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

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

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

revision 447 by ph10, Tue Sep 15 18:17:54 2009 UTC revision 459 by ph10, Sun Oct 4 09:21:39 2009 UTC
# Line 341  static const char error_texts[] = Line 341  static const char error_texts[] =
341    "number is too big\0"    "number is too big\0"
342    "subpattern name expected\0"    "subpattern name expected\0"
343    "digit expected after (?+\0"    "digit expected after (?+\0"
344    "] is an invalid data character in JavaScript compatibility mode";    "] is an invalid data character in JavaScript compatibility mode\0"
345      /* 65 */
346      "different names for subpatterns of the same number are not allowed";
347    
348    
349  /* 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 1315  for (;;) Line 1317  for (;;)
1317    
1318      case OP_CALLOUT:      case OP_CALLOUT:
1319      case OP_CREF:      case OP_CREF:
1320        case OP_NCREF:
1321      case OP_RREF:      case OP_RREF:
1322        case OP_NRREF:
1323      case OP_DEF:      case OP_DEF:
1324      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1325      break;      break;
# Line 1331  for (;;) Line 1335  for (;;)
1335    
1336    
1337  /*************************************************  /*************************************************
1338  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1339  *************************************************/  *************************************************/
1340    
1341  /* 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,
1342  if the length is fixed. This is needed for dealing with backward assertions.  if the length is fixed. This is needed for dealing with backward assertions.
1343  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
1344    temporarily terminated with OP_END when this function is called.
1345    
1346    This function is called when a backward assertion is encountered, so that if it
1347    fails, the error message can point to the correct place in the pattern.
1348    However, we cannot do this when the assertion contains subroutine calls,
1349    because they can be forward references. We solve this by remembering this case
1350    and doing the check at the end; a flag specifies which mode we are running in.
1351    
1352  Arguments:  Arguments:
1353    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1354    options  the compiling options    options  the compiling options
1355      atend    TRUE if called when the pattern is complete
1356      cd       the "compile data" structure
1357    
1358  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1359                 or -1 if there is no fixed length,
1360               or -2 if \C was encountered               or -2 if \C was encountered
1361                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1362  */  */
1363    
1364  static int  static int
1365  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1366  {  {
1367  int length = -1;  int length = -1;
1368    
# Line 1360  branch, check the length against that of Line 1375  branch, check the length against that of
1375  for (;;)  for (;;)
1376    {    {
1377    int d;    int d;
1378      uschar *ce, *cs;
1379    register int op = *cc;    register int op = *cc;
1380    switch (op)    switch (op)
1381      {      {
# Line 1367  for (;;) Line 1383  for (;;)
1383      case OP_BRA:      case OP_BRA:
1384      case OP_ONCE:      case OP_ONCE:
1385      case OP_COND:      case OP_COND:
1386      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1387      if (d < 0) return d;      if (d < 0) return d;
1388      branchlength += d;      branchlength += d;
1389      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1389  for (;;) Line 1405  for (;;)
1405      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1406      branchlength = 0;      branchlength = 0;
1407      break;      break;
1408    
1409        /* A true recursion implies not fixed length, but a subroutine call may
1410        be OK. If the subroutine is a forward reference, we can't deal with
1411        it until the end of the pattern, so return -3. */
1412    
1413        case OP_RECURSE:
1414        if (!atend) return -3;
1415        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1416        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1417        if (cc > cs && cc < ce) return -1;                /* Recursion */
1418        d = find_fixedlength(cs + 2, options, atend, cd);
1419        if (d < 0) return d;
1420        branchlength += d;
1421        cc += 1 + LINK_SIZE;
1422        break;
1423    
1424      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1425    
# Line 1403  for (;;) Line 1434  for (;;)
1434    
1435      case OP_REVERSE:      case OP_REVERSE:
1436      case OP_CREF:      case OP_CREF:
1437        case OP_NCREF:
1438      case OP_RREF:      case OP_RREF:
1439        case OP_NRREF:
1440      case OP_DEF:      case OP_DEF:
1441      case OP_OPT:      case OP_OPT:
1442      case OP_CALLOUT:      case OP_CALLOUT:
# Line 1518  for (;;) Line 1551  for (;;)
1551    
1552    
1553  /*************************************************  /*************************************************
1554  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1555  *************************************************/  *************************************************/
1556    
1557  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1558  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1559    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1560    so that it can be called from pcre_study() when finding the minimum matching
1561    length.
