 3203  for (;; ptr++) Line 3317  for (;; ptr++)
3317      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3318    
3319      case CHAR_CIRCUMFLEX_ACCENT:      case CHAR_CIRCUMFLEX_ACCENT:
3320        previous = NULL;
3321      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3322        {        {
3323        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3324          *code++ = OP_CIRCM;
3325        }        }
3326      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3327      break;      break;
3328    
3329      case CHAR_DOLLAR_SIGN:      case CHAR_DOLLAR_SIGN:
3330      previous = NULL;      previous = NULL;
3331      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3332      break;      break;
3333    
3334      /* 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 3534  for (;; ptr++) Line 3649  for (;; ptr++)
3649              continue;              continue;
3650    
3651              /* 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
3652              if it was previously set by something earlier in the character              if it was previously set by something earlier in the character
3653              class. */              class. */
3654    
3655              case ESC_s:              case ESC_s:
3656              classbits[0] |= cbits[cbit_space];              classbits[0] |= cbits[cbit_space];
3657              classbits[1] |= cbits[cbit_space+1] & ~0x08;              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3658              for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3659              continue;              continue;
3660    
# Line 3959  for (;; ptr++) Line 4074  for (;; ptr++)
4074    
4075      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
4076      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4077      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4078      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4079    
4080      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
4081      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.
4082      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
4083      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
4084      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
4085      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4086    
4087  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4088      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
# Line 3978  for (;; ptr++) Line 4093  for (;; ptr++)
4093        {        {
4094        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4095    
4096        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4097    
4098        if (negate_class)        if (negate_class)
4099          {          {
4100          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4101          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4102          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4103          *code++ = class_lastchar;          *code++ = class_lastchar;
4104          break;          break;
4105          }          }
# Line 4112  for (;; ptr++) Line 4227  for (;; ptr++)
4227      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4228      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4229    
4230      /* 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
4231      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4232    
4233      tempcode = previous;      tempcode = previous;
4234    
# Line 4135  for (;; ptr++) Line 4250  for (;; ptr++)
4250        ptr++;        ptr++;
4251        }        }
4252      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4253    
4254        /* If previous was a recursion call, wrap it in atomic brackets so that
4255        previous becomes the atomic group. All recursions were so wrapped in the
4256        past, but it no longer happens for non-repeated recursions. In fact, the
4257        repeated ones could be re-implemented independently so as not to need this,
4258        but for the moment we rely on the code for repeating groups. */
4259    
4260        if (*previous == OP_RECURSE)
4261          {
4262          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4263          *previous = OP_ONCE;
4264          PUT(previous, 1, 2 + 2*LINK_SIZE);
4265          previous[2 + 2*LINK_SIZE] = OP_KET;
4266          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4267          code += 2 + 2 * LINK_SIZE;
4268          length_prevgroup = 3 + 3*LINK_SIZE;
4269    
4270          /* When actually compiling, we need to check whether this was a forward
4271          reference, and if so, adjust the offset. */
4272    
4273          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4274            {
4275            int offset = GET(cd->hwm, -LINK_SIZE);
4276            if (offset == previous + 1 - cd->start_code)
4277              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4278            }
4279          }
4280    
4281        /* Now handle repetition for the different types of item. */
4282    
4283      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4284      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 4142  for (;; ptr++) Line 4286  for (;; ptr++)
4286      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
4287      instead.  */      instead.  */
4288    
4289      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4290        {        {
4291          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4292    
4293        /* 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
4294        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
4295        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 4188  for (;; ptr++) Line 4334  for (;; ptr++)
4334      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4335      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-
4336      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4337      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
4338      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4339    
4340      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4341        {        {
4342        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4343        c = previous[1];        c = previous[1];
4344        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4345            repeat_max < 0 &&            repeat_max < 0 &&
# Line 4390  for (;; ptr++) Line 4536  for (;; ptr++)
4536  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4537               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4538  #endif  #endif
4539               *previous == OP_REF)               *previous == OP_REF ||
4540                 *previous == OP_REFI)
4541        {        {
4542        if (repeat_max == 0)        if (repeat_max == 0)
4543          {          {
# Line 4424  for (;; ptr++) Line 4571  for (;; ptr++)
4571        }        }
4572    
4573      /* 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
4574      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
4575        opcodes such as BRA and CBRA, as this is the place where they get converted
4576        into the more special varieties such as BRAPOS and SBRA. A test for >=
4577        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4578        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4579        repetition of assertions, but now it does, for Perl compatibility. */
4580    
4581      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4582        {        {
4583        register int i;        register int i;
       int ketoffset = 0;  
4584        int len = (int)(code - previous);        int len = (int)(code - previous);
4585        uschar *bralink = NULL;        uschar *bralink = NULL;
4586          uschar *brazeroptr = NULL;
4587        /* Repeating a DEFINE group is pointless */  
4588          /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4589          we just ignore the repeat. */
4590    
4591        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4592          {          goto END_REPEAT;
         *errorcodeptr = ERR55;  
         goto FAILED;  
         }  
4593    
4594        /* If the maximum repeat count is unlimited, find the end of the bracket        /* There is no sense in actually repeating assertions. The only potential
4595        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,
4596        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
4597        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. */
4598        pointer. */  
4599          if (*previous < OP_ONCE)    /* Assertion */
4600        if (repeat_max == -1)          {
4601          {          if (repeat_min > 0) goto END_REPEAT;
4602          register uschar *ket = previous;          if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
4603          do ket += GET(ket, 1); while (*ket != OP_KET);          }
         ketoffset = (int)(code - ket);  
         }  
4604    
4605        /* 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
4606        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 4473  for (;; ptr++) Line 4620  for (;; ptr++)
4620          **   goto END_REPEAT;          **   goto END_REPEAT;
4621          **   }          **   }
4622    
4623          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
4624          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
4625          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
4626          groups are referenced, we cannot do this selectively.          don't have a list of which groups are referenced, we cannot do this
4627            selectively.
