/[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 721 by ph10, Fri Oct 7 15:51:39 2011 UTC revision 807 by ph10, Sun Dec 18 10:03:38 2011 UTC
# Line 88  so this number is very generous. Line 88  so this number is very generous.
88  The same workspace is used during the second, actual compile phase for  The same workspace is used during the second, actual compile phase for
89  remembering forward references to groups so that they can be filled in at the  remembering forward references to groups so that they can be filled in at the
90  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
91  is 4 there is plenty of room. */  is 4 there is plenty of room for most patterns. However, the memory can get
92    filled up by repetitions of forward references, for example patterns like
93    /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
94    that the workspace is expanded using malloc() in this situation. The value
95    below is therefore a minimum, and we put a maximum on it for safety. The
96    minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
97    kicks in at the same number of forward references in all cases. */
98    
99  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
100    #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
101    
102  /* The overrun tests check for a slightly smaller size so that they detect the  /* The overrun tests check for a slightly smaller size so that they detect the
103  overrun before it actually does run off the end of the data block. */  overrun before it actually does run off the end of the data block. */
104    
105  #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)  #define WORK_SIZE_SAFETY_MARGIN (100)
106    
107    
108  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
# Line 410  static const char error_texts[] = Line 417  static const char error_texts[] =
417    "this version of PCRE is not compiled with PCRE_UCP support\0"    "this version of PCRE is not compiled with PCRE_UCP support\0"
418    "\\c must be followed by an ASCII character\0"    "\\c must be followed by an ASCII character\0"
419    "\\k is not followed by a braced, angle-bracketed, or quoted name\0"    "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
420      /* 70 */
421      "internal error: unknown opcode in find_fixedlength()\0"
422      "\\N is not supported in a class\0"
423      "too many forward references\0"
424    ;    ;
425    
426  /* 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 578  return s; Line 589  return s;
589    
590    
591  /*************************************************  /*************************************************
592    *           Expand the workspace                 *
593    *************************************************/
594    
595    /* This function is called during the second compiling phase, if the number of
596    forward references fills the existing workspace, which is originally a block on
597    the stack. A larger block is obtained from malloc() unless the ultimate limit
598    has been reached or the increase will be rather small.
599    
600    Argument: pointer to the compile data block
601    Returns:  0 if all went well, else an error number
602    */
603    
604    static int
605    expand_workspace(compile_data *cd)
606    {
607    uschar *newspace;
608    int newsize = cd->workspace_size * 2;
609    
610    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
611    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
612        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
613     return ERR72;
614    
615    newspace = (pcre_malloc)(newsize);
616    if (newspace == NULL) return ERR21;
617    
618    memcpy(newspace, cd->start_workspace, cd->workspace_size);
619    cd->hwm = (uschar *)newspace + (cd->hwm - cd->start_workspace);
620    if (cd->workspace_size > COMPILE_WORK_SIZE)
621      (pcre_free)((void *)cd->start_workspace);
622    cd->start_workspace = newspace;
623    cd->workspace_size = newsize;
624    return 0;
625    }
626    
627    
628    
629    /*************************************************
630  *            Check for counted repeat            *  *            Check for counted repeat            *
631  *************************************************/  *************************************************/
632    
# Line 676  else Line 725  else
725    
726      case CHAR_l:      case CHAR_l:
727      case CHAR_L:      case CHAR_L:
728        *errorcodeptr = ERR37;
729        break;
730    
731      case CHAR_u:      case CHAR_u:
732        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
733          {
734          /* In JavaScript, \u must be followed by four hexadecimal numbers.
