/[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 621 by ph10, Mon Jul 18 10:14:09 2011 UTC
# Line 545  static const unsigned char ebcdic_charta Line 545  static const unsigned char ebcdic_charta
545  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
546    
547  static BOOL  static BOOL
548    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int *,
549      int *, int *, branch_chain *, compile_data *, int *);      int *, branch_chain *, compile_data *, int *);
550    
551    
552    
# Line 1403  does not. Line 1403  does not.
1403    
1404  Arguments:  Arguments:
1405    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  
1406    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1407    
1408  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1409  */  */
1410    
1411  static const uschar*  static const uschar*
1412  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1413  {  {
1414  for (;;)  for (;;)
1415    {    {
# Line 1468  and doing the check at the end; a flag s Line 1464  and doing the check at the end; a flag s
1464    
1465  Arguments:  Arguments:
1466    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1467    options  the compiling options    utf8     TRUE in UTF-8 mode
1468    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1469    cd       the "compile data" structure    cd       the "compile data" structure
1470    
# Line 1479  Returns:   the fixed length, Line 1475  Returns:   the fixed length,
1475  */  */
1476    
1477  static int  static int
1478  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1479  {  {
1480  int length = -1;  int length = -1;
1481    
# Line 1496  for (;;) Line 1492  for (;;)
1492    register int op = *cc;    register int op = *cc;
1493    switch (op)    switch (op)
1494      {      {
1495        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1496        OP_BRA (normal non-capturing bracket) because the other variants of these
1497        opcodes are all concerned with unlimited repeated groups, which of course
1498        are not of fixed length. They will cause a -1 response from the default
1499        case of this switch. */
1500    
1501      case OP_CBRA:      case OP_CBRA:
1502      case OP_BRA:      case OP_BRA:
1503      case OP_ONCE:      case OP_ONCE:
1504      case OP_COND:      case OP_COND:
1505      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1506      if (d < 0) return d;      if (d < 0) return d;
1507      branchlength += d;      branchlength += d;
1508      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1509  for (;;) Line 1511  for (;;)
1511    
1512      /* 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
1513      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
1514      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.
1515        Note that we must not include the OP_KETRxxx opcodes here, because they
1516        all imply an unlimited repeat. */
1517    
1518      case OP_ALT:      case OP_ALT:
1519      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1520      case OP_END:      case OP_END:
1521      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1522        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
# Line 1532  for (;;) Line 1534  for (;;)
1534      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1535      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1536      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                /* Recursion */
1537      d = find_fixedlength(cs + 2, options, atend, cd);      d = find_fixedlength(cs + 2, utf8, atend, cd);
1538      if (d < 0) return d;      if (d < 0) return d;
1539      branchlength += d;      branchlength += d;
1540      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1575  for (;;) Line 1577  for (;;)
1577      case OP_CHAR:      case OP_CHAR:
1578      case OP_CHARI:      case OP_CHARI:
1579      case OP_NOT:      case OP_NOT:
1580      case OP_NOTI:      case OP_NOTI:
1581      branchlength++;      branchlength++;
1582      cc += 2;      cc += 2;
1583  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1584      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];  
1585  #endif  #endif
1586      break;      break;
1587    
# Line 1591  for (;;) Line 1592  for (;;)
1592      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1593      cc += 4;      cc += 4;
1594  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1595      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];  
1596  #endif  #endif
1597      break;      break;
1598    
# Line 1694  _pcre_find_bracket(const uschar *code, B Line 1694  _pcre_find_bracket(const uschar *code, B
1694  for (;;)  for (;;)
1695    {    {
1696    register int c = *code;    register int c = *code;
1697    
1698    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1699    
1700    /* 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 1713  for (;;)
1713    
1714    /* Handle capturing bracket */    /* Handle capturing bracket */
1715    
1716    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1717               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1718      {      {
1719      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1720      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1954  could_be_empty_branch(const uschar *code Line 1956  could_be_empty_branch(const uschar *code
1956    compile_data *cd)    compile_data *cd)
1957  {  {
1958  register int c;  register int c;
1959  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
1960       code < endcode;       code < endcode;
1961       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
1962    {    {
1963    const uschar *ccode;    const uschar *ccode;
1964    
