/[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 602 by ph10, Wed May 25 08:29:03 2011 UTC revision 635 by ph10, Sat Jul 23 16:19:50 2011 UTC
# Line 409  static const char error_texts[] = Line 409  static const char error_texts[] =
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 1403  does not. Line 1404  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    {    {
# Line 1468  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 1479  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 1496  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 1509  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 1532  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 1575  for (;;) Line 1578  for (;;)
1578      case OP_CHAR:      case OP_CHAR:
1579      case OP_CHARI:      case OP_CHARI:
1580      case OP_NOT:      case OP_NOT:
1581      case OP_NOTI:      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 1591  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 1694  _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 1712  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 1954  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 1972  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 2000  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 2135  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 2200  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 2682  if (next >= 0) switch(op_code) Line 2720  if (next >= 0) switch(op_code)
2720    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2721    
2722    /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These    /* 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    opcodes are not used for multi-byte characters, because they are coded using
2724    an XCLASS instead. */    an XCLASS instead. */
2725    
2726    case OP_NOT:    case OP_NOT:
2727    return (c = *previous) == next;    return (c = *previous) == next;
2728    
2729    case OP_NOTI:    case OP_NOTI:
2730    if ((c = *previous) == next) return TRUE;    if ((c = *previous) == next) return TRUE;
2731  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2732    if (utf8)    if (utf8)
# Line 3001  int greedy_default, greedy_non_default; Line 3039  int greedy_default, greedy_non_default;
3039  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3040  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3041  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3042  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3043  int after_manual_callout = 0;  int after_manual_callout = 0;
3044  int length_prevgroup = 0;  int length_prevgroup = 0;
3045  register int c;  register int c;
# Line 3019  uschar *previous_callout = NULL; Line 3057  uschar *previous_callout = NULL;
3057  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3058  uschar classbits[32];  uschar classbits[32];
3059    
3060    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3061    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3062    dynamically as we process the pattern. */
3063    
3064  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3065  BOOL class_utf8;  BOOL class_utf8;
3066  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
# Line 3199  for (;; ptr++) Line 3241  for (;; ptr++)
3241      previous_callout = NULL;      previous_callout = NULL;
3242      }      }
3243    
3244    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3245    
3246    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3247      {      {
# Line 4191  for (;; ptr++) Line 4233  for (;; ptr++)
4233        ptr++;        ptr++;
4234        }        }
4235      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4236    
4237        /* If previous was a recursion call, wrap it in atomic brackets so that
4238        previous becomes the atomic group. All recursions were so wrapped in the
4239        past, but it no longer happens for non-repeated recursions. In fact, the
4240        repeated ones could be re-implemented independently so as not to need this,
4241        but for the moment we rely on the code for repeating groups. */
4242    
4243        if (*previous == OP_RECURSE)
4244          {
4245          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4246          *previous = OP_ONCE;
4247          PUT(previous, 1, 2 + 2*LINK_SIZE);
4248          previous[2 + 2*LINK_SIZE] = OP_KET;
4249          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4250          code += 2 + 2 * LINK_SIZE;
4251          length_prevgroup = 3 + 3*LINK_SIZE;
4252    
4253          /* When actually compiling, we need to check whether this was a forward
4254          reference, and if so, adjust the offset. */
4255    
4256          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4257            {
4258            int offset = GET(cd->hwm, -LINK_SIZE);
4259            if (offset == previous + 1 - cd->start_code)
4260              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4261            }
4262          }
4263    
4264        /* Now handle repetition for the different types of item. */
4265    
4266      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4267      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 4201  for (;; ptr++) Line 4272  for (;; ptr++)
4272      if (*previous == OP_CHAR || *previous == OP_CHARI)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4273        {        {
4274        op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;        op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4275    
4276        /* 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
4277        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
4278        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 4246  for (;; ptr++) Line 4317  for (;; ptr++)
4317      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4318      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-
4319      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4320      repeat_type. We can also test for auto-possessification. OP_NOT and OP_NOTI      repeat_type. We can also test for auto-possessification. OP_NOT and OP_NOTI
4321      are currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4322    