4628    
4629          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
4630          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 4496  for (;; ptr++) Line 4644  for (;; ptr++)
4644              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
4645              goto END_REPEAT;              goto END_REPEAT;
4646              }              }
4647              brazeroptr = previous;    /* Save for possessive optimizing */
4648            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4649            }            }
4650    
# Line 4660  for (;; ptr++) Line 4809  for (;; ptr++)
4809            }            }
4810          }          }
4811    
4812        /* 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
4813        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
4814        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
4815        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
4816          deal with possessive ONCEs specially.
4817    
4818          Otherwise, if the quantifier was possessive, we convert the BRA code to
4819          the POS form, and the KET code to KETRPOS. (It turns out to be convenient
4820          at runtime to detect this kind of subpattern at both the start and at the
4821          end.) The use of special opcodes makes it possible to reduce greatly the
4822          stack usage in pcre_exec(). If the group is preceded by OP_BRAZERO,
4823          convert this to OP_BRAPOSZERO. Then cancel the possessive flag so that
4824          the default action below, of wrapping everything inside atomic brackets,
4825          does not happen.
4826    
4827        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
4828        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
4829        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
4830        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
4831        atomic groups at runtime, but in a different way.] */        runtime, but in a different way.] */
4832    
4833        else        else
4834          {          {
4835          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
4836          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
4837          *ketcode = OP_KETRMAX + repeat_type;  
4838          if (lengthptr == NULL && *bracode != OP_ONCE)          if (*bracode == OP_ONCE && possessive_quantifier) *bracode = OP_BRA;
4839            if (*bracode == OP_ONCE)
4840              *ketcode = OP_KETRMAX + repeat_type;
4841            else
4842            {            {
4843            uschar *scode = bracode;            if (possessive_quantifier)
           do  
4844              {              {
4845              if (could_be_empty_branch(scode, ketcode, utf8, cd))              *bracode += 1;                   /* Switch to xxxPOS opcodes */
4846                *ketcode = OP_KETRPOS;
4847                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
4848                possessive_quantifier = FALSE;
4849                }
4850              else *ketcode = OP_KETRMAX + repeat_type;
4851    
4852              if (lengthptr == NULL)
4853                {
4854                uschar *scode = bracode;
4855                do
4856                {                {
4857                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
4858                break;                  {
4859                    *bracode += OP_SBRA - OP_BRA;
4860                    break;
4861                    }
4862                  scode += GET(scode, 1);
4863                }                }
4864              scode += GET(scode, 1);              while (*scode == OP_ALT);
4865              }              }
           while (*scode == OP_ALT);  
4866            }            }
4867          }          }
4868        }        }
# Line 4709  for (;; ptr++) Line 4883  for (;; ptr++)
4883        }        }
4884    
4885      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
4886      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
4887      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
4888      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
4889      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
4890      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
4891      tempcode, not at previous, which might be the first part of a string whose  
4892      (former) last char we repeated.      Possessively repeated subpatterns have already been handled in the code
4893        just above, so possessive_quantifier is always FALSE for them at this
4894        stage.
4895    
4896        Note that the repeated item starts at tempcode, not at previous, which
4897        might be the first part of a string whose (former) last char we repeated.