# Line 1972  for (code = first_significant_code(code Line 1974  for (code = first_significant_code(code
1974      continue;      continue;
1975      }      }
1976    
   /* 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;  
     }  
   
1977    /* For a recursion/subroutine call, if its end has been reached, which    /* For a recursion/subroutine call, if its end has been reached, which
1978    implies a subroutine call, we can scan it. */    implies a subroutine call, we can scan it. */
1979    
# Line 2004  for (code = first_significant_code(code Line 1996  for (code = first_significant_code(code
1996      continue;      continue;
1997      }      }
1998    
1999      /* Groups with zero repeats can of course be empty; skip them. */
2000    
2001      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2002          c == OP_BRAPOSZERO)
2003        {
2004        code += _pcre_OP_lengths[c];
2005        do code += GET(code, 1); while (*code == OP_ALT);
2006        c = *code;
2007        continue;
2008        }
2009    
2010      /* A nested group that is already marked as "could be empty" can just be
2011      skipped. */
2012    
2013      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2014          c == OP_SCBRA || c == OP_SCBRAPOS)
2015        {
2016        do code += GET(code, 1); while (*code == OP_ALT);
2017        c = *code;
2018        continue;
2019        }
2020    
2021    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2022    
2023    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2024          c == OP_CBRA || c == OP_CBRAPOS ||
2025          c == OP_ONCE || c == OP_COND)
2026      {      {
2027      BOOL empty_branch;      BOOL empty_branch;
2028      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 2151  for (code = first_significant_code(code
2151      case OP_KET:      case OP_KET:
2152      case OP_KETRMAX:      case OP_KETRMAX:
2153      case OP_KETRMIN:      case OP_KETRMIN:
2154        case OP_KETRPOS:
2155      case OP_ALT:      case OP_ALT:
2156      return TRUE;      return TRUE;
2157    
# Line 2682  if (next >= 0) switch(op_code) Line 2699  if (next >= 0) switch(op_code)
2699    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2700    
2701    /* 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
2702    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
2703    an XCLASS instead. */    an XCLASS instead. */
2704    
2705    case OP_NOT:    case OP_NOT:
2706    return (c = *previous) == next;    return (c = *previous) == next;
2707    
2708    case OP_NOTI:    case OP_NOTI:
2709    if ((c = *previous) == next) return TRUE;    if ((c = *previous) == next) return TRUE;
2710  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2711    if (utf8)    if (utf8)
# Line 4191  for (;; ptr++) Line 4208  for (;; ptr++)
4208        ptr++;        ptr++;
4209        }        }
4210      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4211    
4212        /* If previous was a recursion call, wrap it in atomic brackets so that
4213        previous becomes the atomic group. All recursions were so wrapped in the
4214        past, but it no longer happens for non-repeated recursions. In fact, the
4215        repeated ones could be re-implemented independently so as not to need this,
4216        but for the moment we rely on the code for repeating groups. */
4217    
4218        if (*previous == OP_RECURSE)
4219          {
4220          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4221          *previous = OP_ONCE;
4222          PUT(previous, 1, 2 + 2*LINK_SIZE);
4223          previous[2 + 2*LINK_SIZE] = OP_KET;
4224          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4225          code += 2 + 2 * LINK_SIZE;
4226          length_prevgroup = 3 + 3*LINK_SIZE;
4227    
4228          /* When actually compiling, we need to check whether this was a forward
4229          reference, and if so, adjust the offset. */
4230    
4231          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4232            {
4233            int offset = GET(cd->hwm, -LINK_SIZE);
4234            if (offset == previous + 1 - cd->start_code)
4235              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4236            }
4237          }
4238    
4239        /* Now handle repetition for the different types of item. */
4240    
4241      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4242      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 4247  for (;; ptr++)
4247      if (*previous == OP_CHAR || *previous == OP_CHARI)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4248        {        {
4249        op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;        op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4250    
4251        /* 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
4252        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
4253        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 4292  for (;; ptr++)
4292      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4293      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-
4294      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4295      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
4296      are currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4297    
4298      else if (*previous == OP_NOT || *previous == OP_NOTI)      else if (*previous == OP_NOT || *previous == OP_NOTI)
# Line 4483  for (;; ptr++) Line 4529  for (;; ptr++)
4529        }        }
4530    
4531      /* 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
4532      cases. */      cases. Note that at this point we can encounter only the "basic" BRA and
4533        KET opcodes, as this is the place where they get converted into the more