735          Otherwise it is a lowercase u letter. */
736          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0
737               && (digitab[ptr[3]] & ctype_xdigit) != 0 && (digitab[ptr[4]] & ctype_xdigit) != 0)
738            {
739            c = 0;
740            for (i = 0; i < 4; ++i)
741              {
742              register int cc = *(++ptr);
743    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
744              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
745              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
746    #else           /* EBCDIC coding */
747              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
748              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
749    #endif
750              }
751            }
752          }
753        else
754          *errorcodeptr = ERR37;
755        break;
756    
757      case CHAR_U:      case CHAR_U:
758      *errorcodeptr = ERR37;      /* In JavaScript, \U is an uppercase U letter. */
759        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
760      break;      break;
761    
762      /* In a character class, \g is just a literal "g". Outside a character      /* In a character class, \g is just a literal "g". Outside a character
# Line 828  else Line 906  else
906      treated as a data character. */      treated as a data character. */
907    
908      case CHAR_x:      case CHAR_x:
909        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
910          {
911          /* In JavaScript, \x must be followed by two hexadecimal numbers.
912          Otherwise it is a lowercase x letter. */
913          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0)
914            {
915            c = 0;
916            for (i = 0; i < 2; ++i)
917              {
918              register int cc = *(++ptr);
919    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
920              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
921              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
922    #else           /* EBCDIC coding */
923              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
924              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
925    #endif
926              }
927            }
928          break;
929          }
930    
931      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
932        {        {
933        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
# Line 1475  Arguments: Line 1575  Arguments:
1575    
1576  Returns:   the fixed length,  Returns:   the fixed length,
1577               or -1 if there is no fixed length,               or -1 if there is no fixed length,
1578               or -2 if \C was encountered               or -2 if \C was encountered (in UTF-8 mode only)
1579               or -3 if an OP_RECURSE item was encountered and atend is FALSE               or -3 if an OP_RECURSE item was encountered and atend is FALSE
1580                 or -4 if an unknown opcode was encountered (internal error)
1581  */  */
1582    
1583  static int  static int
# Line 1500  for (;;) Line 1601  for (;;)
1601      /* We only need to continue for OP_CBRA (normal capturing bracket) and      /* We only need to continue for OP_CBRA (normal capturing bracket) and
1602      OP_BRA (normal non-capturing bracket) because the other variants of these      OP_BRA (normal non-capturing bracket) because the other variants of these
1603      opcodes are all concerned with unlimited repeated groups, which of course      opcodes are all concerned with unlimited repeated groups, which of course
1604      are not of fixed length. They will cause a -1 response from the default      are not of fixed length. */
     case of this switch. */  
1605    
1606      case OP_CBRA:      case OP_CBRA:
1607      case OP_BRA:      case OP_BRA:
1608      case OP_ONCE:      case OP_ONCE:
1609        case OP_ONCE_NC:
1610      case OP_COND:      case OP_COND:
1611      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1612      if (d < 0) return d;      if (d < 0) return d;
# Line 1514  for (;;) Line 1615  for (;;)
1615      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1616      break;      break;
1617    
1618      /* 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 call.
1619      call. If it's ALT it is an alternation in a nested call. If it is      If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1620      END it's the end of the outer call. All can be handled by the same code.      an ALT. If it is END it's the end of the outer call. All can be handled by
1621      Note that we must not include the OP_KETRxxx opcodes here, because they      the same code. Note that we must not include the OP_KETRxxx opcodes here,
1622      all imply an unlimited repeat. */      because they all imply an unlimited repeat. */
1623    
1624      case OP_ALT:      case OP_ALT:
1625      case OP_KET:      case OP_KET:
1626      case OP_END:      case OP_END:
1627        case OP_ACCEPT:
1628        case OP_ASSERT_ACCEPT:
1629      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1630        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1631      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1552  for (;;) Line 1655  for (;;)
1655      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1656      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1657      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1658      /* Fall through */      cc += _pcre_OP_lengths[*cc];
1659        break;
1660    
1661      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1662    
1663      case OP_REVERSE:      case OP_MARK:
1664      case OP_CREF:      case OP_PRUNE_ARG:
1665      case OP_NCREF:      case OP_SKIP_ARG:
1666      case OP_RREF:      case OP_THEN_ARG:
1667      case OP_NRREF:      cc += cc[1] + _pcre_OP_lengths[*cc];
1668      case OP_DEF:      break;
1669    
1670      case OP_CALLOUT:      case OP_CALLOUT:
     case OP_SOD:  
     case OP_SOM:  
     case OP_SET_SOM:  
     case OP_EOD:  
     case OP_EODN:  
1671      case OP_CIRC:      case OP_CIRC:
1672      case OP_CIRCM:      case OP_CIRCM:
1673        case OP_CLOSE:
1674        case OP_COMMIT:
1675        case OP_CREF:
1676        case OP_DEF:
1677      case OP_DOLL:      case OP_DOLL:
1678      case OP_DOLLM:      case OP_DOLLM:
1679        case OP_EOD:
1680        case OP_EODN:
1681        case OP_FAIL:
1682        case OP_NCREF:
1683        case OP_NRREF:
1684      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1685        case OP_PRUNE:
1686        case OP_REVERSE:
1687        case OP_RREF:
1688        case OP_SET_SOM:
1689        case OP_SKIP:
1690        case OP_SOD:
1691        case OP_SOM:
1692        case OP_THEN:
1693      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1694      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
1695      break;      break;
# Line 1594  for (;;) Line 1711  for (;;)
1711      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1712    
1713      case OP_EXACT:      case OP_EXACT:
1714        case OP_EXACTI:
1715        case OP_NOTEXACT:
1716        case OP_NOTEXACTI:
1717      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1718      cc += 4;      cc += 4;
1719  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1614  for (;;) Line 1734  for (;;)
1734      cc += 2;      cc += 2;
1735      /* Fall through */      /* Fall through */
1736    
1737        case OP_HSPACE:
1738        case OP_VSPACE:
1739        case OP_NOT_HSPACE:
1740        case OP_NOT_VSPACE:
1741      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1742      case OP_DIGIT:      case OP_DIGIT:
1743      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1626  for (;;) Line 1750  for (;;)
1750      cc++;      cc++;
1751      break;      break;
1752    
1753      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1754        otherwise \C is coded as OP_ALLANY. */
1755    
1756      case OP_ANYBYTE:      case OP_ANYBYTE:
1757      return -2;      return -2;
# Line 1645  for (;;) Line 1770  for (;;)
1770    
1771      switch (*cc)      switch (*cc)
1772        {        {
1773          case OP_CRPLUS:
1774          case OP_CRMINPLUS:
1775        case OP_CRSTAR:        case OP_CRSTAR:
1776        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1777        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1665  for (;;) Line 1792  for (;;)
1792    
1793      /* Anything else is variable length */      /* Anything else is variable length */
1794    
1795      default:      case OP_ANYNL:
1796        case OP_BRAMINZERO:
1797        case OP_BRAPOS:
1798        case OP_BRAPOSZERO:
1799        case OP_BRAZERO:
1800        case OP_CBRAPOS:
1801        case OP_EXTUNI:
1802        case OP_KETRMAX:
1803        case OP_KETRMIN:
1804        case OP_KETRPOS:
1805        case OP_MINPLUS:
1806        case OP_MINPLUSI:
1807        case OP_MINQUERY:
1808        case OP_MINQUERYI:
1809        case OP_MINSTAR:
1810        case OP_MINSTARI:
1811        case OP_MINUPTO:
1812        case OP_MINUPTOI:
1813        case OP_NOTMINPLUS:
1814        case OP_NOTMINPLUSI:
1815        case OP_NOTMINQUERY:
1816        case OP_NOTMINQUERYI:
1817        case OP_NOTMINSTAR:
1818        case OP_NOTMINSTARI:
1819        case OP_NOTMINUPTO:
1820        case OP_NOTMINUPTOI:
1821        case OP_NOTPLUS:
1822        case OP_NOTPLUSI:
1823        case OP_NOTPOSPLUS:
1824        case OP_NOTPOSPLUSI:
1825        case OP_NOTPOSQUERY:
1826        case OP_NOTPOSQUERYI:
1827        case OP_NOTPOSSTAR:
1828        case OP_NOTPOSSTARI:
1829        case OP_NOTPOSUPTO:
1830        case OP_NOTPOSUPTOI:
1831        case OP_NOTQUERY:
1832        case OP_NOTQUERYI:
1833        case OP_NOTSTAR:
1834        case OP_NOTSTARI:
1835        case OP_NOTUPTO:
1836        case OP_NOTUPTOI:
1837        case OP_PLUS:
1838        case OP_PLUSI:
1839        case OP_POSPLUS:
1840        case OP_POSPLUSI:
1841        case OP_POSQUERY:
1842        case OP_POSQUERYI:
1843        case OP_POSSTAR:
1844        case OP_POSSTARI:
1845        case OP_POSUPTO:
1846        case OP_POSUPTOI:
1847        case OP_QUERY:
1848        case OP_QUERYI:
1849        case OP_REF:
1850        case OP_REFI:
1851        case OP_SBRA:
1852        case OP_SBRAPOS:
1853        case OP_SCBRA:
1854        case OP_SCBRAPOS:
1855        case OP_SCOND:
1856        case OP_SKIPZERO:
1857        case OP_STAR:
1858        case OP_STARI:
1859        case OP_TYPEMINPLUS:
1860        case OP_TYPEMINQUERY:
1861        case OP_TYPEMINSTAR:
1862        case OP_TYPEMINUPTO:
1863        case OP_TYPEPLUS:
1864        case OP_TYPEPOSPLUS:
1865        case OP_TYPEPOSQUERY:
1866        case OP_TYPEPOSSTAR:
1867        case OP_TYPEPOSUPTO:
1868        case OP_TYPEQUERY:
1869        case OP_TYPESTAR:
1870        case OP_TYPEUPTO:
1871        case OP_UPTO:
1872        case OP_UPTOI:
1873      return -1;      return -1;
1874    
1875        /* Catch unrecognized opcodes so that when new ones are added they
1876        are not forgotten, as has happened in the past. */
1877    
1878        default:
1879        return -4;
1880      }      }
1881    }    }
1882  /* Control never gets here */  /* Control never gets here */
# Line 2045  for (code = first_significant_code(code Line 2255  for (code = first_significant_code(code
2255    
2256    if (c == OP_BRA  || c == OP_BRAPOS ||    if (c == OP_BRA  || c == OP_BRAPOS ||
2257        c == OP_CBRA || c == OP_CBRAPOS ||        c == OP_CBRA || c == OP_CBRAPOS ||
2258        c == OP_ONCE || c == OP_COND)        c == OP_ONCE || c == OP_ONCE_NC ||
2259          c == OP_COND)
2260      {      {
2261      BOOL empty_branch;      BOOL empty_branch;
2262      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 3142  for (;; ptr++) Line 3353  for (;; ptr++)
3353    int subfirstbyte;    int subfirstbyte;
3354    int terminator;    int terminator;
3355    int mclength;    int mclength;
3356      int tempbracount;
3357    uschar mcbuffer[8];    uschar mcbuffer[8];
3358    
3359    /* Get next byte in the pattern */    /* Get next byte in the pattern */
# Line 3166  for (;; ptr++) Line 3378  for (;; ptr++)
3378  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
3379      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3380  #endif  #endif
3381      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */      if (code > cd->start_workspace + cd->workspace_size -
3382            WORK_SIZE_SAFETY_MARGIN)                       /* Check for overrun */
3383        {        {
3384        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3385        goto FAILED;        goto FAILED;
# Line 3189  for (;; ptr++) Line 3402  for (;; ptr++)
3402        }        }
3403    
3404      *lengthptr += (int)(code - last_code);      *lengthptr += (int)(code - last_code);
3405      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, (int)(code - last_code),
3406          c));
3407    
3408      /* If "previous" is set and it is not at the start of the work space, move      /* If "previous" is set and it is not at the start of the work space, move
3409      it back to there, in order to avoid filling up the work space. Otherwise,      it back to there, in order to avoid filling up the work space. Otherwise,
# Line 3215  for (;; ptr++) Line 3429  for (;; ptr++)
3429    /* In the real compile phase, just check the workspace used by the forward    /* In the real compile phase, just check the workspace used by the forward
3430    reference list. */    reference list. */
3431    
3432    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)    else if (cd->hwm > cd->start_workspace + cd->workspace_size -
3433               WORK_SIZE_SAFETY_MARGIN)
3434      {      {
3435      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3436      goto FAILED;      goto FAILED;
# Line 3469  for (;; ptr++) Line 3684  for (;; ptr++)
3684    
3685        if (lengthptr != NULL)        if (lengthptr != NULL)
3686          {          {
3687          *lengthptr += class_utf8data - class_utf8data_base;          *lengthptr += (int)(class_utf8data - class_utf8data_base);
3688          class_utf8data = class_utf8data_base;          class_utf8data = class_utf8data_base;
3689          }          }
3690    
# Line 3605  for (;; ptr++) Line 3820  for (;; ptr++)
3820          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3821    
3822          if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */          if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */
3823            else if (-c == ESC_N)            /* \N is not supported in a class */
3824              {
3825              *errorcodeptr = ERR71;
