/[pcre]/code/branches/pcre16/pcre_compile.c
ViewVC logotype

Diff of /code/branches/pcre16/pcre_compile.c

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

revision 574 by ph10, Sat Nov 20 17:47:27 2010 UTC revision 629 by ph10, Fri Jul 22 09:18:11 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 408  static const char error_texts[] = Line 408  static const char error_texts[] =
408    "different names for subpatterns of the same number are not allowed\0"    "different names for subpatterns of the same number are not allowed\0"
409    "(*MARK) must have an argument\0"    "(*MARK) must have an argument\0"
410    "this version of PCRE is not compiled with PCRE_UCP support\0"    "this version of PCRE is not compiled with PCRE_UCP support\0"
411    "\\c must be followed by an ASCII character\0"    "\\c must be followed by an ASCII character\0"
412      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
413    ;    ;
414    
415  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 545  static const unsigned char ebcdic_charta Line 546  static const unsigned char ebcdic_charta
546  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
547    
548  static BOOL  static BOOL
549    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int *,
550      int *, int *, branch_chain *, compile_data *, int *);      int *, branch_chain *, compile_data *, int *);
551    
552    
553    
# Line 857  else Line 858  else
858      if (c > 127)  /* Excludes all non-ASCII in either mode */      if (c > 127)  /* Excludes all non-ASCII in either mode */
859        {        {
860        *errorcodeptr = ERR68;        *errorcodeptr = ERR68;
861        break;        break;
862        }        }
863      if (c >= CHAR_a && c <= CHAR_z) c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
864      c ^= 0x40;      c ^= 0x40;
865  #else             /* EBCDIC coding */  #else             /* EBCDIC coding */
# Line 1105  top-level call starts at the beginning o Line 1106  top-level call starts at the beginning o
1106  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
1107  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
1108  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
1109  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
1110  encountered, the name will be terminated by '>' because that is checked in the  track of subpatterns that reset the capturing group numbers - the (?| feature.
1111  first pass. Recursion is used to keep track of subpatterns that reset the  
1112  capturing group numbers - the (?| feature.  This function was originally called only from the second pass, in which we know
1113    that if (?< or (?' or (?P< is encountered, the name will be correctly
1114    terminated because that is checked in the first pass. There is now one call to
1115    this function in the first pass, to check for a recursive back reference by
1116    name (so that we can make the whole group atomic). In this case, we need check
1117    only up to the current position in the pattern, and that is still OK because
1118    and previous occurrences will have been checked. To make this work, the test
1119    for "end of pattern" is a check against cd->end_pattern in the main loop,
1120    instead of looking for a binary zero. This means that the special first-pass
1121    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1122    processing items within the loop are OK, because afterwards the main loop will
1123    terminate.)
1124    
1125  Arguments:  Arguments:
1126    ptrptr       address of the current character pointer (updated)    ptrptr       address of the current character pointer (updated)
# Line 1116  Arguments: Line 1128  Arguments:
1128    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1129    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1130    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1131    utf8         TRUE if we are in UTF-8 mode    utf8         TRUE if we are in UTF-8 mode
1132    count        pointer to the current capturing subpattern number (updated)    count        pointer to the current capturing subpattern number (updated)
1133    
1134  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
# Line 1209  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1221  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1221    }    }
1222    
1223  /* Past any initial parenthesis handling, scan for parentheses or vertical  /* Past any initial parenthesis handling, scan for parentheses or vertical
1224  bars. */  bars. Stop if we get to cd->end_pattern. Note that this is important for the
1225    first-pass call when this value is temporarily adjusted to stop at the current
1226    position. So DO NOT change this to a test for binary zero. */
1227    
1228  for (; *ptr != 0; ptr++)  for (; ptr < cd->end_pattern; ptr++)
1229    {    {
1230    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1231    
# Line 1285  for (; *ptr != 0; ptr++) Line 1299  for (; *ptr != 0; ptr++)
1299    
1300    if (xmode && *ptr == CHAR_NUMBER_SIGN)    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1301      {      {
1302      ptr++;      ptr++;
1303      while (*ptr != 0)      while (*ptr != 0)
1304        {        {
1305        if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }        if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1306        ptr++;        ptr++;
1307  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1308        if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;        if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1309  #endif  #endif
1310        }        }
# Line 1348  Arguments: Line 1362  Arguments:
1362    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1363    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1364    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1365    utf8         TRUE if we are in UTF-8 mode    utf8         TRUE if we are in UTF-8 mode
1366    
1367  Returns:       the number of the found subpattern, or -1 if not found  Returns:       the number of the found subpattern, or -1 if not found
1368  */  */
# Line 1384  return rc; Line 1398  return rc;
1398    
1399  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1400  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
1401  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
1402  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
1403  assertions, and also the \b assertion; for others it does not.  does not.
1404    
1405  Arguments:  Arguments:
1406    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  
1407    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1408    
1409  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1410  */  */
1411    
1412  static const uschar*  static const uschar*
1413  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1414  {  {
1415  for (;;)  for (;;)
1416    {    {
1417    switch ((int)*code)    switch ((int)*code)
1418      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1419      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1420      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1421      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 1461  and doing the check at the end; a flag s Line 1465  and doing the check at the end; a flag s
1465    
1466  Arguments:  Arguments:
1467    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1468    options  the compiling options    utf8     TRUE in UTF-8 mode
1469    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1470    cd       the "compile data" structure    cd       the "compile data" structure
1471    
# Line 1472  Returns:   the fixed length, Line 1476  Returns:   the fixed length,
1476  */  */
1477    
1478  static int  static int
1479  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1480  {  {
1481  int length = -1;  int length = -1;
1482    
# Line 1489  for (;;) Line 1493  for (;;)
1493    register int op = *cc;    register int op = *cc;
1494    switch (op)    switch (op)
1495      {      {
1496        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1497        OP_BRA (normal non-capturing bracket) because the other variants of these
1498        opcodes are all concerned with unlimited repeated groups, which of course
1499        are not of fixed length. They will cause a -1 response from the default
1500        case of this switch. */
1501    
1502      case OP_CBRA:      case OP_CBRA:
1503      case OP_BRA:      case OP_BRA:
1504      case OP_ONCE:      case OP_ONCE:
1505      case OP_COND:      case OP_COND:
1506      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1507      if (d < 0) return d;      if (d < 0) return d;
1508      branchlength += d;      branchlength += d;
1509      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1502  for (;;) Line 1512  for (;;)
1512    
1513      /* 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
1514      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
1515      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.
