/[pcre]/code/trunk/pcre_compile.c

Diff of /code/trunk/pcre_compile.c

Parent Directory | Revision Log | View Patch Patch

-revision 447 by ph10,
Tue Sep 15 18:17:54 2009 UTC
+revision 487 by ph10,
Wed Jan  6 10:26:55 2010 UTC
 Line 6
  and semantics are as close as possible to those of the Perl 5 language.
                         Written by Philip Hazel
-            Copyright (c) 1997-2009 University of Cambridge
+            Copyright (c) 1997-2010 University of Cambridge
  -----------------------------------------------------------------------------
  Redistribution and use in source and binary forms, with or without
 Line 53 
 supporting internal functions that are n
  #include "pcre_internal.h"
- /* When DEBUG is defined, we need the pcre_printint() function, which is also
+ /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
- used by pcretest. DEBUG is not defined when building a production library. */
+ also used by pcretest. PCRE_DEBUG is not defined when building a production
+ library. */
- #ifdef DEBUG
+ #ifdef PCRE_DEBUG
  #include "pcre_printint.src"
  #endif
-Line 341 
 static const char error_texts[] =
+Line 342 
 static const char error_texts[] =
    "number is too big\0"
    "subpattern name expected\0"
    "digit expected after (?+\0"
-   "] is an invalid data character in JavaScript compatibility mode";
+   "] is an invalid data character in JavaScript compatibility mode\0"
+   /* 65 */
+   "different names for subpatterns of the same number are not allowed";
  /* Table to identify digits and hex digits. This is used when compiling
-Line 1100 
 if (ptr[0] == CHAR_LEFT_PARENTHESIS)
+Line 1103 
 if (ptr[0] == CHAR_LEFT_PARENTHESIS)
        if (name != NULL && lorn == ptr - thisname &&
            strncmp((const char *)name, (const char *)thisname, lorn) == 0)
          return *count;
        term++;
        }
      }
    }
-Line 1146 
 for (; *ptr != 0; ptr++)
+Line 1149 
 for (; *ptr != 0; ptr++)
            break;
          }
        else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
          {
          negate_class = TRUE;
          ptr++;
          }
        else break;
        }
-Line 1315 
 for (;;)
+Line 1318 
 for (;;)
      case OP_CALLOUT:
      case OP_CREF:
+     case OP_NCREF:
      case OP_RREF:
+     case OP_NRREF:
      case OP_DEF:
      code += _pcre_OP_lengths[*code];
      break;
-Line 1331 
 for (;;)
+Line 1336 
 for (;;)
  /*************************************************
- *        Find the fixed length of a pattern      *
+ *        Find the fixed length of a branch       *
  *************************************************/
- /* Scan a pattern and compute the fixed length of subject that will match it,
+ /* Scan a branch and compute the fixed length of subject that will match it,
  if the length is fixed. This is needed for dealing with backward assertions.
- In UTF8 mode, the result is in characters rather than bytes.
+ In UTF8 mode, the result is in characters rather than bytes. The branch is
+ temporarily terminated with OP_END when this function is called.
+ This function is called when a backward assertion is encountered, so that if it
+ fails, the error message can point to the correct place in the pattern.
+ However, we cannot do this when the assertion contains subroutine calls,
+ because they can be forward references. We solve this by remembering this case
+ and doing the check at the end; a flag specifies which mode we are running in.
  Arguments:
    code     points to the start of the pattern (the bracket)
    options  the compiling options
+   atend    TRUE if called when the pattern is complete
+   cd       the "compile data" structure
- Returns:   the fixed length, or -1 if there is no fixed length,
+ Returns:   the fixed length,
+              or -1 if there is no fixed length,
               or -2 if \C was encountered
+              or -3 if an OP_RECURSE item was encountered and atend is FALSE
  */
  static int
- find_fixedlength(uschar *code, int options)
+ find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
  {
  int length = -1;
-Line 1360 
 branch, check the length against that of
+Line 1376 
 branch, check the length against that of
  for (;;)
    {
    int d;
+   uschar *ce, *cs;
    register int op = *cc;
    switch (op)
      {
-Line 1367 
 for (;;)
+Line 1384 
 for (;;)
      case OP_BRA:
      case OP_ONCE:
      case OP_COND:
-     d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);