4534        special varieties. */
4535    
4536      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous == OP_BRA  || *previous == OP_CBRA ||
4537               *previous == OP_ONCE || *previous == OP_COND)               *previous == OP_ONCE || *previous == OP_COND)
4538        {        {
4539        register int i;        register int i;
       int ketoffset = 0;  
4540        int len = (int)(code - previous);        int len = (int)(code - previous);
4541        uschar *bralink = NULL;        uschar *bralink = NULL;
4542          uschar *brazeroptr = NULL;
4543    
4544        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless */
4545    
4546        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 4549  for (;; ptr++)
4549          goto FAILED;          goto FAILED;
4550          }          }
4551    
       /* 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);  
         }  
   
4552        /* 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
4553        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
4554        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 4590  for (;; ptr++)
4590              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
4591              goto END_REPEAT;              goto END_REPEAT;
4592              }              }
4593              brazeroptr = previous;    /* Save for possessive optimizing */
4594            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4595            }            }
4596    
# Line 4717  for (;; ptr++) Line 4755  for (;; ptr++)
4755            }            }
4756          }          }
4757    
4758        /* 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
4759        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
4760        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
4761        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
4762          deal with possessive ONCEs specially.
4763    
4764          Otherwise, if the quantifier was possessive, we convert the BRA code to
4765          the POS form, and the KET code to KETRPOS. (It turns out to be convenient
4766          at runtime to detect this kind of subpattern at both the start and at the
4767          end.) The use of special opcodes makes it possible to reduce greatly the
4768          stack usage in pcre_exec(). If the group is preceded by OP_BRAZERO,
4769          convert this to OP_BRAPOSZERO. Then cancel the possessive flag so that
4770          the default action below, of wrapping everything inside atomic brackets,
4771          does not happen.
4772    
4773        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
4774        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
4775        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
4776        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
4777        atomic groups at runtime, but in a different way.] */        runtime, but in a different way.] */
4778    
4779        else        else
4780          {          {
4781          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
4782          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
4783          *ketcode = OP_KETRMAX + repeat_type;  
4784          if (lengthptr == NULL && *bracode != OP_ONCE)          if (*bracode == OP_ONCE && possessive_quantifier) *bracode = OP_BRA;
4785            if (*bracode == OP_ONCE)
4786              *ketcode = OP_KETRMAX + repeat_type;
4787            else
4788            {            {
4789            uschar *scode = bracode;            if (possessive_quantifier)
4790            do              {
4791                *bracode += 1;                   /* Switch to xxxPOS opcodes */
4792                *ketcode = OP_KETRPOS;
4793                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
4794                possessive_quantifier = FALSE;
4795                }
4796              else *ketcode = OP_KETRMAX + repeat_type;
4797    
4798              if (lengthptr == NULL)
4799              {              {
4800              if (could_be_empty_branch(scode, ketcode, utf8, cd))              uschar *scode = bracode;
4801                do
4802                {                {
4803                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
4804                break;                  {
4805                    *bracode += OP_SBRA - OP_BRA;
4806                    break;
4807                    }
4808                  scode += GET(scode, 1);
4809                }                }
4810              scode += GET(scode, 1);              while (*scode == OP_ALT);
4811              }              }
           while (*scode == OP_ALT);  
4812            }            }
4813          }          }
4814        }        }
# Line 4766  for (;; ptr++) Line 4829  for (;; ptr++)
4829        }        }
4830    
4831      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
4832      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
4833      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
4834      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
4835      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
4836      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
4837      tempcode, not at previous, which might be the first part of a string whose  
4838      (former) last char we repeated.      Possessively repeated subpatterns have already been handled in the code
4839        just above, so possessive_quantifier is always FALSE for them at this
4840        stage.
4841    
4842        Note that the repeated item starts at tempcode, not at previous, which
4843        might be the first part of a string whose (former) last char we repeated.