1516        Note that we must not include the OP_KETRxxx opcodes here, because they
1517        all imply an unlimited repeat. */
1518    
1519      case OP_ALT:      case OP_ALT:
1520      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1521      case OP_END:      case OP_END:
1522      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1523        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
# Line 1525  for (;;) Line 1535  for (;;)
1535      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1536      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1537      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                /* Recursion */
1538      d = find_fixedlength(cs + 2, options, atend, cd);      d = find_fixedlength(cs + 2, utf8, atend, cd);
1539      if (d < 0) return d;      if (d < 0) return d;
1540      branchlength += d;      branchlength += d;
1541      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1548  for (;;) Line 1558  for (;;)
1558      case OP_RREF:      case OP_RREF:
1559      case OP_NRREF:      case OP_NRREF:
1560      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
1561      case OP_CALLOUT:      case OP_CALLOUT:
1562      case OP_SOD:      case OP_SOD:
1563      case OP_SOM:      case OP_SOM:
# Line 1556  for (;;) Line 1565  for (;;)
1565      case OP_EOD:      case OP_EOD:
1566      case OP_EODN:      case OP_EODN:
1567      case OP_CIRC:      case OP_CIRC:
1568        case OP_CIRCM:
1569      case OP_DOLL:      case OP_DOLL:
1570        case OP_DOLLM:
1571      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1572      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1573      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 1565  for (;;) Line 1576  for (;;)
1576      /* Handle literal characters */      /* Handle literal characters */
1577    
1578      case OP_CHAR:      case OP_CHAR:
1579      case OP_CHARNC:      case OP_CHARI:
1580      case OP_NOT:      case OP_NOT:
1581        case OP_NOTI:
1582      branchlength++;      branchlength++;
1583      cc += 2;      cc += 2;
1584  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1585      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];  
1586  #endif  #endif
1587      break;      break;
1588    
# Line 1582  for (;;) Line 1593  for (;;)
1593      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1594      cc += 4;      cc += 4;
1595  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1596      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];  
1597  #endif  #endif
1598      break;      break;
1599    
# Line 1685  _pcre_find_bracket(const uschar *code, B Line 1695  _pcre_find_bracket(const uschar *code, B
1695  for (;;)  for (;;)
1696    {    {
1697    register int c = *code;    register int c = *code;
1698    
1699    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1700    
1701    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1703  for (;;) Line 1714  for (;;)
1714    
1715    /* Handle capturing bracket */    /* Handle capturing bracket */
1716    
1717    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1718               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1719      {      {
1720      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1721      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1761  for (;;) Line 1773  for (;;)
1773      if (utf8) switch(c)      if (utf8) switch(c)
1774        {        {
1775        case OP_CHAR:        case OP_CHAR:
1776        case OP_CHARNC:        case OP_CHARI:
1777        case OP_EXACT:        case OP_EXACT:
1778          case OP_EXACTI:
1779        case OP_UPTO:        case OP_UPTO:
1780          case OP_UPTOI:
1781        case OP_MINUPTO:        case OP_MINUPTO:
1782          case OP_MINUPTOI:
1783        case OP_POSUPTO:        case OP_POSUPTO:
1784          case OP_POSUPTOI:
1785        case OP_STAR:        case OP_STAR:
1786          case OP_STARI:
1787        case OP_MINSTAR:        case OP_MINSTAR:
1788          case OP_MINSTARI:
1789        case OP_POSSTAR:        case OP_POSSTAR:
1790          case OP_POSSTARI:
1791        case OP_PLUS:        case OP_PLUS:
1792          case OP_PLUSI:
1793        case OP_MINPLUS:        case OP_MINPLUS:
1794          case OP_MINPLUSI:
1795        case OP_POSPLUS:        case OP_POSPLUS:
1796          case OP_POSPLUSI:
1797        case OP_QUERY:        case OP_QUERY:
1798          case OP_QUERYI:
1799        case OP_MINQUERY:        case OP_MINQUERY:
1800          case OP_MINQUERYI:
1801        case OP_POSQUERY:        case OP_POSQUERY:
1802          case OP_POSQUERYI:
1803        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1804        break;        break;
1805        }        }
# Line 1867  for (;;) Line 1892  for (;;)
1892      if (utf8) switch(c)      if (utf8) switch(c)
1893        {        {
1894        case OP_CHAR:        case OP_CHAR:
1895        case OP_CHARNC:        case OP_CHARI:
1896        case OP_EXACT:        case OP_EXACT:
1897          case OP_EXACTI:
1898        case OP_UPTO:        case OP_UPTO:
1899          case OP_UPTOI:
1900        case OP_MINUPTO:        case OP_MINUPTO:
1901          case OP_MINUPTOI:
1902        case OP_POSUPTO:        case OP_POSUPTO:
1903          case OP_POSUPTOI:
1904        case OP_STAR:        case OP_STAR:
1905          case OP_STARI:
1906        case OP_MINSTAR:        case OP_MINSTAR:
1907          case OP_MINSTARI:
1908        case OP_POSSTAR:        case OP_POSSTAR:
1909          case OP_POSSTARI:
1910        case OP_PLUS:        case OP_PLUS:
1911          case OP_PLUSI:
1912        case OP_MINPLUS:        case OP_MINPLUS:
1913          case OP_MINPLUSI:
1914        case OP_POSPLUS:        case OP_POSPLUS:
1915          case OP_POSPLUSI:
1916        case OP_QUERY:        case OP_QUERY:
1917          case OP_QUERYI:
1918        case OP_MINQUERY:        case OP_MINQUERY:
1919          case OP_MINQUERYI:
1920        case OP_POSQUERY:        case OP_POSQUERY:
1921          case OP_POSQUERYI:
1922        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1923        break;        break;
1924        }        }
# Line 1919  could_be_empty_branch(const uschar *code Line 1957  could_be_empty_branch(const uschar *code
1957    compile_data *cd)    compile_data *cd)
1958  {  {
1959  register int c;  register int c;
1960  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
1961       code < endcode;       code < endcode;
1962       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
1963    {    {
1964    const uschar *ccode;    const uschar *ccode;
1965    
# Line 1937  for (code = first_significant_code(code Line 1975  for (code = first_significant_code(code
1975      continue;      continue;
1976      }      }
1977    
   /* 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;  
     }  
   
1978    /* For a recursion/subroutine call, if its end has been reached, which    /* For a recursion/subroutine call, if its end has been reached, which
1979    implies a subroutine call, we can scan it. */    implies a backward reference subroutine call, we can scan it. If it's a
1980      forward reference subroutine call, we can't. To detect forward reference
1981      we have to scan up the list that is kept in the workspace. This function is
1982      called only when doing the real compile, not during the pre-compile that
1983      measures the size of the compiled pattern. */
1984    
1985    if (c == OP_RECURSE)    if (c == OP_RECURSE)
1986      {      {
1987      BOOL empty_branch = FALSE;      const uschar *scode;
1988      const uschar *scode = cd->start_code + GET(code, 1);      BOOL empty_branch;
1989    
1990        /* Test for forward reference */
1991    
1992        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
1993          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
1994    
1995        /* Not a forward reference, test for completed backward reference */
1996    
1997        empty_branch = FALSE;
1998        scode = cd->start_code + GET(code, 1);
1999      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2000    
2001        /* Completed backwards reference */
2002    
2003      do      do
2004        {        {
2005        if (could_be_empty_branch(scode, endcode, utf8, cd))        if (could_be_empty_branch(scode, endcode, utf8, cd))
# Line 1965  for (code = first_significant_code(code Line 2010  for (code = first_significant_code(code
2010        scode += GET(scode, 1);        scode += GET(scode, 1);
2011        }        }
2012      while (*scode == OP_ALT);      while (*scode == OP_ALT);
2013    
2014      if (!empty_branch) return FALSE;  /* All branches are non-empty */      if (!empty_branch) return FALSE;  /* All branches are non-empty */
2015      continue;      continue;
2016      }      }
2017    
2018      /* Groups with zero repeats can of course be empty; skip them. */
2019    
2020      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2021          c == OP_BRAPOSZERO)
2022        {