+     d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
      if (d < 0) return d;
      branchlength += d;
      do cc += GET(cc, 1); while (*cc == OP_ALT);
-Line 1390 
 for (;;)
+Line 1407 
 for (;;)
      branchlength = 0;
      break;
+     /* A true recursion implies not fixed length, but a subroutine call may
+     be OK. If the subroutine is a forward reference, we can't deal with
+     it until the end of the pattern, so return -3. */
+     case OP_RECURSE:
+     if (!atend) return -3;
+     cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
+     do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
+     if (cc > cs && cc < ce) return -1;                /* Recursion */
+     d = find_fixedlength(cs + 2, options, atend, cd);
+     if (d < 0) return d;
+     branchlength += d;
+     cc += 1 + LINK_SIZE;
+     break;
      /* Skip over assertive subpatterns */
      case OP_ASSERT:
-Line 1403 
 for (;;)
+Line 1435 
 for (;;)
      case OP_REVERSE:
      case OP_CREF:
+     case OP_NCREF:
      case OP_RREF:
+     case OP_NRREF:
      case OP_DEF:
      case OP_OPT:
      case OP_CALLOUT:
-Line 1426 
 for (;;)
+Line 1460 
 for (;;)
      branchlength++;
      cc += 2;
  #ifdef SUPPORT_UTF8
      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
  #endif
      break;
-Line 1438 
 for (;;)
+Line 1472 
 for (;;)
      branchlength += GET2(cc,1);
      cc += 4;
  #ifdef SUPPORT_UTF8
      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
  #endif
      break;
-Line 1518 
 for (;;)
+Line 1552 
 for (;;)
  /*************************************************
- *    Scan compiled regex for numbered bracket    *
+ *    Scan compiled regex for specific bracket    *
  *************************************************/
  /* This little function scans through a compiled pattern until it finds a
- capturing bracket with the given number.
+ capturing bracket with the given number, or, if the number is negative, an
+ instance of OP_REVERSE for a lookbehind. The function is global in the C sense
+ so that it can be called from pcre_study() when finding the minimum matching
+ length.
  Arguments:
    code        points to start of expression
    utf8        TRUE in UTF-8 mode
-   number      the required bracket number
+   number      the required bracket number or negative to find a lookbehind
  Returns:      pointer to the opcode for the bracket, or NULL if not found
  */
- static const uschar *
+ const uschar *
- find_bracket(const uschar *code, BOOL utf8, int number)
+ _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
  {
  for (;;)
    {
-Line 1546 
 for (;;)
+Line 1583 
 for (;;)
    if (c == OP_XCLASS) code += GET(code, 1);
+   /* Handle recursion */
+   else if (c == OP_REVERSE)
+     {
+     if (number < 0) return (uschar *)code;
+     code += _pcre_OP_lengths[c];
+     }
    /* Handle capturing bracket */
    else if (c == OP_CBRA)
-Line 1913 
 for (code = first_significant_code(code
+Line 1958 
 for (code = first_significant_code(code
      case OP_POSQUERY:
      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
      break;
      case OP_UPTO:
      case OP_MINUPTO:
      case OP_POSUPTO:
-Line 1950 
 static BOOL
+Line 1995 
 static BOOL
  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
    BOOL utf8)
  {
- while (bcptr != NULL && bcptr->current >= code)
+ while (bcptr != NULL && bcptr->current_branch >= code)
    {
-   if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;
+   if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8))
+     return FALSE;
    bcptr = bcptr->outer;
    }
  return TRUE;
-Line 2614 
 BOOL utf8 = FALSE;
+Line 2660 
 BOOL utf8 = FALSE;
  uschar *utf8_char = NULL;
  #endif
- #ifdef DEBUG
+ #ifdef PCRE_DEBUG
  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
  #endif
-Line 2673 
 for (;; ptr++)
+Line 2719 
 for (;; ptr++)
    if (lengthptr != NULL)
      {
- #ifdef DEBUG
+ #ifdef PCRE_DEBUG
      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
  #endif
      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */
-Line 3871 
 we set the flag only if there is a liter
+Line 3917 
 we set the flag only if there is a liter
        if (repeat_max == 0) goto END_REPEAT;
        /*--------------------------------------------------------------------*/
        /* This code is obsolete from release 8.00; the restriction was finally
        removed: */
        /* All real repeats make it impossible to handle partial matching (maybe