4323      else if (*previous == OP_NOT || *previous == OP_NOTI)      else if (*previous == OP_NOT || *previous == OP_NOTI)
# Line 4483  for (;; ptr++) Line 4554  for (;; ptr++)
4554        }        }
4555    
4556      /* 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
4557      cases. */      cases. Note that at this point we can encounter only the "basic" BRA and
4558        KET opcodes, as this is the place where they get converted into the more
4559        special varieties. */
4560    
4561      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous == OP_BRA  || *previous == OP_CBRA ||
4562               *previous == OP_ONCE || *previous == OP_COND)               *previous == OP_ONCE || *previous == OP_COND)
4563        {        {
4564        register int i;        register int i;
       int ketoffset = 0;  
4565        int len = (int)(code - previous);        int len = (int)(code - previous);
4566        uschar *bralink = NULL;        uschar *bralink = NULL;
4567          uschar *brazeroptr = NULL;
4568    
4569        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless */
4570    
4571        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
# Line 4501  for (;; ptr++) Line 4574  for (;; ptr++)
4574          goto FAILED;          goto FAILED;
4575          }          }
4576    
       /* 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. */  
   
       if (repeat_max == -1)  
         {  
         register uschar *ket = previous;  
         do ket += GET(ket, 1); while (*ket != OP_KET);  
         ketoffset = (int)(code - ket);  
         }  
   
4577        /* 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
4578        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
4579        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 4553  for (;; ptr++) Line 4615  for (;; ptr++)
4615              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
4616              goto END_REPEAT;              goto END_REPEAT;
4617              }              }
4618              brazeroptr = previous;    /* Save for possessive optimizing */
4619            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4620            }            }
4621    
# Line 4717  for (;; ptr++) Line 4780  for (;; ptr++)
4780            }            }
4781          }          }
4782    
4783        /* 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
4784        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
4785        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
4786        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
4787          deal with possessive ONCEs specially.
4788    
4789          Otherwise, if the quantifier was possessive, we convert the BRA code to
4790          the POS form, and the KET code to KETRPOS. (It turns out to be convenient
4791          at runtime to detect this kind of subpattern at both the start and at the
4792          end.) The use of special opcodes makes it possible to reduce greatly the
4793          stack usage in pcre_exec(). If the group is preceded by OP_BRAZERO,
4794          convert this to OP_BRAPOSZERO. Then cancel the possessive flag so that
4795          the default action below, of wrapping everything inside atomic brackets,
4796          does not happen.
4797    
4798        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
4799        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
4800        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
4801        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
4802        atomic groups at runtime, but in a different way.] */        runtime, but in a different way.] */
4803    
4804        else        else
4805          {          {
4806          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
4807          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
4808          *ketcode = OP_KETRMAX + repeat_type;  
4809          if (lengthptr == NULL && *bracode != OP_ONCE)          if (*bracode == OP_ONCE && possessive_quantifier) *bracode = OP_BRA;
4810            if (*bracode == OP_ONCE)
4811              *ketcode = OP_KETRMAX + repeat_type;
4812            else
4813            {            {
4814            uschar *scode = bracode;            if (possessive_quantifier)
           do  
4815              {              {
4816              if (could_be_empty_branch(scode, ketcode, utf8, cd))              *bracode += 1;                   /* Switch to xxxPOS opcodes */
4817                *ketcode = OP_KETRPOS;
4818                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
4819                possessive_quantifier = FALSE;
4820                }
4821              else *ketcode = OP_KETRMAX + repeat_type;
4822    
4823              if (lengthptr == NULL)
4824                {
4825                uschar *scode = bracode;
4826                do
4827                {                {
4828                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
4829                break;                  {
4830                    *bracode += OP_SBRA - OP_BRA;
4831                    break;
4832                    }
4833                  scode += GET(scode, 1);
4834                }                }
4835              scode += GET(scode, 1);              while (*scode == OP_ALT);
4836              }              }
           while (*scode == OP_ALT);  
4837            }            }
4838          }          }
4839        }        }
# Line 4766  for (;; ptr++) Line 4854  for (;; ptr++)
4854        }        }
4855    
4856      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
4857      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
4858      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
4859      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
4860      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
4861      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
4862      tempcode, not at previous, which might be the first part of a string whose  
4863      (former) last char we repeated.      Possessively repeated subpatterns have already been handled in the code
4864        just above, so possessive_quantifier is always FALSE for them at this
4865        stage.