2023        code += _pcre_OP_lengths[c];
2024        do code += GET(code, 1); while (*code == OP_ALT);
2025        c = *code;
2026        continue;
2027        }
2028    
2029      /* A nested group that is already marked as "could be empty" can just be
2030      skipped. */
2031    
2032      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2033          c == OP_SCBRA || c == OP_SCBRAPOS)
2034        {
2035        do code += GET(code, 1); while (*code == OP_ALT);
2036        c = *code;
2037        continue;
2038        }
2039    
2040    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2041    
2042    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2043          c == OP_CBRA || c == OP_CBRAPOS ||
2044          c == OP_ONCE || c == OP_COND)
2045      {      {
2046      BOOL empty_branch;      BOOL empty_branch;
2047      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 2058  for (code = first_significant_code(code Line 2128  for (code = first_significant_code(code
2128      case OP_ALLANY:      case OP_ALLANY:
2129      case OP_ANYBYTE:      case OP_ANYBYTE:
2130      case OP_CHAR:      case OP_CHAR:
2131      case OP_CHARNC:      case OP_CHARI:
2132      case OP_NOT:      case OP_NOT:
2133        case OP_NOTI:
2134      case OP_PLUS:      case OP_PLUS:
2135      case OP_MINPLUS:      case OP_MINPLUS:
2136      case OP_POSPLUS:      case OP_POSPLUS:
# Line 2099  for (code = first_significant_code(code Line 2170  for (code = first_significant_code(code
2170      case OP_KET:      case OP_KET:
2171      case OP_KETRMAX:      case OP_KETRMAX:
2172      case OP_KETRMIN:      case OP_KETRMIN:
2173        case OP_KETRPOS:
2174      case OP_ALT:      case OP_ALT:
2175      return TRUE;      return TRUE;
2176    
# Line 2107  for (code = first_significant_code(code Line 2179  for (code = first_significant_code(code
2179    
2180  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2181      case OP_STAR:      case OP_STAR:
2182        case OP_STARI:
2183      case OP_MINSTAR:      case OP_MINSTAR:
2184        case OP_MINSTARI:
2185      case OP_POSSTAR:      case OP_POSSTAR:
2186        case OP_POSSTARI:
2187      case OP_QUERY:      case OP_QUERY:
2188        case OP_QUERYI:
2189      case OP_MINQUERY:      case OP_MINQUERY:
2190        case OP_MINQUERYI:
2191      case OP_POSQUERY:      case OP_POSQUERY:
2192        case OP_POSQUERYI:
2193      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2194      break;      break;
2195    
2196      case OP_UPTO:      case OP_UPTO:
2197        case OP_UPTOI:
2198      case OP_MINUPTO:      case OP_MINUPTO:
2199        case OP_MINUPTOI:
2200      case OP_POSUPTO:      case OP_POSUPTO:
2201        case OP_POSUPTOI:
2202      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2203      break;      break;
2204  #endif  #endif
# Line 2155  return TRUE; Line 2236  return TRUE;
2236  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
2237  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,
2238  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.
2239    This function is called only during the real compile, not during the
2240    pre-compile.
2241    
2242  Arguments:  Arguments:
2243    code        points to start of the recursion    code        points to start of the recursion
# Line 2532  if ((options & PCRE_EXTENDED) != 0) Line 2615  if ((options & PCRE_EXTENDED) != 0)
2615      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2616      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2617        {        {
2618        ptr++;        ptr++;
2619        while (*ptr != 0)        while (*ptr != 0)
2620          {          {
2621          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2622          ptr++;          ptr++;
2623  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2624          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2625  #endif  #endif
2626          }          }
# Line 2576  if ((options & PCRE_EXTENDED) != 0) Line 2659  if ((options & PCRE_EXTENDED) != 0)
2659      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2660      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2661        {        {
2662        ptr++;        ptr++;
2663        while (*ptr != 0)        while (*ptr != 0)
2664          {          {
2665          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2666          ptr++;          ptr++;
2667  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2668          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2669  #endif  #endif
2670          }          }
# Line 2609  if (next >= 0) switch(op_code) Line 2692  if (next >= 0) switch(op_code)
2692  #endif  #endif
2693    return c != next;    return c != next;
2694    
2695    /* 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
2696    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
2697    high-valued characters. */    high-valued characters. */
2698    
2699    case OP_CHARNC:    case OP_CHARI:
2700  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2701    GETCHARTEST(c, previous);    GETCHARTEST(c, previous);
2702  #else  #else
# Line 2636  if (next >= 0) switch(op_code) Line 2719  if (next >= 0) switch(op_code)
2719  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2720    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2721    
2722    /* 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
2723      opcodes are not used for multi-byte characters, because they are coded using
2724      an XCLASS instead. */
2725    
2726    case OP_NOT:    case OP_NOT:
2727      return (c = *previous) == next;
2728    
2729      case OP_NOTI:
2730    if ((c = *previous) == next) return TRUE;    if ((c = *previous) == next) return TRUE;
   if ((options & PCRE_CASELESS) == 0) return FALSE;  
2731  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2732    if (utf8)    if (utf8)
2733      {      {
# Line 2745  replaced by OP_PROP codes when PCRE_UCP Line 2832  replaced by OP_PROP codes when PCRE_UCP
2832  switch(op_code)  switch(op_code)
2833    {    {
2834    case OP_CHAR:    case OP_CHAR:
2835    case OP_CHARNC:    case OP_CHARI:
2836  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2837    GETCHARTEST(c, previous);    GETCHARTEST(c, previous);
2838  #else  #else
# Line 3157  for (;; ptr++) Line 3244  for (;; ptr++)
3244      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3245      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
3246        {        {
3247        ptr++;        ptr++;
3248        while (*ptr != 0)        while (*ptr != 0)
3249          {          {
3250          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3251          ptr++;          ptr++;
3252  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3253          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3254  #endif  #endif
3255          }          }
# Line 3209  for (;; ptr++) Line 3296  for (;; ptr++)
3296      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3297    
3298      case CHAR_CIRCUMFLEX_ACCENT:      case CHAR_CIRCUMFLEX_ACCENT:
3299        previous = NULL;
3300      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3301        {        {
3302        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3303          *code++ = OP_CIRCM;
3304        }        }
3305      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3306      break;      break;
3307    
3308      case CHAR_DOLLAR_SIGN:      case CHAR_DOLLAR_SIGN:
3309      previous = NULL;      previous = NULL;
3310      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3311      break;      break;
3312    
3313      /* There can never be a first char if '.' is first, whatever happens about      /* There can never be a first char if '.' is first, whatever happens about
# Line 3540  for (;; ptr++) Line 3628  for (;; ptr++)
3628              continue;              continue;
3629    
3630              /* 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
3631              if it was previously set by something earlier in the character              if it was previously set by something earlier in the character
3632              class. */              class. */
3633    
3634              case ESC_s:              case ESC_s:
3635              classbits[0] |= cbits[cbit_space];              classbits[0] |= cbits[cbit_space];
3636              classbits[1] |= cbits[cbit_space+1] & ~0x08;              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3637              for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3638              continue;              continue;
3639    
# Line 3965  for (;; ptr++) Line 4053  for (;; ptr++)
4053    
4054      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
4055      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4056      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4057      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4058    
4059      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
4060      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.