4844    
4845      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
4846      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 4965  for (;; ptr++)
4965          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4966              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4967            {            {
4968            /* Check for open captures before ACCEPT */            /* Check for open captures before ACCEPT and convert it to
4969              ASSERT_ACCEPT if in an assertion. */
4970    
4971            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
4972              {              {
4973              open_capitem *oc;              open_capitem *oc;
4974                if (arglen != 0)
4975                  {
4976                  *errorcodeptr = ERR59;
4977                  goto FAILED;
4978                  }
4979              cd->had_accept = TRUE;              cd->had_accept = TRUE;
4980              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4981                {                {
4982                *code++ = OP_CLOSE;                *code++ = OP_CLOSE;
4983                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
4984                }                }
4985                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
4986              }              }
4987    
4988            /* Handle the cases with/without an argument */            /* Handle other cases with/without an argument */
4989    
4990            if (arglen == 0)            else if (arglen == 0)
4991              {              {
4992              if (verbs[i].op < 0)   /* Argument is mandatory */              if (verbs[i].op < 0)   /* Argument is mandatory */
4993                {                {
# Line 5201  for (;; ptr++) Line 5276  for (;; ptr++)
5276          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5277          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5278          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5279            cd->assert_depth += 1;
5280          ptr++;          ptr++;
5281          break;          break;
5282    
# Line 5215  for (;; ptr++) Line 5291  for (;; ptr++)
5291            continue;            continue;
5292            }            }
5293          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5294            cd->assert_depth += 1;
5295          break;          break;
5296    
5297    
# Line 5224  for (;; ptr++) Line 5301  for (;; ptr++)
5301            {            {
5302            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5303            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5304              cd->assert_depth += 1;
5305            ptr += 2;            ptr += 2;
5306            break;            break;
5307    
5308            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5309            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5310              cd->assert_depth += 1;
5311            ptr += 2;            ptr += 2;
5312            break;            break;
5313    
# Line 5619  for (;; ptr++) Line 5698  for (;; ptr++)
5698    
5699                /* 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
5700                offset below, what it actually inserted is the reference number                offset below, what it actually inserted is the reference number
5701                of the group. */                of the group. Then remember the forward reference. */
5702    
5703                called = cd->start_code + recno;                called = cd->start_code + recno;
5704                PUTINC(cd->hwm, 0, (int)(code + 2 + LINK_SIZE - cd->start_code));                PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
5705                }                }
5706    
5707              /* 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 5716  for (;; ptr++)
5716                }                }
5717              }              }
5718    
5719            /* Insert the recursion/subroutine item, automatically wrapped inside            /* Insert the recursion/subroutine item. */
5720            "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;  
   
5721            *code = OP_RECURSE;            *code = OP_RECURSE;
5722            PUT(code, 1, (int)(called - cd->start_code));            PUT(code, 1, (int)(called - cd->start_code));
5723            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;  
5724            }            }
5725    
5726          /* Can't determine a first byte now */          /* Can't determine a first byte now */
# Line 5714  for (;; ptr++) Line 5781  for (;; ptr++)
5781          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
5782          both phases.          both phases.
5783    
5784          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
5785          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. */  
5786    
5787          if (*ptr == CHAR_RIGHT_PARENTHESIS)          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5788            {            {
# Line 5733  for (;; ptr++) Line 5799  for (;; ptr++)
5799              }              }
5800    
5801            /* Change options at this level, and pass them back for use            /* Change options at this level, and pass them back for use
5802            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). */  
5803    
5804            *optionsptr = options = newoptions;            *optionsptr = options = newoptions;