4866    
4867        Note that the repeated item starts at tempcode, not at previous, which
4868        might be the first part of a string whose (former) last char we repeated.
4869    
4870      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
4871      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 4897  for (;; ptr++) Line 4990  for (;; ptr++)
4990          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4991              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4992            {            {
4993            /* Check for open captures before ACCEPT */            /* Check for open captures before ACCEPT and convert it to
4994              ASSERT_ACCEPT if in an assertion. */
4995    
4996            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
4997              {              {
4998              open_capitem *oc;              open_capitem *oc;
4999                if (arglen != 0)
5000                  {
5001                  *errorcodeptr = ERR59;
5002                  goto FAILED;
5003                  }
5004              cd->had_accept = TRUE;              cd->had_accept = TRUE;
5005              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5006                {                {
5007                *code++ = OP_CLOSE;                *code++ = OP_CLOSE;
5008                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
5009                }                }
5010                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5011              }              }
5012    
5013            /* Handle the cases with/without an argument */            /* Handle other cases with/without an argument */
5014    
5015            if (arglen == 0)            else if (arglen == 0)
5016              {              {
5017              if (verbs[i].op < 0)   /* Argument is mandatory */              if (verbs[i].op < 0)   /* Argument is mandatory */
5018                {                {
# Line 5201  for (;; ptr++) Line 5301  for (;; ptr++)
5301          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5302          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5303          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5304            cd->assert_depth += 1;
5305          ptr++;          ptr++;
5306          break;          break;
5307    
# Line 5215  for (;; ptr++) Line 5316  for (;; ptr++)
5316            continue;            continue;
5317            }            }
5318          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5319            cd->assert_depth += 1;
5320          break;          break;
5321    
5322    
# Line 5224  for (;; ptr++) Line 5326  for (;; ptr++)
5326            {            {
5327            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5328            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5329              cd->assert_depth += 1;
5330            ptr += 2;            ptr += 2;
5331            break;            break;
5332    
5333            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5334            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5335              cd->assert_depth += 1;
5336            ptr += 2;            ptr += 2;
5337            break;            break;
5338    
# Line 5472  for (;; ptr++) Line 5576  for (;; ptr++)
5576    
5577            temp = cd->end_pattern;            temp = cd->end_pattern;
5578            cd->end_pattern = ptr;            cd->end_pattern = ptr;
5579            recno = find_parens(cd, name, namelen,            recno = find_parens(cd, name, namelen,
5580              (options & PCRE_EXTENDED) != 0, utf8);              (options & PCRE_EXTENDED) != 0, utf8);
5581            cd->end_pattern = temp;            cd->end_pattern = temp;
5582            if (recno < 0) recno = 0;    /* Forward ref; set dummy number */            if (recno < 0) recno = 0;    /* Forward ref; set dummy number */
# Line 5619  for (;; ptr++) Line 5723  for (;; ptr++)
5723    
5724                /* 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
5725                offset below, what it actually inserted is the reference number                offset below, what it actually inserted is the reference number
5726                of the group. */                of the group. Then remember the forward reference. */
5727    
5728                called = cd->start_code + recno;                called = cd->start_code + recno;
5729                PUTINC(cd->hwm, 0, (int)(code + 2 + LINK_SIZE - cd->start_code));                PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
5730                }                }
5731    
5732              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
# Line 5637  for (;; ptr++) Line 5741  for (;; ptr++)
5741                }                }
5742              }              }
5743    
5744            /* Insert the recursion/subroutine item, automatically wrapped inside            /* Insert the recursion/subroutine item. */
5745            "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;  
   
5746            *code = OP_RECURSE;            *code = OP_RECURSE;
5747            PUT(code, 1, (int)(called - cd->start_code));            PUT(code, 1, (int)(called - cd->start_code));
5748            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;  
5749            }            }
5750    
5751          /* Can't determine a first byte now */          /* Can't determine a first byte now */
# Line 5714  for (;; ptr++) Line 5806  for (;; ptr++)
5806          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
5807          both phases.          both phases.