3826              goto FAILED;
3827              }
3828          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3829            {            {
3830            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
# Line 4163  for (;; ptr++) Line 4383  for (;; ptr++)
4383    
4384        /* Now fill in the complete length of the item */        /* Now fill in the complete length of the item */
4385    
4386        PUT(previous, 1, code - previous);        PUT(previous, 1, (int)(code - previous));
4387        break;   /* End of class handling */        break;   /* End of class handling */
4388        }        }
4389  #endif  #endif
# Line 4305  for (;; ptr++) Line 4525  for (;; ptr++)
4525          {          {
4526          uschar *lastchar = code - 1;          uschar *lastchar = code - 1;
4527          while((*lastchar & 0xc0) == 0x80) lastchar--;          while((*lastchar & 0xc0) == 0x80) lastchar--;
4528          c = code - lastchar;            /* Length of UTF-8 character */          c = (int)(code - lastchar);     /* Length of UTF-8 character */
4529          memcpy(utf8_char, lastchar, c); /* Save the char */          memcpy(utf8_char, lastchar, c); /* Save the char */
4530          c |= 0x80;                      /* Flag c as a length */          c |= 0x80;                      /* Flag c as a length */
4531          }          }
# Line 4712  for (;; ptr++) Line 4932  for (;; ptr++)
4932              *lengthptr += delta;              *lengthptr += delta;
4933              }              }
4934    
4935            /* This is compiling for real */            /* This is compiling for real. If there is a set first byte for
4936              the group, and we have not yet set a "required byte", set it. Make
4937              sure there is enough workspace for copying forward references before
4938              doing the copy. */
4939    
4940            else            else
4941              {              {
4942              if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;              if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;
4943    
4944              for (i = 1; i < repeat_min; i++)              for (i = 1; i < repeat_min; i++)
4945                {                {
4946                uschar *hc;                uschar *hc;
4947                uschar *this_hwm = cd->hwm;                uschar *this_hwm = cd->hwm;
4948                memcpy(code, previous, len);                memcpy(code, previous, len);
4949    
4950                  while (cd->hwm > cd->start_workspace + cd->workspace_size -
4951                         WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
4952                    {
4953                    int save_offset = save_hwm - cd->start_workspace;
4954                    int this_offset = this_hwm - cd->start_workspace;
4955                    *errorcodeptr = expand_workspace(cd);
4956                    if (*errorcodeptr != 0) goto FAILED;
4957                    save_hwm = (uschar *)cd->start_workspace + save_offset;
4958                    this_hwm = (uschar *)cd->start_workspace + this_offset;
4959                    }
4960    
4961                for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)                for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
4962                  {                  {
4963                  PUT(cd->hwm, 0, GET(hc, 0) + len);                  PUT(cd->hwm, 0, GET(hc, 0) + len);
# Line 4789  for (;; ptr++) Line 5025  for (;; ptr++)
5025              }              }
5026    
5027            memcpy(code, previous, len);            memcpy(code, previous, len);
5028    
5029              /* Ensure there is enough workspace for forward references before
5030              copying them. */
5031    
5032              while (cd->hwm > cd->start_workspace + cd->workspace_size -
5033                     WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
5034                {
5035                int save_offset = save_hwm - cd->start_workspace;
5036                int this_offset = this_hwm - cd->start_workspace;
5037                *errorcodeptr = expand_workspace(cd);
5038                if (*errorcodeptr != 0) goto FAILED;
5039                save_hwm = (uschar *)cd->start_workspace + save_offset;
5040                this_hwm = (uschar *)cd->start_workspace + this_offset;
5041                }
5042    
5043            for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)            for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
5044              {              {
5045              PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));              PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
# Line 4820  for (;; ptr++) Line 5071  for (;; ptr++)
5071        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
5072        deal with possessive ONCEs specially.        deal with possessive ONCEs specially.