5805            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
# Line 5773  for (;; ptr++) Line 5837  for (;; ptr++)
5837    
5838      /* Process nested bracketed regex. Assertions may not be repeated, but      /* Process nested bracketed regex. Assertions may not be repeated, but
5839      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
5840      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
5841      compilers complain otherwise. Pass in a new setting for the ims options if      because some compilers complain otherwise. */
     they have changed. */  
5842    
5843      previous = (bravalue >= OP_ONCE)? code : NULL;      previous = (bravalue >= OP_ONCE)? code : NULL;
5844      *code = bravalue;      *code = bravalue;
# Line 5785  for (;; ptr++) Line 5848  for (;; ptr++)
5848    
5849      if (!compile_regex(      if (!compile_regex(
5850           newoptions,                   /* The complete new option state */           newoptions,                   /* The complete new option state */
          options & PCRE_IMS,           /* The previous ims option state */  
5851           &tempcode,                    /* Where to put code (updated) */           &tempcode,                    /* Where to put code (updated) */
5852           &ptr,                         /* Input pointer (updated) */           &ptr,                         /* Input pointer (updated) */
5853           errorcodeptr,                 /* Where to put an error message */           errorcodeptr,                 /* Where to put an error message */
# Line 5801  for (;; ptr++) Line 5863  for (;; ptr++)
5863             &length_prevgroup           /* Pre-compile phase */             &length_prevgroup           /* Pre-compile phase */
5864           ))           ))
5865        goto FAILED;        goto FAILED;
5866    
5867        if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
5868          cd->assert_depth -= 1;
5869    
5870      /* 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
5871      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 5937  for (;; ptr++)
5937          goto FAILED;          goto FAILED;
5938          }          }
5939        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
5940        *code++ = OP_BRA;        code++;   /* This already contains bravalue */
5941        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5942        *code++ = OP_KET;        *code++ = OP_KET;
5943        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
# Line 6242  value of lengthptr distinguishes the two Line 6307  value of lengthptr distinguishes the two
6307    
6308  Arguments:  Arguments:
6309    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  
6310    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
6311    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
6312    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
# Line 6260  Returns:         TRUE on success Line 6324  Returns:         TRUE on success
6324  */  */
6325    
6326  static BOOL  static BOOL
6327  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,
6328    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
6329    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,
6330    int *lengthptr)    int *lengthptr)
# Line 6277  int branchfirstbyte, branchreqbyte; Line 6341  int branchfirstbyte, branchreqbyte;
6341  int length;  int length;
6342  int orig_bracount;  int orig_bracount;
6343  int max_bracount;  int max_bracount;
 int old_external_options = cd->external_options;  
6344  branch_chain bc;  branch_chain bc;
6345    
6346  bc.outer = bcptr;  bc.outer = bcptr;
# Line 6301  pre-compile phase to find out whether an Line 6364  pre-compile phase to find out whether an
6364    
6365  /* 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
6366  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
6367  detect groups that contain recursive back references to themselves. */  detect groups that contain recursive back references to themselves. Note that
6368    only OP_CBRA need be tested here; changing this opcode to one of its variants,
6369    e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
6370    
6371  if (*code == OP_CBRA)  if (*code == OP_CBRA)
6372    {    {
# Line 6347  for (;;) Line 6412  for (;;)
6412      return FALSE;      return FALSE;
6413      }      }
6414    
   /* 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;  
   
6415    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6416    has fewer than the rest. */    has fewer than the rest. */
6417    
# Line 6416  for (;;) Line 6472  for (;;)
6472        {        {
6473        int fixed_length;        int fixed_length;
6474        *code = OP_END;        *code = OP_END;
6475        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);        fixed_length = find_fixedlength(last_branch,  (options & PCRE_UTF8) != 0,
6476            FALSE, cd);
6477        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6478        if (fixed_length == -3)        if (fixed_length == -3)
6479          {          {
# Line 6437  for (;;) Line 6494  for (;;)
6494    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
6495    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
6496    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
6497    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. */  
6498    
6499    if (*ptr != CHAR_VERTICAL_LINE)    if (*ptr != CHAR_VERTICAL_LINE)