4061      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
4062      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
4063      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
4064      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4065    
4066  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4067      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
# Line 3984  for (;; ptr++) Line 4072  for (;; ptr++)
4072        {        {
4073        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4074    
4075        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4076    
4077        if (negate_class)        if (negate_class)
4078          {          {
4079          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4080          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4081          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4082          *code++ = class_lastchar;          *code++ = class_lastchar;
4083          break;          break;
4084          }          }
# Line 4141  for (;; ptr++) Line 4229  for (;; ptr++)
4229        ptr++;        ptr++;
4230        }        }
4231      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4232    
4233        /* If previous was a recursion call, wrap it in atomic brackets so that
4234        previous becomes the atomic group. All recursions were so wrapped in the
4235        past, but it no longer happens for non-repeated recursions. In fact, the
4236        repeated ones could be re-implemented independently so as not to need this,
4237        but for the moment we rely on the code for repeating groups. */
4238    
4239        if (*previous == OP_RECURSE)
4240          {
4241          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4242          *previous = OP_ONCE;
4243          PUT(previous, 1, 2 + 2*LINK_SIZE);
4244          previous[2 + 2*LINK_SIZE] = OP_KET;
4245          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4246          code += 2 + 2 * LINK_SIZE;
4247          length_prevgroup = 3 + 3*LINK_SIZE;
4248    
4249          /* When actually compiling, we need to check whether this was a forward
4250          reference, and if so, adjust the offset. */
4251    
4252          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4253            {
4254            int offset = GET(cd->hwm, -LINK_SIZE);
4255            if (offset == previous + 1 - cd->start_code)
4256              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4257            }
4258          }
4259    
4260        /* Now handle repetition for the different types of item. */
4261    
4262      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4263      repeat item instead. If a char item has a minumum of more than one, ensure      repeat item instead. If a char item has a minumum of more than one, ensure
# Line 4148  for (;; ptr++) Line 4265  for (;; ptr++)
4265      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
4266      instead.  */      instead.  */
4267    
4268      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4269        {        {
4270          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4271    
4272        /* 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
4273        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
4274        hold the length of the character in bytes, plus 0x80 to flag that it's a        hold the length of the character in bytes, plus 0x80 to flag that it's a
# Line 4194  for (;; ptr++) Line 4313  for (;; ptr++)
4313      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4314      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-
4315      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4316      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
4317      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4318    
4319      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4320        {        {
4321        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4322        c = previous[1];        c = previous[1];
4323        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4324            repeat_max < 0 &&            repeat_max < 0 &&
# Line 4396  for (;; ptr++) Line 4515  for (;; ptr++)
4515  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4516               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4517  #endif  #endif
4518               *previous == OP_REF)               *previous == OP_REF ||
4519                 *previous == OP_REFI)
4520        {        {
4521        if (repeat_max == 0)        if (repeat_max == 0)
4522          {          {
# Line 4430  for (;; ptr++) Line 4550  for (;; ptr++)
4550        }        }
4551    
4552      /* 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
4553      cases. */      cases. Note that at this point we can encounter only the "basic" BRA and
4554        KET opcodes, as this is the place where they get converted into the more
4555        special varieties. */
4556    
4557      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous == OP_BRA  || *previous == OP_CBRA ||
4558               *previous == OP_ONCE || *previous == OP_COND)               *previous == OP_ONCE || *previous == OP_COND)
4559        {        {
4560        register int i;        register int i;
       int ketoffset = 0;  
4561        int len = (int)(code - previous);        int len = (int)(code - previous);
4562        uschar *bralink = NULL;        uschar *bralink = NULL;
4563          uschar *brazeroptr = NULL;
4564    
4565        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless */
4566    
4567        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
# Line 4448  for (;; ptr++) Line 4570  for (;; ptr++)
4570          goto FAILED;          goto FAILED;
4571          }          }
4572    
       /* If the maximum repeat count is unlimited, find the end of the bracket  
       by scanning through from the start, and compute the offset back to it  
       from the current code pointer. There may be an OP_OPT setting following  
       the final KET, so we can't find the end just by going back from the code  
       pointer. */  
   
       if (repeat_max == -1)  
         {  
         register uschar *ket = previous;  
         do ket += GET(ket, 1); while (*ket != OP_KET);  
         ketoffset = (int)(code - ket);  
         }  
   
4573        /* 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
4574        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
4575        data, whereas in other cases it appears the minimum number of times. For        data, whereas in other cases it appears the minimum number of times. For
# Line 4502  for (;; ptr++) Line 4611  for (;; ptr++)
4611              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
4612              goto END_REPEAT;              goto END_REPEAT;
4613              }              }
4614              brazeroptr = previous;    /* Save for possessive optimizing */
4615            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4616            }            }
4617    
# Line 4666  for (;; ptr++) Line 4776  for (;; ptr++)
4776            }            }
4777          }          }
4778    
4779        /* 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
4780        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
4781        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
4782        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
4783          deal with possessive ONCEs specially.
4784    
4785          Otherwise, if the quantifier was possessive, we convert the BRA code to
4786          the POS form, and the KET code to KETRPOS. (It turns out to be convenient
4787          at runtime to detect this kind of subpattern at both the start and at the
4788          end.) The use of special opcodes makes it possible to reduce greatly the
4789          stack usage in pcre_exec(). If the group is preceded by OP_BRAZERO,
4790          convert this to OP_BRAPOSZERO. Then cancel the possessive flag so that
4791          the default action below, of wrapping everything inside atomic brackets,
4792          does not happen.