4898    
4899      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
4900      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 4746  for (;; ptr++) Line 4925  for (;; ptr++)
4925          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
4926          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4927    
4928          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
4929          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
4930          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
4931          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
4932    
4933          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4934          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4935          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4936          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4937    
4938            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
4939            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
4940            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
4941            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
4942    
4943            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
4944            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
4945            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
4946            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
4947    
4948          /* Because we are moving code along, we must ensure that any          /* Because we are moving code along, we must ensure that any
4949          pending recursive references are updated. */          pending recursive references are updated. */
4950    
# Line 4808  for (;; ptr++) Line 4997  for (;; ptr++)
4997        previous = NULL;        previous = NULL;
4998        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4999        namelen = (int)(ptr - name);        namelen = (int)(ptr - name);
5000    
5001          /* It appears that Perl allows any characters whatsoever, other than
5002          a closing parenthesis, to appear in arguments, so we no longer insist on
5003          letters, digits, and underscores. */
5004    
5005        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
5006          {          {
5007          arg = ++ptr;          arg = ++ptr;
5008          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
           || *ptr == '_') ptr++;  
5009          arglen = (int)(ptr - arg);          arglen = (int)(ptr - arg);
5010          }          }
5011    
5012        if (*ptr != CHAR_RIGHT_PARENTHESIS)        if (*ptr != CHAR_RIGHT_PARENTHESIS)
5013          {          {
5014          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
# Line 4830  for (;; ptr++) Line 5022  for (;; ptr++)
5022          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
5023              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
5024            {            {
5025            /* Check for open captures before ACCEPT */            /* Check for open captures before ACCEPT and convert it to
5026              ASSERT_ACCEPT if in an assertion. */
5027    
5028            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
5029              {              {
5030              open_capitem *oc;              open_capitem *oc;
5031                if (arglen != 0)
5032                  {
5033                  *errorcodeptr = ERR59;
5034                  goto FAILED;
5035                  }
5036              cd->had_accept = TRUE;              cd->had_accept = TRUE;
5037              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5038                {                {
5039                *code++ = OP_CLOSE;                *code++ = OP_CLOSE;
5040                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
5041                }                }
5042                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5043              }              }
5044    
5045            /* Handle the cases with/without an argument */            /* Handle other cases with/without an argument */
5046    
5047            if (arglen == 0)            else if (arglen == 0)
5048              {              {
5049              if (verbs[i].op < 0)   /* Argument is mandatory */              if (verbs[i].op < 0)   /* Argument is mandatory */
5050                {                {
# Line 4856  for (;; ptr++) Line 5055  for (;; ptr++)
5055              if (*code++ == OP_THEN)              if (*code++ == OP_THEN)
5056                {                {
5057                PUT(code, 0, code - bcptr->current_branch - 1);                PUT(code, 0, code - bcptr->current_branch - 1);
5058                code += LINK_SIZE;                code += LINK_SIZE;
5059                }                }
5060              }              }
5061    
5062            else            else
# Line 4871  for (;; ptr++) Line 5070  for (;; ptr++)
5070              if (*code++ == OP_THEN_ARG)              if (*code++ == OP_THEN_ARG)
5071                {                {
5072                PUT(code, 0, code - bcptr->current_branch - 1);                PUT(code, 0, code - bcptr->current_branch - 1);
5073                code += LINK_SIZE;                code += LINK_SIZE;
5074                }                }
5075              *code++ = arglen;              *code++ = arglen;
5076              memcpy(code, arg, arglen);              memcpy(code, arg, arglen);
5077              code += arglen;              code += arglen;
# Line 5134  for (;; ptr++) Line 5333  for (;; ptr++)
5333          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5334          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5335          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5336            cd->assert_depth += 1;
5337          ptr++;          ptr++;
5338          break;          break;
5339    
# Line 5148  for (;; ptr++) Line 5348  for (;; ptr++)
5348            continue;            continue;
5349            }            }
5350          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5351            cd->assert_depth += 1;
5352          break;          break;
5353    
5354    
# Line 5157  for (;; ptr++) Line 5358  for (;; ptr++)
5358            {            {
5359            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5360            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5361              cd->assert_depth += 1;
5362            ptr += 2;            ptr += 2;
5363            break;            break;
5364    
5365            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5366            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5367              cd->assert_depth += 1;
5368            ptr += 2;            ptr += 2;
5369            break;            break;
5370    
# Line 5367  for (;; ptr++) Line 5570  for (;; ptr++)
5570          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5571          namelen = (int)(ptr - name);          namelen = (int)(ptr - name);
5572    
5573          /* 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
5574          reference number. */          a dummy reference number, because it was not used in the first pass.