5073    
5074        Otherwise, if the quantifier was possessive, we convert the BRA code to        Otherwise, when we are doing the actual compile phase, check to see
5075        the POS form, and the KET code to KETRPOS. (It turns out to be convenient        whether this group is one that could match an empty string. If so,
5076        at runtime to detect this kind of subpattern at both the start and at the        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
5077        end.) The use of special opcodes makes it possible to reduce greatly the        that runtime checking can be done. [This check is also applied to ONCE
5078        stack usage in pcre_exec(). If the group is preceded by OP_BRAZERO,        groups at runtime, but in a different way.]
5079        convert this to OP_BRAPOSZERO. Then cancel the possessive flag so that  
5080        the default action below, of wrapping everything inside atomic brackets,        Then, if the quantifier was possessive and the bracket is not a
5081        does not happen.        conditional, we convert the BRA code to the POS form, and the KET code to
5082          KETRPOS. (It turns out to be convenient at runtime to detect this kind of
5083        Then, when we are doing the actual compile phase, check to see whether        subpattern at both the start and at the end.) The use of special opcodes
5084        this group is one that could match an empty string. If so, convert the        makes it possible to reduce greatly the stack usage in pcre_exec(). If
5085        initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so that runtime        the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
5086        checking can be done. [This check is also applied to ONCE groups at  
5087        runtime, but in a different way.] */        Then, if the minimum number of matches is 1 or 0, cancel the possessive
5088          flag so that the default action below, of wrapping everything inside
5089          atomic brackets, does not happen. When the minimum is greater than 1,
5090          there will be earlier copies of the group, and so we still have to wrap
5091          the whole thing. */
5092    
5093        else        else
5094          {          {
5095          uschar *ketcode = code - 1 - LINK_SIZE;          uschar *ketcode = code - 1 - LINK_SIZE;
5096          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
5097    
5098          if (*bracode == OP_ONCE && possessive_quantifier) *bracode = OP_BRA;          /* Convert possessive ONCE brackets to non-capturing */
5099          if (*bracode == OP_ONCE)  
5100            if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
5101                possessive_quantifier) *bracode = OP_BRA;
5102    
5103            /* For non-possessive ONCE brackets, all we need to do is to
5104            set the KET. */
5105    
5106            if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
5107            *ketcode = OP_KETRMAX + repeat_type;            *ketcode = OP_KETRMAX + repeat_type;
5108    
5109            /* Handle non-ONCE brackets and possessive ONCEs (which have been
5110            converted to non-capturing above). */
5111    
5112          else          else
5113            {            {
5114            if (possessive_quantifier)            /* In the compile phase, check for empty string matching. */
             {  
             *bracode += 1;                   /* Switch to xxxPOS opcodes */  
             *ketcode = OP_KETRPOS;  