        one day we will be able to remove this restriction). */
        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
        /*--------------------------------------------------------------------*/
        /* Combine the op_type with the repeat_type */
-Line 4026 
 we set the flag only if there is a liter
+Line 4072 
 we set the flag only if there is a liter
          goto END_REPEAT;
          }
        /*--------------------------------------------------------------------*/
        /* This code is obsolete from release 8.00; the restriction was finally
        removed: */
-Line 4034 
 we set the flag only if there is a liter
+Line 4080 
 we set the flag only if there is a liter
        one day we will be able to remove this restriction). */
        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
        /*--------------------------------------------------------------------*/
        if (repeat_min == 0 && repeat_max == -1)
          *code++ = OP_CRSTAR + repeat_type;
-Line 4169 
 we set the flag only if there is a liter
+Line 4215 
 we set the flag only if there is a liter
            {
            /* In the pre-compile phase, we don't actually do the replication. We
            just adjust the length as if we had. Do some paranoid checks for
-           potential integer overflow. */
+           potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
+           integer type when available, otherwise double. */
            if (lengthptr != NULL)
              {
              int delta = (repeat_min - 1)*length_prevgroup;
-             if ((double)(repeat_min - 1)*(double)length_prevgroup >
+             if ((INT64_OR_DOUBLE)(repeat_min - 1)*
-                                                             (double)INT_MAX ||
+                   (INT64_OR_DOUBLE)length_prevgroup >
+                     (INT64_OR_DOUBLE)INT_MAX ||
                  OFLOW_MAX - *lengthptr < delta)
                {
                *errorcodeptr = ERR20;
-Line 4221 
 we set the flag only if there is a liter
+Line 4269 
 we set the flag only if there is a liter
          just adjust the length as if we had. For each repetition we must add 1
          to the length for BRAZERO and for all but the last repetition we must
          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
-         paranoid checks to avoid integer overflow. */
+         paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
+         a 64-bit integer type when available, otherwise double. */
          if (lengthptr != NULL && repeat_max > 0)
            {
            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
 - 2*LINK_SIZE;   /* Last one doesn't nest */
-           if ((double)repeat_max *
+           if ((INT64_OR_DOUBLE)repeat_max *
-                 (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
+                 (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
-                   > (double)INT_MAX ||
+                   > (INT64_OR_DOUBLE)INT_MAX ||
                OFLOW_MAX - *lengthptr < delta)
              {
              *errorcodeptr = ERR20;
-Line 4349 
 we set the flag only if there is a liter
+Line 4398 
 we set the flag only if there is a liter
      if (possessive_quantifier)
        {
        int len;
        if (*tempcode == OP_TYPEEXACT)
          tempcode += _pcre_OP_lengths[*tempcode] +
            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
          {
          tempcode += _pcre_OP_lengths[*tempcode];
-Line 4361 
 we set the flag only if there is a liter
+Line 4410 
 we set the flag only if there is a liter
          if (utf8 && tempcode[-1] >= 0xc0)
            tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
  #endif
          }
        len = code - tempcode;
        if (len > 0) switch (*tempcode)
          {
-Line 4441 
 we set the flag only if there is a liter
+Line 4490 
 we set the flag only if there is a liter
              strncmp((char *)name, vn, namelen) == 0)
            {
            /* Check for open captures before ACCEPT */
            if (verbs[i].op == OP_ACCEPT)
              {
              open_capitem *oc;
              cd->had_accept = TRUE;
              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
                {
                *code++ = OP_CLOSE;
                PUT2INC(code, 0, oc->number);
                }
              }
            *code++ = verbs[i].op;
            break;
            }
-Line 4614 
 we set the flag only if there is a liter