5575            However, with the change of recursive back references to be atomic,
5576            we have to look for the number so that this state can be identified, as
5577            otherwise the incorrect length is computed. If it's not a backwards
5578            reference, the dummy number will do. */
5579    
5580          if (lengthptr != NULL)          if (lengthptr != NULL)
5581            {            {
5582              const uschar *temp;
5583    
5584            if (namelen == 0)            if (namelen == 0)
5585              {              {
5586              *errorcodeptr = ERR62;              *errorcodeptr = ERR62;
# Line 5387  for (;; ptr++) Line 5596  for (;; ptr++)
5596              *errorcodeptr = ERR48;              *errorcodeptr = ERR48;
5597              goto FAILED;              goto FAILED;
5598              }              }
5599            recno = 0;  
5600              /* The name table does not exist in the first pass, so we cannot
5601              do a simple search as in the code below. Instead, we have to scan the
5602              pattern to find the number. It is important that we scan it only as
5603              far as we have got because the syntax of named subpatterns has not
5604              been checked for the rest of the pattern, and find_parens() assumes
5605              correct syntax. In any case, it's a waste of resources to scan
5606              further. We stop the scan at the current point by temporarily
5607              adjusting the value of cd->endpattern. */
5608    
5609              temp = cd->end_pattern;
5610              cd->end_pattern = ptr;
5611              recno = find_parens(cd, name, namelen,
5612                (options & PCRE_EXTENDED) != 0, utf8);
5613              cd->end_pattern = temp;
5614              if (recno < 0) recno = 0;    /* Forward ref; set dummy number */
5615            }            }
5616    
5617          /* 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 5531  for (;; ptr++) Line 5755  for (;; ptr++)
5755    
5756                /* 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
5757                offset below, what it actually inserted is the reference number                offset below, what it actually inserted is the reference number
5758                of the group. */                of the group. Then remember the forward reference. */
5759    
5760                called = cd->start_code + recno;                called = cd->start_code + recno;
5761                PUTINC(cd->hwm, 0, (int)(code + 2 + LINK_SIZE - cd->start_code));                PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
5762                }                }
5763    
5764              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
5765              this is a recursive call. We check to see if this is a left              this is a recursive call. We check to see if this is a left
5766              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. We
5767                must not, however, do this check if we are in a conditional
5768                subpattern because the condition might be testing for recursion in
5769                a pattern such as /(?(R)a+|(?R)b)/, which is perfectly valid.
5770                Forever loops are also detected at runtime, so those that occur in
5771                conditional subpatterns will be picked up then. */
5772    
5773              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 && cond_depth <= 0 &&
5774                       could_be_empty(called, code, bcptr, utf8, cd))                       could_be_empty(called, code, bcptr, utf8, cd))
5775                {                {
5776                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
# Line 5549  for (;; ptr++) Line 5778  for (;; ptr++)
5778                }                }
5779              }              }
5780    
5781            /* Insert the recursion/subroutine item, automatically wrapped inside            /* Insert the recursion/subroutine item. */
5782            "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;  
   
5783            *code = OP_RECURSE;            *code = OP_RECURSE;
5784            PUT(code, 1, (int)(called - cd->start_code));            PUT(code, 1, (int)(called - cd->start_code));
5785            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;  
5786            }            }
5787    
5788          /* Can't determine a first byte now */          /* Can't determine a first byte now */
# Line 5626  for (;; ptr++) Line 5843  for (;; ptr++)
5843          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
5844          both phases.          both phases.
5845    
5846          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
5847          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. */  
5848    
5849          if (*ptr == CHAR_RIGHT_PARENTHESIS)          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5850            {            {
# Line 5639  for (;; ptr++) Line 5855  for (;; ptr++)
5855              }              }
5856            else            else
5857              {              {
             if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))  
               {  
               *code++ = OP_OPT;  
               *code++ = newoptions & PCRE_IMS;  
               }  
5858              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5859              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5860              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5861              }              }
5862    
5863            /* Change options at this level, and pass them back for use            /* Change options at this level, and pass them back for use
5864            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). */  
5865    
5866            *optionsptr = options = newoptions;            *optionsptr = options = newoptions;
5867            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
# Line 5688  for (;; ptr++) Line 5897  for (;; ptr++)
5897        skipbytes = 2;        skipbytes = 2;
5898        }        }
5899    
5900      /* Process nested bracketed regex. Assertions may not be repeated, but      /* Process nested bracketed regex. Assertions used not to be repeatable,