4793    
4794        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
4795        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
4796        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
4797        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
4798        atomic groups at runtime, but in a different way.] */        runtime, but in a different way.] */
4799    
4800        else        else
4801          {          {
4802          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
4803          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
4804          *ketcode = OP_KETRMAX + repeat_type;  
4805          if (lengthptr == NULL && *bracode != OP_ONCE)          if (*bracode == OP_ONCE && possessive_quantifier) *bracode = OP_BRA;
4806            if (*bracode == OP_ONCE)
4807              *ketcode = OP_KETRMAX + repeat_type;
4808            else
4809            {            {
4810            uschar *scode = bracode;            if (possessive_quantifier)
4811            do              {
4812                *bracode += 1;                   /* Switch to xxxPOS opcodes */
4813                *ketcode = OP_KETRPOS;
4814                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
4815                possessive_quantifier = FALSE;
4816                }
4817              else *ketcode = OP_KETRMAX + repeat_type;
4818    
4819              if (lengthptr == NULL)
4820              {              {
4821              if (could_be_empty_branch(scode, ketcode, utf8, cd))              uschar *scode = bracode;
4822                do
4823                {                {
4824                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
4825                break;                  {
4826                    *bracode += OP_SBRA - OP_BRA;
4827                    break;
4828                    }
4829                  scode += GET(scode, 1);
4830                }                }
4831              scode += GET(scode, 1);              while (*scode == OP_ALT);
4832              }              }
           while (*scode == OP_ALT);  
4833            }            }
4834          }          }
4835        }        }
# Line 4715  for (;; ptr++) Line 4850  for (;; ptr++)
4850        }        }
4851    
4852      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
4853      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
4854      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
4855      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
4856      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
4857      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
4858      tempcode, not at previous, which might be the first part of a string whose  
4859      (former) last char we repeated.      Possessively repeated subpatterns have already been handled in the code
4860        just above, so possessive_quantifier is always FALSE for them at this
4861        stage.
4862    
4863        Note that the repeated item starts at tempcode, not at previous, which
4864        might be the first part of a string whose (former) last char we repeated.
4865    
4866      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
4867      an 'upto' may follow. We skip over an 'exact' item, and then test the      an 'upto' may follow. We skip over an 'exact' item, and then test the
# Line 4752  for (;; ptr++) Line 4892  for (;; ptr++)
4892          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
4893          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4894    
4895          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
4896          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
4897          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
4898          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
4899    
4900          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4901          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4902          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4903          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4904    
4905            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
4906            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
4907            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
4908            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
4909    
4910            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
4911            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
4912            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
4913            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
4914    
4915          /* Because we are moving code along, we must ensure that any          /* Because we are moving code along, we must ensure that any
4916          pending recursive references are updated. */          pending recursive references are updated. */
4917    
# Line 4836  for (;; ptr++) Line 4986  for (;; ptr++)
4986          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4987              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4988            {            {
4989            /* Check for open captures before ACCEPT */            /* Check for open captures before ACCEPT and convert it to
4990              ASSERT_ACCEPT if in an assertion. */
4991    
4992            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
4993              {              {
4994              open_capitem *oc;              open_capitem *oc;
4995                if (arglen != 0)
4996                  {
4997                  *errorcodeptr = ERR59;
4998                  goto FAILED;
4999                  }
5000              cd->had_accept = TRUE;              cd->had_accept = TRUE;
5001              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5002                {                {
5003                *code++ = OP_CLOSE;                *code++ = OP_CLOSE;
5004                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
5005                }                }
5006                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5007              }              }
5008    
5009            /* Handle the cases with/without an argument */            /* Handle other cases with/without an argument */
5010    
5011            if (arglen == 0)            else if (arglen == 0)
5012              {              {
5013              if (verbs[i].op < 0)   /* Argument is mandatory */              if (verbs[i].op < 0)   /* Argument is mandatory */
5014                {                {
# Line 4862  for (;; ptr++) Line 5019  for (;; ptr++)
5019              if (*code++ == OP_THEN)              if (*code++ == OP_THEN)
5020                {                {
5021                PUT(code, 0, code - bcptr->current_branch - 1);                PUT(code, 0, code - bcptr->current_branch - 1);
5022                code += LINK_SIZE;                code += LINK_SIZE;
5023                }                }
5024              }              }
5025    
5026            else            else
# Line 4877  for (;; ptr++) Line 5034  for (;; ptr++)
5034              if (*code++ == OP_THEN_ARG)              if (*code++ == OP_THEN_ARG)
5035                {                {
5036                PUT(code, 0, code - bcptr->current_branch - 1);                PUT(code, 0, code - bcptr->current_branch - 1);
5037                code += LINK_SIZE;                code += LINK_SIZE;
5038                }                }
5039              *code++ = arglen;              *code++ = arglen;
5040              memcpy(code, arg, arglen);              memcpy(code, arg, arglen);
5041              code += arglen;              code += arglen;
# Line 5140  for (;; ptr++) Line 5297  for (;; ptr++)
5297          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5298          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5299          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5300            cd->assert_depth += 1;
5301          ptr++;          ptr++;
5302          break;          break;
5303    
# Line 5154  for (;; ptr++) Line 5312  for (;; ptr++)
5312            continue;            continue;
5313            }            }
5314          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5315            cd->assert_depth += 1;
5316          break;          break;
5317    
5318    
# Line 5163  for (;; ptr++) Line 5322  for (;; ptr++)
5322            {            {
5323            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5324            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5325              cd->assert_depth += 1;
5326            ptr += 2;            ptr += 2;
5327            break;            break;
5328    
5329            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5330            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5331              cd->assert_depth += 1;
5332            ptr += 2;            ptr += 2;
5333            break;            break;
5334    
# Line 5373  for (;; ptr++) Line 5534  for (;; ptr++)
5534          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5535          namelen = (int)(ptr - name);          namelen = (int)(ptr - name);
5536    
5537          /* 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
5538          reference number. */          a dummy reference number, because it was not used in the first pass.