5808    
5809          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
5810          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. */  
5811    
5812          if (*ptr == CHAR_RIGHT_PARENTHESIS)          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5813            {            {
# Line 5733  for (;; ptr++) Line 5824  for (;; ptr++)
5824              }              }
5825    
5826            /* Change options at this level, and pass them back for use            /* Change options at this level, and pass them back for use
5827            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). */  
5828    
5829            *optionsptr = options = newoptions;            *optionsptr = options = newoptions;
5830            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
# Line 5773  for (;; ptr++) Line 5862  for (;; ptr++)
5862    
5863      /* Process nested bracketed regex. Assertions may not be repeated, but      /* Process nested bracketed regex. Assertions may not be repeated, but
5864      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
5865      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
5866      compilers complain otherwise. Pass in a new setting for the ims options if      because some compilers complain otherwise. */
     they have changed. */  
5867    
5868      previous = (bravalue >= OP_ONCE)? code : NULL;      previous = (bravalue >= OP_ONCE)? code : NULL;
5869      *code = bravalue;      *code = bravalue;
# Line 5785  for (;; ptr++) Line 5873  for (;; ptr++)
5873    
5874      if (!compile_regex(      if (!compile_regex(
5875           newoptions,                   /* The complete new option state */           newoptions,                   /* The complete new option state */
          options & PCRE_IMS,           /* The previous ims option state */  
5876           &tempcode,                    /* Where to put code (updated) */           &tempcode,                    /* Where to put code (updated) */
5877           &ptr,                         /* Input pointer (updated) */           &ptr,                         /* Input pointer (updated) */
5878           errorcodeptr,                 /* Where to put an error message */           errorcodeptr,                 /* Where to put an error message */
# Line 5801  for (;; ptr++) Line 5888  for (;; ptr++)
5888             &length_prevgroup           /* Pre-compile phase */             &length_prevgroup           /* Pre-compile phase */
5889           ))           ))
5890        goto FAILED;        goto FAILED;
5891    
5892        if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
5893          cd->assert_depth -= 1;
5894    
5895      /* 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
5896      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 5872  for (;; ptr++) Line 5962  for (;; ptr++)
5962          goto FAILED;          goto FAILED;
5963          }          }
5964        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
5965        *code++ = OP_BRA;        code++;   /* This already contains bravalue */
5966        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5967        *code++ = OP_KET;        *code++ = OP_KET;
5968        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
# Line 6040  for (;; ptr++) Line 6130  for (;; ptr++)
6130          }          }
6131    
6132        /* \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).