5901      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
5902      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
5903      compilers complain otherwise. Pass in a new setting for the ims options if      be able to pass its address because some compilers complain otherwise. */
5904      they have changed. */  
5905        previous = code;                      /* For handling repetition */
5906      previous = (bravalue >= OP_ONCE)? code : NULL;      *code = bravalue;
5907      *code = bravalue;      tempcode = code;
5908      tempcode = code;      tempreqvary = cd->req_varyopt;        /* Save value before bracket */
5909      tempreqvary = cd->req_varyopt;     /* Save value before bracket */      length_prevgroup = 0;                 /* Initialize for pre-compile phase */
5910      length_prevgroup = 0;              /* Initialize for pre-compile phase */  
5911        if (!compile_regex(
5912      if (!compile_regex(           newoptions,                      /* The complete new option state */
5913           newoptions,                   /* The complete new option state */           &tempcode,                       /* Where to put code (updated) */
5914           options & PCRE_IMS,           /* The previous ims option state */           &ptr,                            /* Input pointer (updated) */
5915           &tempcode,                    /* Where to put code (updated) */           errorcodeptr,                    /* Where to put an error message */
          &ptr,                         /* Input pointer (updated) */  
          errorcodeptr,                 /* Where to put an error message */  
5916           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
5917            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
5918           reset_bracount,               /* True if (?| group */           reset_bracount,                  /* True if (?| group */
5919           skipbytes,                    /* Skip over bracket number */           skipbytes,                       /* Skip over bracket number */
5920           &subfirstbyte,                /* For possible first char */           cond_depth +
5921           &subreqbyte,                  /* For possible last char */             ((bravalue == OP_COND)?1:0),   /* Depth of condition subpatterns */
5922           bcptr,                        /* Current branch chain */           &subfirstbyte,                   /* For possible first char */
5923           cd,                           /* Tables block */           &subreqbyte,                     /* For possible last char */
5924           (lengthptr == NULL)? NULL :   /* Actual compile phase */           bcptr,                           /* Current branch chain */
5925             &length_prevgroup           /* Pre-compile phase */           cd,                              /* Tables block */
5926             (lengthptr == NULL)? NULL :      /* Actual compile phase */
5927               &length_prevgroup              /* Pre-compile phase */
5928           ))           ))
5929        goto FAILED;        goto FAILED;
5930    
5931        if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
5932          cd->assert_depth -= 1;
5933    
5934      /* 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
5935      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 5789  for (;; ptr++) Line 6001  for (;; ptr++)
6001          goto FAILED;          goto FAILED;
6002          }          }
6003        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
6004        *code++ = OP_BRA;        code++;   /* This already contains bravalue */
6005        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
6006        *code++ = OP_KET;        *code++ = OP_KET;
6007        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
# Line 5957  for (;; ptr++) Line 6169  for (;; ptr++)
6169          }          }
6170    
6171        /* \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).
6172        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax).  */
6173    
6174        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||        if (-c == ESC_k)
           ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))  
6175          {          {
6176            if ((ptr[1] != CHAR_LESS_THAN_SIGN &&
6177              ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))
6178              {
6179              *errorcodeptr = ERR69;
6180              break;
6181              }
6182          is_recurse = FALSE;          is_recurse = FALSE;
6183          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
6184            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
6185            CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;            CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
6186          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
6187          }          }
6188    
6189        /* Back references are handled specially; must disable firstbyte if        /* Back references are handled specially; must disable firstbyte if
6190        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 5981  for (;; ptr++) Line 6198  for (;; ptr++)
6198          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
6199          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
6200          previous = code;          previous = code;
6201          *code++ = OP_REF;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_REFI : OP_REF;
6202          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
6203          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
6204          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
# Line 6089  for (;; ptr++) Line 6306  for (;; ptr++)
6306    
6307      ONE_CHAR:      ONE_CHAR:
6308      previous = code;      previous = code;
6309      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARI : OP_CHAR;
6310      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
6311    
6312      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
# Line 6153  return FALSE; Line 6370  return FALSE;
6370  /* 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
6371  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
6372  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.  
   
6373  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