5539            However, with the change of recursive back references to be atomic,
5540            we have to look for the number so that this state can be identified, as
5541            otherwise the incorrect length is computed. If it's not a backwards
5542            reference, the dummy number will do. */
5543    
5544          if (lengthptr != NULL)          if (lengthptr != NULL)
5545            {            {
5546              const uschar *temp;
5547    
5548            if (namelen == 0)            if (namelen == 0)
5549              {              {
5550              *errorcodeptr = ERR62;              *errorcodeptr = ERR62;
# Line 5393  for (;; ptr++) Line 5560  for (;; ptr++)
5560              *errorcodeptr = ERR48;              *errorcodeptr = ERR48;
5561              goto FAILED;              goto FAILED;
5562              }              }
5563            recno = 0;  
5564              /* The name table does not exist in the first pass, so we cannot
5565              do a simple search as in the code below. Instead, we have to scan the
5566              pattern to find the number. It is important that we scan it only as
5567              far as we have got because the syntax of named subpatterns has not
5568              been checked for the rest of the pattern, and find_parens() assumes
5569              correct syntax. In any case, it's a waste of resources to scan
5570              further. We stop the scan at the current point by temporarily
5571              adjusting the value of cd->endpattern. */
5572    
5573              temp = cd->end_pattern;
5574              cd->end_pattern = ptr;
5575              recno = find_parens(cd, name, namelen,
5576                (options & PCRE_EXTENDED) != 0, utf8);
5577              cd->end_pattern = temp;
5578              if (recno < 0) recno = 0;    /* Forward ref; set dummy number */
5579            }            }
5580    
5581          /* In the real compile, seek the name in the table. We check the name          /* In the real compile, seek the name in the table. We check the name
# Line 5537  for (;; ptr++) Line 5719  for (;; ptr++)
5719    
5720                /* 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
5721                offset below, what it actually inserted is the reference number                offset below, what it actually inserted is the reference number
5722                of the group. */                of the group. Then remember the forward reference. */
5723    
5724                called = cd->start_code + recno;                called = cd->start_code + recno;
5725                PUTINC(cd->hwm, 0, (int)(code + 2 + LINK_SIZE - cd->start_code));                PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
5726                }                }
5727    
5728              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
# Line 5555  for (;; ptr++) Line 5737  for (;; ptr++)
5737                }                }
5738              }              }
5739    
5740            /* Insert the recursion/subroutine item, automatically wrapped inside            /* Insert the recursion/subroutine item. */
5741            "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;  
   
5742            *code = OP_RECURSE;            *code = OP_RECURSE;
5743            PUT(code, 1, (int)(called - cd->start_code));            PUT(code, 1, (int)(called - cd->start_code));
5744            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;  
5745            }            }
5746    
5747          /* Can't determine a first byte now */          /* Can't determine a first byte now */
# Line 5632  for (;; ptr++) Line 5802  for (;; ptr++)
5802          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
5803          both phases.          both phases.
5804    
5805          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
5806          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. */  
5807    
5808          if (*ptr == CHAR_RIGHT_PARENTHESIS)          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5809            {            {
# Line 5645  for (;; ptr++) Line 5814  for (;; ptr++)
5814              }              }
5815            else            else
5816              {              {
             if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))  
               {  
               *code++ = OP_OPT;  
               *code++ = newoptions & PCRE_IMS;  
               }  
5817              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5818              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5819              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5820              }              }
5821    
5822            /* Change options at this level, and pass them back for use            /* Change options at this level, and pass them back for use
5823            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). */  
5824    
5825            *optionsptr = options = newoptions;            *optionsptr = options = newoptions;
5826            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
# Line 5696  for (;; ptr++) Line 5858  for (;; ptr++)
5858    
5859      /* Process nested bracketed regex. Assertions may not be repeated, but      /* Process nested bracketed regex. Assertions may not be repeated, but
5860      other kinds can be. All their opcodes are >= OP_ONCE. We copy code into a      other kinds can be. All their opcodes are >= OP_ONCE. We copy code into a
5861      non-register variable in order to be able to pass its address because some      non-register variable (tempcode) in order to be able to pass its address
5862      compilers complain otherwise. Pass in a new setting for the ims options if      because some compilers complain otherwise. */
     they have changed. */  
5863    
5864      previous = (bravalue >= OP_ONCE)? code : NULL;      previous = (bravalue >= OP_ONCE)? code : NULL;
5865      *code = bravalue;      *code = bravalue;
# Line 5708  for (;; ptr++) Line 5869  for (;; ptr++)
5869    
5870      if (!compile_regex(      if (!compile_regex(
5871           newoptions,                   /* The complete new option state */           newoptions,                   /* The complete new option state */
          options & PCRE_IMS,           /* The previous ims option state */  
5872           &tempcode,                    /* Where to put code (updated) */           &tempcode,                    /* Where to put code (updated) */
5873           &ptr,                         /* Input pointer (updated) */           &ptr,                         /* Input pointer (updated) */
5874           errorcodeptr,                 /* Where to put an error message */           errorcodeptr,                 /* Where to put an error message */
# Line 5724  for (;; ptr++) Line 5884  for (;; ptr++)
5884             &length_prevgroup           /* Pre-compile phase */             &length_prevgroup           /* Pre-compile phase */
5885           ))           ))
5886        goto FAILED;        goto FAILED;
5887    
5888        if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
5889          cd->assert_depth -= 1;
5890    
5891      /* 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
5892      group, while tempcode has been updated to point past the end of the group      group, while tempcode has been updated to point past the end of the group
# Line 5795  for (;; ptr++) Line 5958  for (;; ptr++)
5958          goto FAILED;          goto FAILED;
5959          }          }
5960        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
5961        *code++ = OP_BRA;        code++;   /* This already contains bravalue */
5962        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5963        *code++ = OP_KET;        *code++ = OP_KET;
5964        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
# Line 5963  for (;; ptr++) Line 6126  for (;; ptr++)
6126          }          }
6127    
6128        /* \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).
6129        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax).  */
6130    
6131        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||        if (-c == ESC_k)
           ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))  
6132          {          {
6133            if ((ptr[1] != CHAR_LESS_THAN_SIGN &&
6134              ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))
6135              {
6136              *errorcodeptr = ERR69;
6137              break;
6138              }
6139          is_recurse = FALSE;          is_recurse = FALSE;
6140          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
6141            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
6142            CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;            CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
6143          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
6144          }          }
6145    
6146        /* Back references are handled specially; must disable firstbyte if        /* Back references are handled specially; must disable firstbyte if
6147        not set to cope with cases like (?=(\w+))\1: which would otherwise set        not set to cope with cases like (?=(\w+))\1: which would otherwise set
# Line 5987  for (;; ptr++) Line 6155  for (;; ptr++)
6155          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
6156          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
6157          previous = code;          previous = code;
6158          *code++ = OP_REF;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_REFI : OP_REF;
6159          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
6160          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
6161          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
# Line 6095  for (;; ptr++) Line 6263  for (;; ptr++)
6263    
6264      ONE_CHAR:      ONE_CHAR:
6265      previous = code;      previous = code;
6266      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARI : OP_CHAR;
6267      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
6268    
6269      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
# Line 6159  return FALSE; Line 6327  return FALSE;
6327  /* 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
6328  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
6329  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.  
   
6330  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
6331  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
6332  value of lengthptr distinguishes the two phases.  value of lengthptr distinguishes the two phases.