6133        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax).  */
6134    
6135        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||        if (-c == ESC_k)
           ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))  
6136          {          {
6137            if ((ptr[1] != CHAR_LESS_THAN_SIGN &&
6138              ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))
6139              {
6140              *errorcodeptr = ERR69;
6141              break;
6142              }
6143          is_recurse = FALSE;          is_recurse = FALSE;
6144          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
6145            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
6146            CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;            CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
6147          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
6148          }          }
6149    
6150        /* Back references are handled specially; must disable firstbyte if        /* Back references are handled specially; must disable firstbyte if
6151        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 6242  value of lengthptr distinguishes the two Line 6337  value of lengthptr distinguishes the two
6337    
6338  Arguments:  Arguments:
6339    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  
6340    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
6341    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
6342    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
# Line 6260  Returns:         TRUE on success Line 6354  Returns:         TRUE on success
6354  */  */
6355    
6356  static BOOL  static BOOL
6357  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,
6358    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
6359    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,
6360    int *lengthptr)    int *lengthptr)
# Line 6277  int branchfirstbyte, branchreqbyte; Line 6371  int branchfirstbyte, branchreqbyte;
6371  int length;  int length;
6372  int orig_bracount;  int orig_bracount;
6373  int max_bracount;  int max_bracount;
 int old_external_options = cd->external_options;  
6374  branch_chain bc;  branch_chain bc;
6375    
6376  bc.outer = bcptr;  bc.outer = bcptr;
# Line 6301  pre-compile phase to find out whether an Line 6394  pre-compile phase to find out whether an
6394    
6395  /* 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
6396  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
6397  detect groups that contain recursive back references to themselves. */  detect groups that contain recursive back references to themselves. Note that
6398    only OP_CBRA need be tested here; changing this opcode to one of its variants,
6399    e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
6400    
6401  if (*code == OP_CBRA)  if (*code == OP_CBRA)
6402    {    {
# Line 6347  for (;;) Line 6442  for (;;)
6442      return FALSE;      return FALSE;
6443      }      }
6444    
   /* 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;  
   
6445    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6446    has fewer than the rest. */    has fewer than the rest. */
6447    
# Line 6416  for (;;) Line 6502  for (;;)
6502        {        {
6503        int fixed_length;        int fixed_length;
6504        *code = OP_END;        *code = OP_END;
6505        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);        fixed_length = find_fixedlength(last_branch,  (options & PCRE_UTF8) != 0,
6506            FALSE, cd);
6507        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6508        if (fixed_length == -3)        if (fixed_length == -3)
6509          {          {
# Line 6437  for (;;) Line 6524  for (;;)
6524    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
6525    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
6526    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
6527    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. */  
6528    
6529    if (*ptr != CHAR_VERTICAL_LINE)    if (*ptr != CHAR_VERTICAL_LINE)
6530      {      {
# Line 6564  of the more common cases more precisely. Line 6649  of the more common cases more precisely.
6649    
6650  Arguments:  Arguments:
6651    code           points to start of expression (the bracket)    code           points to start of expression (the bracket)
   options        points to the options setting  
6652    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
6653                    handles up to substring 31; after that we just have to take                    handles up to substring 31; after that we just have to take
6654                    the less precise approach                    the less precise approach
# Line 6574  Returns:     TRUE or FALSE Line 6658  Returns:     TRUE or FALSE
6658  */  */
6659    
6660  static BOOL  static BOOL
6661  is_anchored(register const uschar *code, int *options, unsigned int bracket_map,  is_anchored(register const uschar *code, unsigned int bracket_map,
6662    unsigned int backref_map)    unsigned int backref_map)
6663  {  {
6664  do {  do {
6665     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6666       options, PCRE_MULTILINE, FALSE);       FALSE);
6667     register int op = *scode;     register int op = *scode;
6668    
6669     /* Non-capturing brackets */     /* Non-capturing brackets */
6670    
6671     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6672           op == OP_SBRA || op == OP_SBRAPOS)
6673       {       {
6674       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6675       }       }
6676    
6677     /* Capturing brackets */     /* Capturing brackets */
6678    
6679     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6680                op == OP_SCBRA || op == OP_SCBRAPOS)
6681       {       {
6682       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6683       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
6684       if (!is_anchored(scode, options, new_map, backref_map)) return FALSE;       if (!is_anchored(scode, new_map, backref_map)) return FALSE;
6685       }       }
6686    
6687     /* Other brackets */     /* Other brackets */
6688    
6689     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)
6690       {       {
6691       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6692       }       }
6693    
6694     /* .* 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 6653  is_startline(const uschar *code, unsigne Line 6739  is_startline(const uschar *code, unsigne
6739  {  {
6740  do {  do {
6741     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6742       NULL, 0, FALSE);       FALSE);
6743     register int op = *scode;     register int op = *scode;
6744    
6745     /* 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 6680  do { Line 6766  do {
6766         scode += 1 + LINK_SIZE;         scode += 1 + LINK_SIZE;
6767         break;         break;
6768         }         }
6769       scode = first_significant_code(scode, NULL, 0, FALSE);       scode = first_significant_code(scode, FALSE);
6770       op = *scode;       op = *scode;
6771       }       }
6772    
6773     /* Non-capturing brackets */     /* Non-capturing brackets */
6774    
6775     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6776           op == OP_SBRA || op == OP_SBRAPOS)
6777       {       {
6778       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6779       }       }
6780    
6781     /* Capturing brackets */     /* Capturing brackets */
6782    
6783     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6784                op == OP_SCBRA || op == OP_SCBRAPOS)
6785       {       {
6786       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6787       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