6374  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
6375  value of lengthptr distinguishes the two phases.  value of lengthptr distinguishes the two phases.
6376    
6377  Arguments:  Arguments:
6378    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  
6379    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
6380    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
6381    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
6382    lookbehind     TRUE if this is a lookbehind assertion    lookbehind     TRUE if this is a lookbehind assertion
6383    reset_bracount TRUE to reset the count for each branch    reset_bracount TRUE to reset the count for each branch
6384    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes      skip this many bytes at start (for brackets and OP_COND)
6385      cond_depth     depth of nesting for conditional subpatterns
6386    firstbyteptr   place to put the first required character, or a negative number    firstbyteptr   place to put the first required character, or a negative number
6387    reqbyteptr     place to put the last required character, or a negative number    reqbyteptr     place to put the last required character, or a negative number
6388    bcptr          pointer to the chain of currently open branches    bcptr          pointer to the chain of currently open branches
# Line 6182  Returns: TRUE on success Line 6394  Returns: TRUE on success
6394  */  */
6395    
6396  static BOOL  static BOOL
6397  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,
6398    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
6399    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,    int cond_depth, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
6400    int *lengthptr)    compile_data *cd, int *lengthptr)
6401  {  {
6402  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
6403  uschar *code = *codeptr;  uschar *code = *codeptr;
# Line 6199  int branchfirstbyte, branchreqbyte; Line 6411  int branchfirstbyte, branchreqbyte;
6411  int length;  int length;
6412  int orig_bracount;  int orig_bracount;
6413  int max_bracount;  int max_bracount;
 int old_external_options = cd->external_options;  
6414  branch_chain bc;  branch_chain bc;
6415    
6416  bc.outer = bcptr;  bc.outer = bcptr;
# Line 6223  pre-compile phase to find out whether an Line 6434  pre-compile phase to find out whether an
6434    
6435  /* 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
6436  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
6437  detect groups that contain recursive back references to themselves. */  detect groups that contain recursive back references to themselves. Note that
6438    only OP_CBRA need be tested here; changing this opcode to one of its variants,
6439    e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
6440    
6441  if (*code == OP_CBRA)  if (*code == OP_CBRA)
6442    {    {
# Line 6249  for (;;) Line 6462  for (;;)
6462    
6463    if (reset_bracount) cd->bracount = orig_bracount;    if (reset_bracount) cd->bracount = orig_bracount;
6464    
   /* 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;  
     }  
   
6465    /* Set up dummy OP_REVERSE if lookbehind assertion */    /* Set up dummy OP_REVERSE if lookbehind assertion */
6466    
6467    if (lookbehind)    if (lookbehind)
# Line 6272  for (;;) Line 6476  for (;;)
6476    into the length. */    into the length. */
6477    
6478    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,
6479          &branchreqbyte, &bc, cd, (lengthptr == NULL)? NULL : &length))          &branchreqbyte, &bc, cond_depth, cd,
6480            (lengthptr == NULL)? NULL : &length))
6481      {      {
6482      *ptrptr = ptr;      *ptrptr = ptr;
6483      return FALSE;      return FALSE;
6484      }      }
6485    
   /* 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;  
   
6486    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6487    has fewer than the rest. */    has fewer than the rest. */
6488    
# Line 6347  for (;;) Line 6543  for (;;)
6543        {        {
6544        int fixed_length;        int fixed_length;
6545        *code = OP_END;        *code = OP_END;
6546        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);        fixed_length = find_fixedlength(last_branch,  (options & PCRE_UTF8) != 0,
6547            FALSE, cd);
6548        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6549        if (fixed_length == -3)        if (fixed_length == -3)
6550          {          {
# Line 6368  for (;;) Line 6565  for (;;)
6565    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
6566    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
6567    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
6568    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. */  
6569    
6570    if (*ptr != CHAR_VERTICAL_LINE)    if (*ptr != CHAR_VERTICAL_LINE)
6571      {      {
# Line 6414  for (;;) Line 6609  for (;;)
6609        cd->open_caps = cd->open_caps->next;        cd->open_caps = cd->open_caps->next;
6610        }        }
6611    
     /* Reset options if needed. */  
   
     if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)  
       {  
       *code++ = OP_OPT;  
       *code++ = oldims;  
       length += 2;  
       }  
   
6612      /* Retain the highest bracket number, in case resetting was used. */      /* Retain the highest bracket number, in case resetting was used. */
6613    
6614      cd->bracount = max_bracount;      cd->bracount = max_bracount;
# Line 6482  for (;;) Line 6668  for (;;)
6668  /* 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
6669  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
6670  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
6671  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
6672  counts, since OP_CIRC can match in the middle.  be found, because ^ generates OP_CIRCM in that mode.
6673    
6674  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.
6675  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 6504  of the more common cases more precisely. Line 6690  of the more common cases more precisely.
6690    
6691  Arguments:  Arguments:
6692    code           points to start of expression (the bracket)    code           points to start of expression (the bracket)
   options        points to the options setting  