6333    
6334  Arguments:  Arguments:
6335    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  
6336    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
6337    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
6338    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
# Line 6188  Returns:         TRUE on success Line 6350  Returns:         TRUE on success
6350  */  */
6351    
6352  static BOOL  static BOOL
6353  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,
6354    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
6355    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,
6356    int *lengthptr)    int *lengthptr)
# Line 6205  int branchfirstbyte, branchreqbyte; Line 6367  int branchfirstbyte, branchreqbyte;
6367  int length;  int length;
6368  int orig_bracount;  int orig_bracount;
6369  int max_bracount;  int max_bracount;
 int old_external_options = cd->external_options;  
6370  branch_chain bc;  branch_chain bc;
6371    
6372  bc.outer = bcptr;  bc.outer = bcptr;
# Line 6229  pre-compile phase to find out whether an Line 6390  pre-compile phase to find out whether an
6390    
6391  /* 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
6392  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
6393  detect groups that contain recursive back references to themselves. */  detect groups that contain recursive back references to themselves. Note that
6394    only OP_CBRA need be tested here; changing this opcode to one of its variants,
6395    e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
6396    
6397  if (*code == OP_CBRA)  if (*code == OP_CBRA)
6398    {    {
# Line 6255  for (;;) Line 6418  for (;;)
6418    
6419    if (reset_bracount) cd->bracount = orig_bracount;    if (reset_bracount) cd->bracount = orig_bracount;
6420    
   /* 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;  
     }  
   
6421    /* Set up dummy OP_REVERSE if lookbehind assertion */    /* Set up dummy OP_REVERSE if lookbehind assertion */
6422    
6423    if (lookbehind)    if (lookbehind)
# Line 6284  for (;;) Line 6438  for (;;)
6438      return FALSE;      return FALSE;
6439      }      }
6440    
   /* 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;  
   
6441    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6442    has fewer than the rest. */    has fewer than the rest. */
6443    
# Line 6353  for (;;) Line 6498  for (;;)
6498        {        {
6499        int fixed_length;        int fixed_length;
6500        *code = OP_END;        *code = OP_END;
6501        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);        fixed_length = find_fixedlength(last_branch,  (options & PCRE_UTF8) != 0,
6502            FALSE, cd);
6503        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6504        if (fixed_length == -3)        if (fixed_length == -3)
6505          {          {
# Line 6374  for (;;) Line 6520  for (;;)
6520    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
6521    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
6522    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
6523    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. */  
6524    
6525    if (*ptr != CHAR_VERTICAL_LINE)    if (*ptr != CHAR_VERTICAL_LINE)
6526      {      {
# Line 6420  for (;;) Line 6564  for (;;)
6564        cd->open_caps = cd->open_caps->next;        cd->open_caps = cd->open_caps->next;
6565        }        }
6566    
     /* Reset options if needed. */  
   
     if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)  
       {  
       *code++ = OP_OPT;  
       *code++ = oldims;  
       length += 2;  
       }  
   
6567      /* Retain the highest bracket number, in case resetting was used. */      /* Retain the highest bracket number, in case resetting was used. */
6568    
6569      cd->bracount = max_bracount;      cd->bracount = max_bracount;
# Line 6488  for (;;) Line 6623  for (;;)
6623  /* 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
6624  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
6625  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
6626  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
6627  counts, since OP_CIRC can match in the middle.  be found, because ^ generates OP_CIRCM in that mode.
6628    
6629  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.
6630  This is the code for \G, which means "match at start of match position, taking  This is the code for \G, which means "match at start of match position, taking
# Line 6510  of the more common cases more precisely. Line 6645  of the more common cases more precisely.
6645    
6646  Arguments:  Arguments:
6647    code           points to start of expression (the bracket)    code           points to start of expression (the bracket)
   options        points to the options setting  
6648    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
6649                    handles up to substring 31; after that we just have to take                    handles up to substring 31; after that we just have to take
6650                    the less precise approach                    the less precise approach
# Line 6520  Returns:     TRUE or FALSE Line 6654  Returns:     TRUE or FALSE
6654  */  */
6655    
6656  static BOOL  static BOOL
6657  is_anchored(register const uschar *code, int *options, unsigned int bracket_map,  is_anchored(register const uschar *code, unsigned int bracket_map,
6658    unsigned int backref_map)    unsigned int backref_map)
6659  {  {
6660  do {  do {
6661     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6662       options, PCRE_MULTILINE, FALSE);       FALSE);
6663     register int op = *scode;     register int op = *scode;
6664    
6665     /* Non-capturing brackets */     /* Non-capturing brackets */
6666    
6667     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6668           op == OP_SBRA || op == OP_SBRAPOS)
6669       {       {
6670       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6671       }       }
6672    
6673     /* Capturing brackets */     /* Capturing brackets */
6674    
6675     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6676                op == OP_SCBRA || op == OP_SCBRAPOS)
6677       {       {
6678       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6679       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
6680       if (!is_anchored(scode, options, new_map, backref_map)) return FALSE;       if (!is_anchored(scode, new_map, backref_map)) return FALSE;
6681       }       }
6682    
6683     /* Other brackets */     /* Other brackets */
6684    
6685     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)
6686       {       {
6687       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6688       }       }
6689    
6690     /* .* is not anchored unless DOTALL is set (which generates OP_ALLANY) and     /* .* is not anchored unless DOTALL is set (which generates OP_ALLANY) and
# Line 6563  do { Line 6699  do {
6699    
6700     /* Check for explicit anchoring */     /* Check for explicit anchoring */
6701    
6702     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;  
6703     code += GET(code, 1);     code += GET(code, 1);
6704     }     }
6705  while (*code == OP_ALT);   /* Loop for each alternative */  while (*code == OP_ALT);   /* Loop for each alternative */
# Line 6601  is_startline(const uschar *code, unsigne Line 6735  is_startline(const uschar *code, unsigne
6735  {  {
6736  do {  do {
6737     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6738       NULL, 0, FALSE);       FALSE);
6739     register int op = *scode;     register int op = *scode;
6740    
6741     /* If we are at the start of a conditional assertion group, *both* the     /* If we are at the start of a conditional assertion group, *both* the
# Line 6628  do { Line 6762  do {
6762         scode += 1 + LINK_SIZE;         scode += 1 + LINK_SIZE;
6763         break;         break;
6764         }         }
6765       scode = first_significant_code(scode, NULL, 0, FALSE);       scode = first_significant_code(scode, FALSE);
6766       op = *scode;       op = *scode;
6767       }       }
6768    
6769     /* Non-capturing brackets */     /* Non-capturing brackets */
6770    
6771     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6772           op == OP_SBRA || op == OP_SBRAPOS)
6773       {       {
6774       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6775       }       }
6776    
6777     /* Capturing brackets */     /* Capturing brackets */
6778    
6779     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6780                op == OP_SCBRA || op == OP_SCBRAPOS)
6781       {       {
6782       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6783       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
# Line 6665  do { Line 6801  do {
6801    
6802     /* Check for explicit circumflex */     /* Check for explicit circumflex */
6803    
6804     else if (op != OP_CIRC) return FALSE;     else if (op != OP_CIRC && op != OP_CIRCM) return FALSE;
6805    
6806     /* Move on to the next alternative */     /* Move on to the next alternative */