+Line 4663 
 we set the flag only if there is a liter
            }
          /* Otherwise (did not start with "+" or "-"), start by looking for the
-         name. */
+         name. If we find a name, add one to the opcode to change OP_CREF or
+         OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
+         except they record that the reference was originally to a name. The
+         information is used to check duplicate names. */
          slot = cd->name_table;
          for (i = 0; i < cd->names_found; i++)
-Line 4629 
 we set the flag only if there is a liter
+Line 4681 
 we set the flag only if there is a liter
            {
            recno = GET2(slot, 0);
            PUT2(code, 2+LINK_SIZE, recno);
+           code[1+LINK_SIZE]++;
            }
          /* Search the pattern for a forward reference */
-Line 4637 
 we set the flag only if there is a liter
+Line 4690 
 we set the flag only if there is a liter
                          (options & PCRE_EXTENDED) != 0)) > 0)
            {
            PUT2(code, 2+LINK_SIZE, i);
+           code[1+LINK_SIZE]++;
            }
          /* If terminator == 0 it means that the name followed directly after
-Line 4829 
 we set the flag only if there is a liter
+Line 4883 
 we set the flag only if there is a liter
                }
              }
-           /* In the real compile, create the entry in the table */
+           /* In the real compile, create the entry in the table, maintaining
+           alphabetical order. Duplicate names for different numbers are
+           permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
+           number are always OK. (An existing number can be re-used if (?|
+           appears in the pattern.) In either event, a duplicate name results in
+           a duplicate entry in the table, even if the number is the same. This
+           is because the number of names, and hence the table size, is computed
+           in the pre-compile, and it affects various numbers and pointers which
+           would all have to be modified, and the compiled code moved down, if
+           duplicates with the same number were omitted from the table. This
+           doesn't seem worth the hassle. However, *different* names for the
+           same number are not permitted. */
            else
              {
+             BOOL dupname = FALSE;
              slot = cd->name_table;
              for (i = 0; i < cd->names_found; i++)
                {
                int crc = memcmp(name, slot+2, namelen);
-Line 4841 
 we set the flag only if there is a liter
+Line 4908 
 we set the flag only if there is a liter
                  {
                  if (slot[2+namelen] == 0)
                    {
-                   if ((options & PCRE_DUPNAMES) == 0)
+                   if (GET2(slot, 0) != cd->bracount + 1 &&
+                       (options & PCRE_DUPNAMES) == 0)
                      {
                      *errorcodeptr = ERR43;
                      goto FAILED;
                      }
+                   else dupname = TRUE;
                    }
-                 else crc = -1;      /* Current name is substring */
+                 else crc = -1;      /* Current name is a substring */
                  }
+               /* Make space in the table and break the loop for an earlier
+               name. For a duplicate or later name, carry on. We do this for
+               duplicates so that in the simple case (when ?(| is not used) they
+               are in order of their numbers. */
                if (crc < 0)
                  {
                  memmove(slot + cd->name_entry_size, slot,
                    (cd->names_found - i) * cd->name_entry_size);
                  break;
                  }
+               /* Continue the loop for a later or duplicate name */
                slot += cd->name_entry_size;
                }
+             /* For non-duplicate names, check for a duplicate number before
+             adding the new name. */
+             if (!dupname)
+               {
+               uschar *cslot = cd->name_table;
+               for (i = 0; i < cd->names_found; i++)
+                 {
+                 if (cslot != slot)
+                   {
+                   if (GET2(cslot, 0) == cd->bracount + 1)
+                     {
+                     *errorcodeptr = ERR65;
+                     goto FAILED;
+                     }
+                   }