1562    
1563  Arguments:  Arguments:
1564    code        points to start of expression    code        points to start of expression
1565    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1566    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1567    
1568  Returns:      pointer to the opcode for the bracket, or NULL if not found  Returns:      pointer to the opcode for the bracket, or NULL if not found
1569  */  */
1570    
1571  static const uschar *  const uschar *
1572  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1573  {  {
1574  for (;;)  for (;;)
1575    {    {
# Line 1545  for (;;) Line 1581  for (;;)
1581    the table is zero; the actual length is stored in the compiled code. */    the table is zero; the actual length is stored in the compiled code. */
1582    
1583    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1584    
1585      /* Handle recursion */
1586    
1587      else if (c == OP_REVERSE)
1588        {
1589        if (number < 0) return (uschar *)code;
1590        code += _pcre_OP_lengths[c];
1591        }
1592    
1593    /* Handle capturing bracket */    /* Handle capturing bracket */
1594    
# Line 4614  we set the flag only if there is a liter Line 4658  we set the flag only if there is a liter
4658            }            }
4659    
4660          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
4661          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
4662            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
4663            except they record that the reference was originally to a name. The
4664            information is used to check duplicate names. */
4665    
4666          slot = cd->name_table;          slot = cd->name_table;
4667          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 4629  we set the flag only if there is a liter Line 4676  we set the flag only if there is a liter
4676            {            {
4677            recno = GET2(slot, 0);            recno = GET2(slot, 0);
4678            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4679              code[1+LINK_SIZE]++;
4680            }            }
4681    
4682          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
# Line 4637  we set the flag only if there is a liter Line 4685  we set the flag only if there is a liter
4685                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4686            {            {
4687            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
4688              code[1+LINK_SIZE]++;
4689            }            }
4690    
4691          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 4829  we set the flag only if there is a liter Line 4878  we set the flag only if there is a liter
4878                }                }
4879              }              }
4880    
4881            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
4882              alphabetical order. Duplicate names for different numbers are
4883              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
4884              number are always OK. (An existing number can be re-used if (?|
4885              appears in the pattern.) In either event, a duplicate name results in
4886              a duplicate entry in the table, even if the number is the same. This
4887              is because the number of names, and hence the table size, is computed
4888              in the pre-compile, and it affects various numbers and pointers which
4889              would all have to be modified, and the compiled code moved down, if
4890              duplicates with the same number were omitted from the table. This
4891              doesn't seem worth the hassle. However, *different* names for the
4892              same number are not permitted. */
4893    
4894            else            else
4895              {              {
4896                BOOL dupname = FALSE;
4897              slot = cd->name_table;              slot = cd->name_table;
4898    
4899              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
4900                {                {
4901                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 4841  we set the flag only if there is a liter Line 4903  we set the flag only if there is a liter
4903                  {                  {
4904                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
4905                    {                    {
4906                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
4907                          (options & PCRE_DUPNAMES) == 0)
4908                      {                      {
4909                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
4910                      goto FAILED;                      goto FAILED;
4911                      }                      }
4912                      else dupname = TRUE;
4913                    }                    }
4914                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