6500      {      {
# Line 6564  of the more common cases more precisely. Line 6619  of the more common cases more precisely.
6619    
6620  Arguments:  Arguments:
6621    code           points to start of expression (the bracket)    code           points to start of expression (the bracket)
   options        points to the options setting  
6622    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
6623                    handles up to substring 31; after that we just have to take                    handles up to substring 31; after that we just have to take
6624                    the less precise approach                    the less precise approach
# Line 6574  Returns:     TRUE or FALSE Line 6628  Returns:     TRUE or FALSE
6628  */  */
6629    
6630  static BOOL  static BOOL
6631  is_anchored(register const uschar *code, int *options, unsigned int bracket_map,  is_anchored(register const uschar *code, unsigned int bracket_map,
6632    unsigned int backref_map)    unsigned int backref_map)
6633  {  {
6634  do {  do {
6635     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6636       options, PCRE_MULTILINE, FALSE);       FALSE);
6637     register int op = *scode;     register int op = *scode;
6638    
6639     /* Non-capturing brackets */     /* Non-capturing brackets */
6640    
6641     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6642           op == OP_SBRA || op == OP_SBRAPOS)
6643       {       {
6644       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6645       }       }
6646    
6647     /* Capturing brackets */     /* Capturing brackets */
6648    
6649     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6650                op == OP_SCBRA || op == OP_SCBRAPOS)
6651       {       {
6652       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6653       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
6654       if (!is_anchored(scode, options, new_map, backref_map)) return FALSE;       if (!is_anchored(scode, new_map, backref_map)) return FALSE;
6655       }       }
6656    
6657     /* Other brackets */     /* Other brackets */
6658    
6659     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)
6660       {       {
6661       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6662       }       }
6663    
6664     /* .* 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 6709  is_startline(const uschar *code, unsigne
6709  {  {
6710  do {  do {
6711     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6712       NULL, 0, FALSE);       FALSE);
6713     register int op = *scode;     register int op = *scode;
6714    
6715     /* 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 6736  do {
6736         scode += 1 + LINK_SIZE;         scode += 1 + LINK_SIZE;
6737         break;         break;
6738         }         }
6739       scode = first_significant_code(scode, NULL, 0, FALSE);       scode = first_significant_code(scode, FALSE);
6740       op = *scode;       op = *scode;
6741       }       }
6742    
6743     /* Non-capturing brackets */     /* Non-capturing brackets */
6744    
6745     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6746           op == OP_SBRA || op == OP_SBRAPOS)
6747       {       {
6748       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6749       }       }
6750    
6751     /* Capturing brackets */     /* Capturing brackets */
6752    
6753     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6754                op == OP_SCBRA || op == OP_SCBRAPOS)
6755       {       {
6756       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6757       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 6801  we return that char, otherwise -1.
6801    
6802  Arguments:  Arguments:
6803    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)  
6804    inassert   TRUE if in an assertion    inassert   TRUE if in an assertion
6805    
6806  Returns:     -1 or the fixed first char  Returns:     -1 or the fixed first char
6807  */  */
6808    
6809  static int  static int
6810  find_firstassertedchar(const uschar *code, int *options, BOOL inassert)  find_firstassertedchar(const uschar *code, BOOL inassert)
6811  {  {
6812  register int c = -1;  register int c = -1;
6813  do {  do {
6814     int d;     int d;
6815     const uschar *scode =     int xl = (*code == OP_CBRA || *code == OP_SCBRA ||
6816       first_significant_code(code + 1+LINK_SIZE, options, PCRE_CASELESS, TRUE);               *code == OP_CBRAPOS || *code == OP_SCBRAPOS)? 2:0;
6817       const uschar *scode = first_significant_code(code + 1+LINK_SIZE + xl, TRUE);
6818     register int op = *scode;     register int op = *scode;
6819    
6820     switch(op)     switch(op)
# Line 6765  do { Line 6823  do {
6823       return -1;       return -1;
6824    
6825       case OP_BRA:       case OP_BRA:
6826         case OP_BRAPOS:
6827       case OP_CBRA:       case OP_CBRA:
6828         case OP_SCBRA:
6829         case OP_CBRAPOS:
6830         case OP_SCBRAPOS:
6831       case OP_ASSERT:       case OP_ASSERT:
6832       case OP_ONCE:       case OP_ONCE:
6833       case OP_COND:       case OP_COND:
6834       if ((d = find_firstassertedchar(scode, options, op == OP_ASSERT)) < 0)       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)
6835         return -1;         return -1;