# Line 6743  we return that char, otherwise -1. Line 6831  we return that char, otherwise -1.
6831    
6832  Arguments:  Arguments:
6833    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)  
6834    inassert   TRUE if in an assertion    inassert   TRUE if in an assertion
6835    
6836  Returns:     -1 or the fixed first char  Returns:     -1 or the fixed first char
6837  */  */
6838    
6839  static int  static int
6840  find_firstassertedchar(const uschar *code, int *options, BOOL inassert)  find_firstassertedchar(const uschar *code, BOOL inassert)
6841  {  {
6842  register int c = -1;  register int c = -1;
6843  do {  do {
6844     int d;     int d;
6845     const uschar *scode =     int xl = (*code == OP_CBRA || *code == OP_SCBRA ||
6846       first_significant_code(code + 1+LINK_SIZE, options, PCRE_CASELESS, TRUE);               *code == OP_CBRAPOS || *code == OP_SCBRAPOS)? 2:0;
6847       const uschar *scode = first_significant_code(code + 1+LINK_SIZE + xl, TRUE);
6848     register int op = *scode;     register int op = *scode;
6849    
6850     switch(op)     switch(op)
# Line 6765  do { Line 6853  do {
6853       return -1;       return -1;
6854    
6855       case OP_BRA:       case OP_BRA:
6856         case OP_BRAPOS:
6857       case OP_CBRA:       case OP_CBRA:
6858         case OP_SCBRA:
6859         case OP_CBRAPOS:
6860         case OP_SCBRAPOS:
6861       case OP_ASSERT:       case OP_ASSERT:
6862       case OP_ONCE:       case OP_ONCE:
6863       case OP_COND:       case OP_COND:
6864       if ((d = find_firstassertedchar(scode, options, op == OP_ASSERT)) < 0)       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)
6865         return -1;         return -1;
6866       if (c < 0) c = d; else if (c != d) return -1;       if (c < 0) c = d; else if (c != d) return -1;
6867       break;       break;
6868    
6869       case OP_EXACT:       /* Fall through */       case OP_EXACT:
6870       scode += 2;       scode += 2;
6871         /* Fall through */
6872    
6873       case OP_CHAR:       case OP_CHAR:
      case OP_CHARI:  
6874       case OP_PLUS:       case OP_PLUS:
6875       case OP_MINPLUS:       case OP_MINPLUS:
6876       case OP_POSPLUS:       case OP_POSPLUS:
6877       if (!inassert) return -1;       if (!inassert) return -1;
6878       if (c < 0)       if (c < 0) c = scode[1];
6879         {         else if (c != scode[1]) return -1;
6880         c = scode[1];       break;
6881         if ((*options & PCRE_CASELESS) != 0) c |= REQ_CASELESS;  
6882         }       case OP_EXACTI:
6883       else if (c != scode[1]) return -1;       scode += 2;
6884         /* Fall through */
6885    
6886         case OP_CHARI:
6887         case OP_PLUSI:
6888         case OP_MINPLUSI:
6889         case OP_POSPLUSI:
6890         if (!inassert) return -1;
6891         if (c < 0) c = scode[1] | REQ_CASELESS;
6892           else if (c != scode[1]) return -1;
6893       break;       break;
6894       }       }
6895    
# Line 6939  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7040  while (ptr[skipatstart] == CHAR_LEFT_PAR
7040    
7041  utf8 = (options & PCRE_UTF8) != 0;  utf8 = (options & PCRE_UTF8) != 0;
7042    
7043  /* Can't support UTF8 unless PCRE has been compiled to include the code. The  /* Can't support UTF8 unless PCRE has been compiled to include the code. The
7044  return of an error code from _pcre_valid_utf8() is a new feature, introduced in  return of an error code from _pcre_valid_utf8() is a new feature, introduced in
7045  release 8.13. The only use we make of it here is to adjust the offset value to  release 8.13. It is passed back from pcre_[dfa_]exec(), but at the moment is