4915                  }                  }
4916    
4917                  /* Make space in the table and break the loop for an earlier
4918                  name. For a duplicate or later name, carry on. We do this for
4919                  duplicates so that in the simple case (when ?(| is not used) they
4920                  are in order of their numbers. */
4921    
4922                if (crc < 0)                if (crc < 0)
4923                  {                  {
4924                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
4925                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
4926                  break;                  break;
4927                  }                  }
4928    
4929                  /* Continue the loop for a later or duplicate name */
4930    
4931                slot += cd->name_entry_size;                slot += cd->name_entry_size;
4932                }                }
4933    
4934                /* For non-duplicate names, check for a duplicate number before
4935                adding the new name. */
4936    
4937                if (!dupname)
4938                  {
4939                  uschar *cslot = cd->name_table;
4940                  for (i = 0; i < cd->names_found; i++)
4941                    {
4942                    if (cslot != slot)
4943                      {
4944                      if (GET2(cslot, 0) == cd->bracount + 1)
4945                        {
4946                        *errorcodeptr = ERR65;
4947                        goto FAILED;
4948                        }
4949                      }
4950                    else i--;
4951                    cslot += cd->name_entry_size;
4952                    }
4953                  }
4954    
4955              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
4956              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
# Line 4864  we set the flag only if there is a liter Line 4958  we set the flag only if there is a liter
4958              }              }
4959            }            }
4960    
4961          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
4962            encountered. */
4963    
         ptr++;                    /* Move past > or ' */  
4964          cd->names_found++;          cd->names_found++;
4965            ptr++;                    /* Move past > or ' */
4966          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
4967    
4968    
# Line 5036  we set the flag only if there is a liter Line 5131  we set the flag only if there is a liter
5131            if (lengthptr == NULL)            if (lengthptr == NULL)
5132              {              {
5133              *code = OP_END;              *code = OP_END;
5134              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);              if (recno != 0)
5135                  called = _pcre_find_bracket(cd->start_code, utf8, recno);
5136    
5137              /* Forward reference */              /* Forward reference */
5138    
# Line 5813  for (;;) Line 5909  for (;;)
5909    
5910      /* If lookbehind, check that this branch matches a fixed-length string, and      /* If lookbehind, check that this branch matches a fixed-length string, and
5911      put the length into the OP_REVERSE item. Temporarily mark the end of the      put the length into the OP_REVERSE item. Temporarily mark the end of the
5912      branch with OP_END. */      branch with OP_END. If the branch contains OP_RECURSE, the result is -3
5913        because there may be forward references that we can't check here. Set a
5914        flag to cause another lookbehind check at the end. Why not do it all at the
5915        end? Because common, erroneous checks are picked up here and the offset of
5916        the problem can be shown. */
5917    
5918      if (lookbehind)      if (lookbehind)
5919        {        {
5920        int fixed_length;        int fixed_length;
5921        *code = OP_END;        *code = OP_END;
5922        fixed_length = find_fixedlength(last_branch, options);        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);
5923        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
5924        if (fixed_length < 0)        if (fixed_length == -3)
5925            {
5926            cd->check_lookbehind = TRUE;
5927            }
5928          else if (fixed_length < 0)
5929          {          {
5930          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;
5931          *ptrptr = ptr;          *ptrptr = ptr;
5932          return FALSE;          return FALSE;
5933          }          }
5934        PUT(reverse_count, 0, fixed_length);        else { PUT(reverse_count, 0, fixed_length); }
5935        }        }
5936      }      }
5937    
# Line 6061  do { Line 6165  do {
6165       switch (*scode)       switch (*scode)
6166         {         {
6167         case OP_CREF:         case OP_CREF:
6168           case OP_NCREF:
6169         case OP_RREF:         case OP_RREF:
6170           case OP_NRREF:
6171         case OP_DEF:         case OP_DEF:
6172         return FALSE;         return FALSE;
6173    
# Line 6230  int length = 1;  /* For final END opcode Line 6336  int length = 1;  /* For final END opcode
6336  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
6337  int errorcode = 0;  int errorcode = 0;
6338  int skipatstart = 0;  int skipatstart = 0;
6339  #ifdef SUPPORT_UTF8  BOOL utf8 = (options & PCRE_UTF8) != 0;
 BOOL utf8;  
 #endif  
6340  size_t size;  size_t size;
6341  uschar *code;  uschar *code;
6342  const uschar *codestart;  const uschar *codestart;
# Line 6329  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 6433  while (ptr[skipatstart] == CHAR_LEFT_PAR
6433  /* Can't support UTF8 unless PCRE has been compiled to include the code. */  /* Can't support UTF8 unless PCRE has been compiled to include the code. */
6434    
6435  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
 utf8 = (options & PCRE_UTF8) != 0;  
6436  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
6437       (*erroroffset = _pcre_valid_utf8((uschar *)pattern, -1)) >= 0)       (*erroroffset = _pcre_valid_utf8((uschar *)pattern, -1)) >= 0)
6438    {    {
# Line 6337  if (utf8 && (options & PCRE_NO_UTF8_CHEC Line 6440  if (utf8 && (options & PCRE_NO_UTF8_CHEC
6440    goto PCRE_EARLY_ERROR_RETURN2;    goto PCRE_EARLY_ERROR_RETURN2;
6441    }    }
6442  #else  #else
6443  if ((options & PCRE_UTF8) != 0)  if (utf8)
6444    {    {
6445    errorcode = ERR32;    errorcode = ERR32;
6446    goto PCRE_EARLY_ERROR_RETURN;    goto PCRE_EARLY_ERROR_RETURN;
# Line 6501  cd->start_code = codestart; Line 6604  cd->start_code = codestart;
6604  cd->hwm = cworkspace;  cd->hwm = cworkspace;
6605  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6606  cd->had_accept = FALSE;  cd->had_accept = FALSE;
6607    cd->check_lookbehind = FALSE;
6608  cd->open_caps = NULL;  cd->open_caps = NULL;
6609    
6610  /* Set up a starting, non-extracting bracket, then compile the expression. On  /* Set up a starting, non-extracting bracket, then compile the expression. On
# Line 6540  while (errorcode == 0 && cd->hwm > cwork Line 6644  while (errorcode == 0 && cd->hwm > cwork
6644    cd->hwm -= LINK_SIZE;    cd->hwm -= LINK_SIZE;
6645    offset = GET(cd->hwm, 0);    offset = GET(cd->hwm, 0);
6646    recno = GET(codestart, offset);    recno = GET(codestart, offset);
6647    groupptr = find_bracket(codestart, (re->options & PCRE_UTF8) != 0, recno);    groupptr = _pcre_find_bracket(codestart, utf8, recno);
6648    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
6649      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, groupptr - codestart);
6650    }    }
# Line 6550  subpattern. */ Line 6654  subpattern. */
6654    
6655  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;
6656    
6657    /* If there were any lookbehind assertions that contained OP_RECURSE
6658    (recursions or subroutine calls), a flag is set for them to be checked here,
6659    because they may contain forward references. Actual recursions can't be fixed
6660    length, but subroutine calls can. It is done like this so that those without
6661    OP_RECURSE that are not fixed length get a diagnosic with a useful offset. The
6662    exceptional ones forgo this. We scan the pattern to check that they are fixed
6663    length, and set their lengths. */
6664    
6665    if (cd->check_lookbehind)
6666      {
6667      uschar *cc = (uschar *)codestart;
6668    
6669      /* Loop, searching for OP_REVERSE items, and process those that do not have
6670      their length set. (Actually, it will also re-process any that have a length
6671      of zero, but that is a pathological case, and it does no harm.) When we find
6672      one, we temporarily terminate the branch it is in while we scan it. */
6673    
6674      for (cc = (uschar *)_pcre_find_bracket(codestart, utf8, -1);
6675           cc != NULL;
6676           cc = (uschar *)_pcre_find_bracket(cc, utf8, -1))
6677        {
6678        if (GET(cc, 1) == 0)
6679          {
6680          int fixed_length;
6681          uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
6682          int end_op = *be;
6683          *be = OP_END;
6684          fixed_length = find_fixedlength(cc, re->options, TRUE, cd);
6685          *be = end_op;
6686          DPRINTF(("fixed length = %d\n", fixed_length));
6687          if (fixed_length < 0)
6688            {
6689            errorcode = (fixed_length == -2)? ERR36 : ERR25;
6690            break;
6691            }
6692          PUT(cc, 1, fixed_length);
6693          }
6694        cc += 1 + LINK_SIZE;
6695        }
6696      }
6697    
6698  /* Failed to compile, or error while post-processing */  /* Failed to compile, or error while post-processing */
6699    
6700  if (errorcode != 0)  if (errorcode != 0)

Legend:
Removed from v.447  
changed lines
  Added in v.459

  ViewVC Help
Powered by ViewVC 1.1.5