+                 else i--;
+                 cslot += cd->name_entry_size;
+                 }
+               }
              PUT2(slot, 0, cd->bracount + 1);
              memcpy(slot + 2, name, namelen);
              slot[2+namelen] = 0;
              }
            }
-         /* In both cases, count the number of names we've encountered. */
+         /* In both pre-compile and compile, count the number of names we've
+         encountered. */
-         ptr++;                    /* Move past > or ' */
          cd->names_found++;
+         ptr++;                    /* Move past > or ' */
          goto NUMBERED_GROUP;
-Line 5036 
 we set the flag only if there is a liter
+Line 5136 
 we set the flag only if there is a liter
            if (lengthptr == NULL)
              {
              *code = OP_END;
-             if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);
+             if (recno != 0)
+               called = _pcre_find_bracket(cd->start_code, utf8, recno);
              /* Forward reference */
-Line 5152 
 we set the flag only if there is a liter
+Line 5253 
 we set the flag only if there is a liter
              {
              cd->external_options = newoptions;
              }
-          else
+           else
              {
              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))
                {
-Line 5687 
 int branchfirstbyte, branchreqbyte;
+Line 5788 
 int branchfirstbyte, branchreqbyte;
  int length;
  int orig_bracount;
  int max_bracount;
+ int old_external_options = cd->external_options;
  branch_chain bc;
  bc.outer = bcptr;
- bc.current = code;
+ bc.current_branch = code;
  firstbyte = reqbyte = REQ_UNSET;
-Line 5716 
 if (*code == OP_CBRA)
+Line 5818 
 if (*code == OP_CBRA)
    capnumber = GET2(code, 1 + LINK_SIZE);
    capitem.number = capnumber;
    capitem.next = cd->open_caps;
    cd->open_caps = &capitem;
    }
  /* Offset is set zero to mark that this bracket is still open */
-Line 5763 
 for (;;)
+Line 5865 
 for (;;)
      return FALSE;
      }
+   /* 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;
    /* Keep the highest bracket count in case (?| was used and some branch
    has fewer than the rest. */
-Line 5813 
 for (;;)
+Line 5924 
 for (;;)
      /* If lookbehind, check that this branch matches a fixed-length string, and
      put the length into the OP_REVERSE item. Temporarily mark the end of the
-     branch with OP_END. */
+     branch with OP_END. If the branch contains OP_RECURSE, the result is -3
+     because there may be forward references that we can't check here. Set a
+     flag to cause another lookbehind check at the end. Why not do it all at the
+     end? Because common, erroneous checks are picked up here and the offset of
+     the problem can be shown. */
      if (lookbehind)
        {
        int fixed_length;
        *code = OP_END;
-       fixed_length = find_fixedlength(last_branch, options);
+       fixed_length = find_fixedlength(last_branch, options, FALSE, cd);
        DPRINTF(("fixed length = %d\n", fixed_length));
-       if (fixed_length < 0)
+       if (fixed_length == -3)
+         {
+         cd->check_lookbehind = TRUE;
+         }
+       else if (fixed_length < 0)
          {
          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;
          *ptrptr = ptr;
          return FALSE;
          }
-       PUT(reverse_count, 0, fixed_length);
+       else { PUT(reverse_count, 0, fixed_length); }
        }
      }
-Line 5854 
 for (;;)
+Line 5973 
 for (;;)
          }
        while (branch_length > 0);
        }
      /* If it was a capturing subpattern, remove it from the chain. */
      if (capnumber > 0) cd->open_caps = cd->open_caps->next;
      /* Fill in the ket */
-Line 5865 
 for (;;)
+Line 5984 
 for (;;)
      PUT(code, 1, code - start_bracket);
      code += 1 + LINK_SIZE;
-     /* Resetting option if needed */
+     /* Reset options if needed. */
      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)
        {
-Line 5914 
 for (;;)
+Line 6033 
 for (;;)
      {
      *code = OP_ALT;
      PUT(code, 1, code - last_branch);
-     bc.current = last_branch = code;
+     bc.current_branch = last_branch = code;
      code += 1 + LINK_SIZE;
      }
-Line 6061 
 do {
+Line 6180 
 do {
       switch (*scode)
         {
         case OP_CREF:
+        case OP_NCREF:
         case OP_RREF:
+        case OP_NRREF:
         case OP_DEF:
         return FALSE;
-Line 6230 