6693    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
6694                    handles up to substring 31; after that we just have to take                    handles up to substring 31; after that we just have to take
6695                    the less precise approach                    the less precise approach
# Line 6514  Returns: TRUE or FALSE Line 6699  Returns: TRUE or FALSE
6699  */  */
6700    
6701  static BOOL  static BOOL
6702  is_anchored(register const uschar *code, int *options, unsigned int bracket_map,  is_anchored(register const uschar *code, unsigned int bracket_map,
6703    unsigned int backref_map)    unsigned int backref_map)
6704  {  {
6705  do {  do {
6706     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6707       options, PCRE_MULTILINE, FALSE);       FALSE);
6708     register int op = *scode;     register int op = *scode;
6709    
6710     /* Non-capturing brackets */     /* Non-capturing brackets */
6711    
6712     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6713           op == OP_SBRA || op == OP_SBRAPOS)
6714       {       {
6715       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6716       }       }
6717    
6718     /* Capturing brackets */     /* Capturing brackets */
6719    
6720     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6721                op == OP_SCBRA || op == OP_SCBRAPOS)
6722       {       {
6723       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6724       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
6725       if (!is_anchored(scode, options, new_map, backref_map)) return FALSE;       if (!is_anchored(scode, new_map, backref_map)) return FALSE;
6726       }       }
6727    
6728     /* Other brackets */     /* Other brackets */
6729    
6730     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)
6731       {       {
6732       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6733       }       }
6734    
6735     /* .* 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 6557  do { Line 6744  do {
6744    
6745     /* Check for explicit anchoring */     /* Check for explicit anchoring */
6746    
6747     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;  
6748     code += GET(code, 1);     code += GET(code, 1);
6749     }     }
6750  while (*code == OP_ALT);   /* Loop for each alternative */  while (*code == OP_ALT);   /* Loop for each alternative */
# Line 6595  is_startline(const uschar *code, unsigne Line 6780  is_startline(const uschar *code, unsigne
6780  {  {
6781  do {  do {
6782     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6783       NULL, 0, FALSE);       FALSE);
6784     register int op = *scode;     register int op = *scode;
6785    
6786     /* 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 6622  do { Line 6807  do {
6807         scode += 1 + LINK_SIZE;         scode += 1 + LINK_SIZE;
6808         break;         break;
6809         }         }
6810       scode = first_significant_code(scode, NULL, 0, FALSE);       scode = first_significant_code(scode, FALSE);
6811       op = *scode;       op = *scode;
6812       }       }
6813    
6814     /* Non-capturing brackets */     /* Non-capturing brackets */
6815    
6816     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6817           op == OP_SBRA || op == OP_SBRAPOS)
6818       {       {
6819       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6820       }       }
6821    
6822     /* Capturing brackets */     /* Capturing brackets */
6823    
6824     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6825                op == OP_SCBRA || op == OP_SCBRAPOS)
6826       {       {
6827       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6828       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
# Line 6659  do { Line 6846  do {
6846    
6847     /* Check for explicit circumflex */     /* Check for explicit circumflex */
6848    
6849     else if (op != OP_CIRC) return FALSE;     else if (op != OP_CIRC && op != OP_CIRCM) return FALSE;
6850    
6851     /* Move on to the next alternative */     /* Move on to the next alternative */
6852    
# Line 6685  we return that char, otherwise -1. Line 6872  we return that char, otherwise -1.
6872    
6873  Arguments:  Arguments:
6874    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)  
6875    inassert   TRUE if in an assertion    inassert   TRUE if in an assertion
6876    
6877  Returns:     -1 or the fixed first char  Returns:     -1 or the fixed first char
6878  */  */
6879    
6880  static int  static int
6881  find_firstassertedchar(const uschar *code, int *options, BOOL inassert)  find_firstassertedchar(const uschar *code, BOOL inassert)
6882  {  {
6883  register int c = -1;  register int c = -1;
6884  do {  do {
6885     int d;     int d;
6886     const uschar *scode =     int xl = (*code == OP_CBRA || *code == OP_SCBRA ||
6887       first_significant_code(code + 1+LINK_SIZE, options, PCRE_CASELESS, TRUE);               *code == OP_CBRAPOS || *code == OP_SCBRAPOS)? 2:0;
6888       const uschar *scode = first_significant_code(code + 1+LINK_SIZE + xl, TRUE);
6889     register int op = *scode;     register int op = *scode;
6890    
6891     switch(op)     switch(op)
# Line 6707  do { Line 6894  do {
6894       return -1;       return -1;
6895    
6896       case OP_BRA:       case OP_BRA:
6897         case OP_BRAPOS:
6898       case OP_CBRA:       case OP_CBRA:
6899         case OP_SCBRA:
6900         case OP_CBRAPOS:
6901         case OP_SCBRAPOS:
6902       case OP_ASSERT:       case OP_ASSERT:
6903       case OP_ONCE:       case OP_ONCE:
6904       case OP_COND:       case OP_COND:
6905       if ((d = find_firstassertedchar(scode, options, op == OP_ASSERT)) < 0)       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)
6906         return -1;         return -1;
6907       if (c < 0) c = d; else if (c != d) return -1;       if (c < 0) c = d; else if (c != d) return -1;
6908       break;       break;
6909    
6910       case OP_EXACT:       /* Fall through */       case OP_EXACT:
6911       scode += 2;       scode += 2;
6912         /* Fall through */
6913    
6914       case OP_CHAR:       case OP_CHAR:
      case OP_CHARNC:  
6915       case OP_PLUS:       case OP_PLUS:
6916       case OP_MINPLUS:       case OP_MINPLUS:
6917       case OP_POSPLUS:       case OP_POSPLUS:
6918       if (!inassert) return -1;       if (!inassert) return -1;
6919       if (c < 0)       if (c < 0) c = scode[1];
6920         {         else if (c != scode[1]) return -1;
6921         c = scode[1];       break;
6922         if ((*options & PCRE_CASELESS) != 0) c |= REQ_CASELESS;  
6923         }       case OP_EXACTI:
6924       else if (c != scode[1]) return -1;       scode += 2;
6925         /* Fall through */
6926    
6927         case OP_CHARI:
6928         case OP_PLUSI:
6929         case OP_MINPLUSI:
6930         case OP_POSPLUSI:
6931         if (!inassert) return -1;
6932         if (c < 0) c = scode[1] | REQ_CASELESS;
6933           else if (c != scode[1]) return -1;
6934       break;       break;
6935       }       }
6936    
# Line 6853  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7053  while (ptr[skipatstart] == CHAR_LEFT_PAR
7053      { skipatstart += 7; options |= PCRE_UTF8; continue; }      { skipatstart += 7; options |= PCRE_UTF8; continue; }
7054    else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)
7055      { skipatstart += 6; options |= PCRE_UCP; continue; }      { skipatstart += 6; options |= PCRE_UCP; continue; }
7056      else if (strncmp((char *)(ptr+skipatstart+2), STRING_NO_START_OPT_RIGHTPAR, 13) == 0)
7057        { skipatstart += 15; options |= PCRE_NO_START_OPTIMIZE; continue; }
7058    
7059    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
7060      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
# Line 6879  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7081  while (ptr[skipatstart] == CHAR_LEFT_PAR
7081    
7082  utf8 = (options & PCRE_UTF8) != 0;  utf8 = (options & PCRE_UTF8) != 0;
7083    
7084  /* 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
7085    return of an error code from _pcre_valid_utf8() is a new feature, introduced in
7086    release 8.13. It is passed back from pcre_[dfa_]exec(), but at the moment is
7087    not used here. */
7088    
7089  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
7090  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
7091       (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)       (errorcode = _pcre_valid_utf8((USPTR)pattern, -1, erroroffset)) != 0)
7092    {    {
7093    errorcode = ERR44;    errorcode = ERR44;
7094    goto PCRE_EARLY_ERROR_RETURN2;    goto PCRE_EARLY_ERROR_RETURN2;
# Line 6908  if ((options & PCRE_UCP) != 0) Line 7113  if ((options & PCRE_UCP) != 0)
7113    
7114  /* Check validity of \R options. */  /* Check validity of \R options. */
7115    
7116  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  if ((options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) ==
7117         (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
7118    {    {
7119    case 0:    errorcode = ERR56;
7120    case PCRE_BSR_ANYCRLF:    goto PCRE_EARLY_ERROR_RETURN;
   case PCRE_BSR_UNICODE:  
   break;  
   default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;  
7121    }    }
7122    
7123  /* 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 6999  outside can help speed up starting point Line 7202  outside can help speed up starting point
7202  ptr += skipatstart;  ptr += skipatstart;
7203  code = cworkspace;  code = cworkspace;
7204  *code = OP_BRA;  *code = OP_BRA;
7205  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
7206    &code, &ptr, &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd,    FALSE, 0, 0, &firstbyte, &reqbyte, NULL, cd, &length);
   &length);  
7207  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
7208    
7209  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,
# Line 7055  field; this time it's used for rememberi Line 7257  field; this time it's used for rememberi
7257  */  */
7258    
7259  cd->final_bracount = cd->bracount;  /* Save for checking forward references */  cd->final_bracount = cd->bracount;  /* Save for checking forward references */
7260    cd->assert_depth = 0;
7261  cd->bracount = 0;  cd->bracount = 0;
7262  cd->names_found = 0;  cd->names_found = 0;
7263  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
# Line 7073  of the function here. */ Line 7276  of the function here. */
7276  ptr = (const uschar *)pattern + skipatstart;  ptr = (const uschar *)pattern + skipatstart;
7277  code = (uschar *)codestart;  code = (uschar *)codestart;
7278  *code = OP_BRA;  *code = OP_BRA;
7279  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,  (void)compile_regex(re->options, &code, &ptr, &errorcode, FALSE, FALSE, 0, 0,
7280    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);    &firstbyte, &reqbyte, NULL, cd, NULL);
7281  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
7282  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
7283  re->flags = cd->external_flags;  re->flags = cd->external_flags;
# Line 7140  if (cd->check_lookbehind) Line 7343  if (cd->check_lookbehind)
7343        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
7344        int end_op = *be;        int end_op = *be;
7345        *be = OP_END;        *be = OP_END;
7346        fixed_length = find_fixedlength(cc, re->options, TRUE, cd);        fixed_length = find_fixedlength(cc, (re->options & PCRE_UTF8) != 0, TRUE,
7347            cd);
7348        *be = end_op;        *be = end_op;
7349        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
7350        if (fixed_length < 0)        if (fixed_length < 0)
# Line 7179  start with ^. and also when all branches Line 7383  start with ^. and also when all branches
7383    
7384  if ((re->options & PCRE_ANCHORED) == 0)  if ((re->options & PCRE_ANCHORED) == 0)
7385    {    {
7386    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))  
7387      re->options |= PCRE_ANCHORED;      re->options |= PCRE_ANCHORED;
7388    else    else
7389      {      {
7390      if (firstbyte < 0)      if (firstbyte < 0)
7391        firstbyte = find_firstassertedchar(codestart, &temp_options, FALSE);        firstbyte = find_firstassertedchar(codestart, FALSE);
7392      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */
7393        {        {
7394        int ch = firstbyte & 255;        int ch = firstbyte & 255;

Legend:
Removed from v.556  
changed lines
  Added in v.643

webmaster@exim.org
ViewVC Help
Powered by ViewVC 1.1.12