6807    
# Line 6691  we return that char, otherwise -1. Line 6827  we return that char, otherwise -1.
6827    
6828  Arguments:  Arguments:
6829    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)  
6830    inassert   TRUE if in an assertion    inassert   TRUE if in an assertion
6831    
6832  Returns:     -1 or the fixed first char  Returns:     -1 or the fixed first char
6833  */  */
6834    
6835  static int  static int
6836  find_firstassertedchar(const uschar *code, int *options, BOOL inassert)  find_firstassertedchar(const uschar *code, BOOL inassert)
6837  {  {
6838  register int c = -1;  register int c = -1;
6839  do {  do {
6840     int d;     int d;
6841     const uschar *scode =     int xl = (*code == OP_CBRA || *code == OP_SCBRA ||
6842       first_significant_code(code + 1+LINK_SIZE, options, PCRE_CASELESS, TRUE);               *code == OP_CBRAPOS || *code == OP_SCBRAPOS)? 2:0;
6843       const uschar *scode = first_significant_code(code + 1+LINK_SIZE + xl, TRUE);
6844     register int op = *scode;     register int op = *scode;
6845    
6846     switch(op)     switch(op)
# Line 6713  do { Line 6849  do {
6849       return -1;       return -1;
6850    
6851       case OP_BRA:       case OP_BRA:
6852         case OP_BRAPOS:
6853       case OP_CBRA:       case OP_CBRA:
6854         case OP_SCBRA:
6855         case OP_CBRAPOS:
6856         case OP_SCBRAPOS:
6857       case OP_ASSERT:       case OP_ASSERT:
6858       case OP_ONCE:       case OP_ONCE:
6859       case OP_COND:       case OP_COND:
6860       if ((d = find_firstassertedchar(scode, options, op == OP_ASSERT)) < 0)       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)
6861         return -1;         return -1;
6862       if (c < 0) c = d; else if (c != d) return -1;       if (c < 0) c = d; else if (c != d) return -1;
6863       break;       break;
6864    
6865       case OP_EXACT:       /* Fall through */       case OP_EXACT:
6866       scode += 2;       scode += 2;
6867         /* Fall through */
6868    
6869       case OP_CHAR:       case OP_CHAR:
      case OP_CHARNC:  
6870       case OP_PLUS:       case OP_PLUS:
6871       case OP_MINPLUS:       case OP_MINPLUS:
6872       case OP_POSPLUS:       case OP_POSPLUS:
6873       if (!inassert) return -1;       if (!inassert) return -1;
6874       if (c < 0)       if (c < 0) c = scode[1];
6875         {         else if (c != scode[1]) return -1;
6876         c = scode[1];       break;
6877         if ((*options & PCRE_CASELESS) != 0) c |= REQ_CASELESS;  
6878         }       case OP_EXACTI:
6879       else if (c != scode[1]) return -1;       scode += 2;
6880         /* Fall through */
6881    
6882         case OP_CHARI:
6883         case OP_PLUSI:
6884         case OP_MINPLUSI:
6885         case OP_POSPLUSI:
6886         if (!inassert) return -1;
6887         if (c < 0) c = scode[1] | REQ_CASELESS;
6888           else if (c != scode[1]) return -1;
6889       break;       break;
6890       }       }
6891    
# Line 6859  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7008  while (ptr[skipatstart] == CHAR_LEFT_PAR
7008      { skipatstart += 7; options |= PCRE_UTF8; continue; }      { skipatstart += 7; options |= PCRE_UTF8; continue; }
7009    else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)
7010      { skipatstart += 6; options |= PCRE_UCP; continue; }      { skipatstart += 6; options |= PCRE_UCP; continue; }
7011      else if (strncmp((char *)(ptr+skipatstart+2), STRING_NO_START_OPT_RIGHTPAR, 13) == 0)
7012        { skipatstart += 15; options |= PCRE_NO_START_OPTIMIZE; continue; }
7013    
7014    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
7015      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
# Line 6885  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7036  while (ptr[skipatstart] == CHAR_LEFT_PAR
7036    
7037  utf8 = (options & PCRE_UTF8) != 0;  utf8 = (options & PCRE_UTF8) != 0;
7038    
7039  /* 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
7040    return of an error code from _pcre_valid_utf8() is a new feature, introduced in
7041    release 8.13. It is passed back from pcre_[dfa_]exec(), but at the moment is
7042    not used here. */
7043    
7044  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
7045  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
7046       (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)       (errorcode = _pcre_valid_utf8((USPTR)pattern, -1, erroroffset)) != 0)
7047    {    {
7048    errorcode = ERR44;    errorcode = ERR44;
7049    goto PCRE_EARLY_ERROR_RETURN2;    goto PCRE_EARLY_ERROR_RETURN2;
# Line 7005  outside can help speed up starting point Line 7159  outside can help speed up starting point
7159  ptr += skipatstart;  ptr += skipatstart;
7160  code = cworkspace;  code = cworkspace;
7161  *code = OP_BRA;  *code = OP_BRA;
7162  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
7163    &code, &ptr, &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd,    FALSE, 0, &firstbyte, &reqbyte, NULL, cd, &length);
   &length);  
7164  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
7165    
7166  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,
# Line 7061  field; this time it's used for rememberi Line 7214  field; this time it's used for rememberi
7214  */  */
7215    
7216  cd->final_bracount = cd->bracount;  /* Save for checking forward references */  cd->final_bracount = cd->bracount;  /* Save for checking forward references */
7217    cd->assert_depth = 0;
7218  cd->bracount = 0;  cd->bracount = 0;
7219  cd->names_found = 0;  cd->names_found = 0;
7220  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
# Line 7079  of the function here. */ Line 7233  of the function here. */
7233  ptr = (const uschar *)pattern + skipatstart;  ptr = (const uschar *)pattern + skipatstart;
7234  code = (uschar *)codestart;  code = (uschar *)codestart;
7235  *code = OP_BRA;  *code = OP_BRA;
7236  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,  (void)compile_regex(re->options, &code, &ptr, &errorcode, FALSE, FALSE, 0,
7237    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);    &firstbyte, &reqbyte, NULL, cd, NULL);
7238  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
7239  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
7240  re->flags = cd->external_flags;  re->flags = cd->external_flags;
# Line 7146  if (cd->check_lookbehind) Line 7300  if (cd->check_lookbehind)
7300        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
7301        int end_op = *be;        int end_op = *be;
7302        *be = OP_END;        *be = OP_END;
7303        fixed_length = find_fixedlength(cc, re->options, TRUE, cd);        fixed_length = find_fixedlength(cc, (re->options & PCRE_UTF8) != 0, TRUE,
7304            cd);
7305        *be = end_op;        *be = end_op;
7306        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
7307        if (fixed_length < 0)        if (fixed_length < 0)
# Line 7185  start with ^. and also when all branches Line 7340  start with ^. and also when all branches
7340    
7341  if ((re->options & PCRE_ANCHORED) == 0)  if ((re->options & PCRE_ANCHORED) == 0)
7342    {    {
7343    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))  
7344      re->options |= PCRE_ANCHORED;      re->options |= PCRE_ANCHORED;
7345    else    else
7346      {      {
7347      if (firstbyte < 0)      if (firstbyte < 0)
7348        firstbyte = find_firstassertedchar(codestart, &temp_options, FALSE);        firstbyte = find_firstassertedchar(codestart, FALSE);
7349      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */
7350        {        {
7351        int ch = firstbyte & 255;        int ch = firstbyte & 255;

Legend:
Removed from v.574  
changed lines
  Added in v.629

  ViewVC Help
Powered by ViewVC 1.1.5