 int length = 1;  /* For final END opcode
+Line 6351 
 int length = 1;  /* For final END opcode
  int firstbyte, reqbyte, newline;
  int errorcode = 0;
  int skipatstart = 0;
- #ifdef SUPPORT_UTF8
+ BOOL utf8 = (options & PCRE_UTF8) != 0;
- BOOL utf8;
- #endif
  size_t size;
  uschar *code;
  const uschar *codestart;
-Line 6329 
 while (ptr[skipatstart] == CHAR_LEFT_PAR
+Line 6448 
 while (ptr[skipatstart] == CHAR_LEFT_PAR
  /* Can't support UTF8 unless PCRE has been compiled to include the code. */
  #ifdef SUPPORT_UTF8
- utf8 = (options & PCRE_UTF8) != 0;
  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
-      (*erroroffset = _pcre_valid_utf8((uschar *)pattern, -1)) >= 0)
+      (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)
    {
    errorcode = ERR44;
    goto PCRE_EARLY_ERROR_RETURN2;
    }
  #else
- if ((options & PCRE_UTF8) != 0)
+ if (utf8)
    {
    errorcode = ERR32;
    goto PCRE_EARLY_ERROR_RETURN;
-Line 6501 
 cd->start_code = codestart;
+Line 6619 
 cd->start_code = codestart;
  cd->hwm = cworkspace;
  cd->req_varyopt = 0;
  cd->had_accept = FALSE;
+ cd->check_lookbehind = FALSE;
  cd->open_caps = NULL;
  /* Set up a starting, non-extracting bracket, then compile the expression. On
-Line 6527 
 if debugging, leave the test till after
+Line 6646 
 if debugging, leave the test till after
  *code++ = OP_END;
- #ifndef DEBUG
+ #ifndef PCRE_DEBUG
  if (code - codestart > length) errorcode = ERR23;
  #endif
-Line 6540 
 while (errorcode == 0 && cd->hwm > cwork
+Line 6659 
 while (errorcode == 0 && cd->hwm > cwork
    cd->hwm -= LINK_SIZE;
    offset = GET(cd->hwm, 0);
    recno = GET(codestart, offset);
-   groupptr = find_bracket(codestart, (re->options & PCRE_UTF8) != 0, recno);
+   groupptr = _pcre_find_bracket(codestart, utf8, recno);
    if (groupptr == NULL) errorcode = ERR53;
      else PUT(((uschar *)codestart), offset, groupptr - codestart);
    }
-Line 6550 
 subpattern. */
+Line 6669 
 subpattern. */
  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;
+ /* If there were any lookbehind assertions that contained OP_RECURSE
+ (recursions or subroutine calls), a flag is set for them to be checked here,
+ because they may contain forward references. Actual recursions can't be fixed
+ length, but subroutine calls can. It is done like this so that those without
+ OP_RECURSE that are not fixed length get a diagnosic with a useful offset. The
+ exceptional ones forgo this. We scan the pattern to check that they are fixed
+ length, and set their lengths. */
+ if (cd->check_lookbehind)
+   {
+   uschar *cc = (uschar *)codestart;
+   /* Loop, searching for OP_REVERSE items, and process those that do not have
+   their length set. (Actually, it will also re-process any that have a length
+   of zero, but that is a pathological case, and it does no harm.) When we find
+   one, we temporarily terminate the branch it is in while we scan it. */
+   for (cc = (uschar *)_pcre_find_bracket(codestart, utf8, -1);
+        cc != NULL;
+        cc = (uschar *)_pcre_find_bracket(cc, utf8, -1))
+     {
+     if (GET(cc, 1) == 0)
+       {
+       int fixed_length;
+       uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
+       int end_op = *be;
+       *be = OP_END;
+       fixed_length = find_fixedlength(cc, re->options, TRUE, cd);
+       *be = end_op;
+       DPRINTF(("fixed length = %d\n", fixed_length));
+       if (fixed_length < 0)
+         {
+         errorcode = (fixed_length == -2)? ERR36 : ERR25;
+         break;
+         }
+       PUT(cc, 1, fixed_length);
+       }
+     cc += 1 + LINK_SIZE;
+     }
+   }
  /* Failed to compile, or error while post-processing */
  if (errorcode != 0)
-Line 6610 
 if (reqbyte >= 0 &&
+Line 6770 
 if (reqbyte >= 0 &&
  /* Print out the compiled data if debugging is enabled. This is never the
  case when building a production library. */
- #ifdef DEBUG
+ #ifdef PCRE_DEBUG
  printf("Length = %d top_bracket = %d top_backref = %d\n",
    length, re->top_bracket, re->top_backref);
-Line 6648 
 if (code - codestart > length)
+Line 6808 
 if (code - codestart > length)
    if (errorcodeptr != NULL) *errorcodeptr = ERR23;
    return NULL;
    }
- #endif   /* DEBUG */
+ #endif   /* PCRE_DEBUG */
  return (pcre *)re;
  }

 Legend:



Removed from v.447
 


changed lines


 
Added in v.487
 Legend:



Removed from v.447
 


changed lines


 
Added in v.487
-Removed from v.447
+Added in v.487

webmaster@exim.org	ViewVC Help
Powered by ViewVC 1.1.12