             if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;  
             possessive_quantifier = FALSE;  
             }  
           else *ketcode = OP_KETRMAX + repeat_type;  
5115    
5116            if (lengthptr == NULL)            if (lengthptr == NULL)
5117              {              {
# Line 4868  for (;; ptr++) Line 5127  for (;; ptr++)
5127                }                }
5128              while (*scode == OP_ALT);              while (*scode == OP_ALT);
5129              }              }
5130    
5131              /* Handle possessive quantifiers. */
5132    
5133              if (possessive_quantifier)
5134                {
5135                /* For COND brackets, we wrap the whole thing in a possessively
5136                repeated non-capturing bracket, because we have not invented POS
5137                versions of the COND opcodes. Because we are moving code along, we
5138                must ensure that any pending recursive references are updated. */
5139    
5140                if (*bracode == OP_COND || *bracode == OP_SCOND)
5141                  {
5142                  int nlen = (int)(code - bracode);
5143                  *code = OP_END;
5144                  adjust_recurse(bracode, 1 + LINK_SIZE, utf8, cd, save_hwm);
5145                  memmove(bracode + 1+LINK_SIZE, bracode, nlen);
5146                  code += 1 + LINK_SIZE;
5147                  nlen += 1 + LINK_SIZE;
5148                  *bracode = OP_BRAPOS;
5149                  *code++ = OP_KETRPOS;
5150                  PUTINC(code, 0, nlen);
5151                  PUT(bracode, 1, nlen);
5152                  }
5153    
5154                /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
5155    
5156                else
5157                  {
5158                  *bracode += 1;              /* Switch to xxxPOS opcodes */
5159                  *ketcode = OP_KETRPOS;
5160                  }
5161    
5162                /* If the minimum is zero, mark it as possessive, then unset the
5163                possessive flag when the minimum is 0 or 1. */
5164    
5165                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
5166                if (repeat_min < 2) possessive_quantifier = FALSE;
5167                }
5168    
5169              /* Non-possessive quantifier */
5170    
5171              else *ketcode = OP_KETRMAX + repeat_type;
5172            }            }
5173          }          }
5174        }        }
# Line 4894  for (;; ptr++) Line 5195  for (;; ptr++)
5195      notation is just syntactic sugar, taken from Sun's Java package, but the      notation is just syntactic sugar, taken from Sun's Java package, but the
5196      special opcodes can optimize it.      special opcodes can optimize it.
5197    
5198      Possessively repeated subpatterns have already been handled in the code      Some (but not all) possessively repeated subpatterns have already been
5199      just above, so possessive_quantifier is always FALSE for them at this      completely handled in the code just above. For them, possessive_quantifier
5200      stage.      is always FALSE at this stage.
5201    
5202      Note that the repeated item starts at tempcode, not at previous, which      Note that the repeated item starts at tempcode, not at previous, which
5203      might be the first part of a string whose (former) last char we repeated.      might be the first part of a string whose (former) last char we repeated.