7046  the end of the string for a short string error, for compatibility with previous  not used here. */
 versions. */  
7047    
7048  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
7049  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
7050       (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1, &errorcode)) >= 0)       (errorcode = _pcre_valid_utf8((USPTR)pattern, -1, erroroffset)) != 0)
7051    {    {
7052    errorcode = ERR44;    errorcode = ERR44;
7053    goto PCRE_EARLY_ERROR_RETURN2;    goto PCRE_EARLY_ERROR_RETURN2;
# Line 7063  outside can help speed up starting point Line 7163  outside can help speed up starting point
7163  ptr += skipatstart;  ptr += skipatstart;
7164  code = cworkspace;  code = cworkspace;
7165  *code = OP_BRA;  *code = OP_BRA;
7166  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
7167    &code, &ptr, &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd,    FALSE, 0, &firstbyte, &reqbyte, NULL, cd, &length);
   &length);  
7168  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
7169    
7170  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,
# Line 7119  field; this time it's used for rememberi Line 7218  field; this time it's used for rememberi
7218  */  */
7219    
7220  cd->final_bracount = cd->bracount;  /* Save for checking forward references */  cd->final_bracount = cd->bracount;  /* Save for checking forward references */
7221    cd->assert_depth = 0;
7222  cd->bracount = 0;  cd->bracount = 0;
7223  cd->names_found = 0;  cd->names_found = 0;
7224  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
# Line 7137  of the function here. */ Line 7237  of the function here. */
7237  ptr = (const uschar *)pattern + skipatstart;  ptr = (const uschar *)pattern + skipatstart;
7238  code = (uschar *)codestart;  code = (uschar *)codestart;
7239  *code = OP_BRA;  *code = OP_BRA;
7240  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,  (void)compile_regex(re->options, &code, &ptr, &errorcode, FALSE, FALSE, 0,
7241    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);    &firstbyte, &reqbyte, NULL, cd, NULL);
7242  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
7243  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
7244  re->flags = cd->external_flags;  re->flags = cd->external_flags;
# Line 7204  if (cd->check_lookbehind) Line 7304  if (cd->check_lookbehind)
7304        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
7305        int end_op = *be;        int end_op = *be;
7306        *be = OP_END;        *be = OP_END;
7307        fixed_length = find_fixedlength(cc, re->options, TRUE, cd);        fixed_length = find_fixedlength(cc, (re->options & PCRE_UTF8) != 0, TRUE,
7308            cd);
7309        *be = end_op;        *be = end_op;
7310        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
7311        if (fixed_length < 0)        if (fixed_length < 0)
# Line 7243  start with ^. and also when all branches Line 7344  start with ^. and also when all branches
7344    
7345  if ((re->options & PCRE_ANCHORED) == 0)  if ((re->options & PCRE_ANCHORED) == 0)
7346    {    {
7347    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))  
7348      re->options |= PCRE_ANCHORED;      re->options |= PCRE_ANCHORED;
7349    else    else
7350      {      {
7351      if (firstbyte < 0)      if (firstbyte < 0)
7352        firstbyte = find_firstassertedchar(codestart, &temp_options, FALSE);        firstbyte = find_firstassertedchar(codestart, FALSE);
7353      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */
7354        {        {
7355        int ch = firstbyte & 255;        int ch = firstbyte & 255;

Legend:
Removed from v.602  
changed lines
  Added in v.635

  ViewVC Help
Powered by ViewVC 1.1.5