6836       if (c < 0) c = d; else if (c != d) return -1;       if (c < 0) c = d; else if (c != d) return -1;
6837       break;       break;
6838    
6839       case OP_EXACT:       /* Fall through */       case OP_EXACT:
6840       scode += 2;       scode += 2;
6841         /* Fall through */
6842    
6843       case OP_CHAR:       case OP_CHAR:
      case OP_CHARI:  
6844       case OP_PLUS:       case OP_PLUS:
6845       case OP_MINPLUS:       case OP_MINPLUS:
6846       case OP_POSPLUS:       case OP_POSPLUS:
6847       if (!inassert) return -1;       if (!inassert) return -1;
6848       if (c < 0)       if (c < 0) c = scode[1];
6849         {         else if (c != scode[1]) return -1;
6850         c = scode[1];       break;
6851         if ((*options & PCRE_CASELESS) != 0) c |= REQ_CASELESS;  
6852         }       case OP_EXACTI:
6853       else if (c != scode[1]) return -1;       scode += 2;
6854         /* Fall through */
6855    
6856         case OP_CHARI:
6857         case OP_PLUSI:
6858         case OP_MINPLUSI:
6859         case OP_POSPLUSI:
6860         if (!inassert) return -1;
6861         if (c < 0) c = scode[1] | REQ_CASELESS;
6862           else if (c != scode[1]) return -1;
6863       break;       break;
6864       }       }
6865    
# Line 6939  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7010  while (ptr[skipatstart] == CHAR_LEFT_PAR
7010    
7011  utf8 = (options & PCRE_UTF8) != 0;  utf8 = (options & PCRE_UTF8) != 0;
7012    
7013  /* 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
7014  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
7015  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
7016  the end of the string for a short string error, for compatibility with previous  not used here. */
 versions. */  
7017    
7018  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
7019  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
7020       (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1, &errorcode)) >= 0)       (errorcode = _pcre_valid_utf8((USPTR)pattern, -1, erroroffset)) != 0)
7021    {    {
7022    errorcode = ERR44;    errorcode = ERR44;
7023    goto PCRE_EARLY_ERROR_RETURN2;    goto PCRE_EARLY_ERROR_RETURN2;
# Line 7063  outside can help speed up starting point Line 7133  outside can help speed up starting point
7133  ptr += skipatstart;  ptr += skipatstart;
7134  code = cworkspace;  code = cworkspace;
7135  *code = OP_BRA;  *code = OP_BRA;
7136  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
7137    &code, &ptr, &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd,    FALSE, 0, &firstbyte, &reqbyte, NULL, cd, &length);
   &length);  
7138  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
7139    
7140  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 7188  field; this time it's used for rememberi
7188  */  */
7189    
7190  cd->final_bracount = cd->bracount;  /* Save for checking forward references */  cd->final_bracount = cd->bracount;  /* Save for checking forward references */
7191    cd->assert_depth = 0;
7192  cd->bracount = 0;  cd->bracount = 0;
7193  cd->names_found = 0;  cd->names_found = 0;
7194  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 7207  of the function here. */
7207  ptr = (const uschar *)pattern + skipatstart;  ptr = (const uschar *)pattern + skipatstart;
7208  code = (uschar *)codestart;  code = (uschar *)codestart;
7209  *code = OP_BRA;  *code = OP_BRA;
7210  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,  (void)compile_regex(re->options, &code, &ptr, &errorcode, FALSE, FALSE, 0,
7211    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);    &firstbyte, &reqbyte, NULL, cd, NULL);
7212  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
7213  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
7214  re->flags = cd->external_flags;  re->flags = cd->external_flags;
# Line 7204  if (cd->check_lookbehind) Line 7274  if (cd->check_lookbehind)
7274        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
7275        int end_op = *be;        int end_op = *be;
7276        *be = OP_END;        *be = OP_END;
7277        fixed_length = find_fixedlength(cc, re->options, TRUE, cd);        fixed_length = find_fixedlength(cc, (re->options & PCRE_UTF8) != 0, TRUE,
7278            cd);
7279        *be = end_op;        *be = end_op;
7280        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
7281        if (fixed_length < 0)        if (fixed_length < 0)
# Line 7243  start with ^. and also when all branches Line 7314  start with ^. and also when all branches
7314    
7315  if ((re->options & PCRE_ANCHORED) == 0)  if ((re->options & PCRE_ANCHORED) == 0)
7316    {    {
7317    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))  
7318      re->options |= PCRE_ANCHORED;      re->options |= PCRE_ANCHORED;
7319    else    else
7320      {      {
7321      if (firstbyte < 0)      if (firstbyte < 0)
7322        firstbyte = find_firstassertedchar(codestart, &temp_options, FALSE);        firstbyte = find_firstassertedchar(codestart, FALSE);
7323      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */
7324        {        {
7325        int ch = firstbyte & 255;        int ch = firstbyte & 255;

Legend:
Removed from v.602  
changed lines
  Added in v.621

  ViewVC Help
Powered by ViewVC 1.1.5