# Line 5386  for (;; ptr++) Line 5687  for (;; ptr++)
5687    
5688          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5689          case CHAR_C:                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
5690          previous_callout = code;  /* Save for later completion */          previous_callout = code;     /* Save for later completion */
5691          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1;    /* Skip one item before completing */
5692          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
5693            {            {
5694            int n = 0;            int n = 0;
# Line 5758  for (;; ptr++) Line 6059  for (;; ptr++)
6059                of the group. Then remember the forward reference. */                of the group. Then remember the forward reference. */
6060    
6061                called = cd->start_code + recno;                called = cd->start_code + recno;
6062                  if (cd->hwm >= cd->start_workspace + cd->workspace_size -
6063                      WORK_SIZE_SAFETY_MARGIN)
6064                    {
6065                    *errorcodeptr = expand_workspace(cd);
6066                    if (*errorcodeptr != 0) goto FAILED;
6067                    }
6068                PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));                PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
6069                }                }
6070    
# Line 5778  for (;; ptr++) Line 6085  for (;; ptr++)
6085                }                }
6086              }              }
6087    
6088            /* Insert the recursion/subroutine item. */            /* Insert the recursion/subroutine item. It does not have a set first
6089              byte (relevant if it is repeated, because it will then be wrapped
6090              with ONCE brackets). */
6091    
6092            *code = OP_RECURSE;            *code = OP_RECURSE;
6093            PUT(code, 1, (int)(called - cd->start_code));            PUT(code, 1, (int)(called - cd->start_code));
6094            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
6095              groupsetfirstbyte = FALSE;
6096            }            }
6097    
6098          /* Can't determine a first byte now */          /* Can't determine a first byte now */
# Line 5906  for (;; ptr++) Line 6216  for (;; ptr++)
6216      *code = bravalue;      *code = bravalue;
6217      tempcode = code;      tempcode = code;
6218      tempreqvary = cd->req_varyopt;        /* Save value before bracket */      tempreqvary = cd->req_varyopt;        /* Save value before bracket */
6219        tempbracount = cd->bracount;          /* Save value before bracket */
6220      length_prevgroup = 0;                 /* Initialize for pre-compile phase */      length_prevgroup = 0;                 /* Initialize for pre-compile phase */
6221    
6222      if (!compile_regex(      if (!compile_regex(
# Line 5928  for (;; ptr++) Line 6239  for (;; ptr++)
6239           ))           ))
6240        goto FAILED;        goto FAILED;
6241    
6242        /* If this was an atomic group and there are no capturing groups within it,
6243        generate OP_ONCE_NC instead of OP_ONCE. */
6244    
6245        if (bravalue == OP_ONCE && cd->bracount <= tempbracount)
6246          *code = OP_ONCE_NC;
6247    
6248      if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)      if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
6249        cd->assert_depth -= 1;        cd->assert_depth -= 1;
6250    
# Line 6134  for (;; ptr++) Line 6451  for (;; ptr++)
6451    
6452          if (ptr[1] != CHAR_PLUS && ptr[1] != CHAR_MINUS)          if (ptr[1] != CHAR_PLUS && ptr[1] != CHAR_MINUS)
6453            {            {
6454            BOOL isnumber = TRUE;            BOOL is_a_number = TRUE;
6455            for (p = ptr + 1; *p != 0 && *p != terminator; p++)            for (p = ptr + 1; *p != 0 && *p != terminator; p++)
6456              {              {
6457              if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;              if ((cd->ctypes[*p] & ctype_digit) == 0) is_a_number = FALSE;
6458              if ((cd->ctypes[*p] & ctype_word) == 0) break;              if ((cd->ctypes[*p] & ctype_word) == 0) break;
6459              }              }
6460            if (*p != terminator)            if (*p != terminator)
# Line 6145  for (;; ptr++) Line 6462  for (;; ptr++)
6462              *errorcodeptr = ERR57;              *errorcodeptr = ERR57;
6463              break;              break;
6464              }              }
6465            if (isnumber)            if (is_a_number)
6466              {              {
6467              ptr++;              ptr++;
6468              goto HANDLE_NUMERICAL_RECURSION;              goto HANDLE_NUMERICAL_RECURSION;
# Line 6257  for (;; ptr++) Line 6574  for (;; ptr++)
6574            }            }
6575          else          else
6576  #endif  #endif
6577            /* In non-UTF-8 mode, we turn \C into OP_ALLANY instead of OP_ANYBYTE
6578            so that it works in DFA mode and in lookbehinds. */
6579    
6580            {            {
6581            previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;            previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;
6582            *code++ = -c;            *code++ = (!utf8 && c == -ESC_C)? OP_ALLANY : -c;
6583            }            }
6584          }          }
6585        continue;        continue;
# Line 6551  for (;;) Line 6871  for (;;)
6871          }          }
6872        else if (fixed_length < 0)        else if (fixed_length < 0)
6873          {          {
6874          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 :
6875                            (fixed_length == -4)? ERR70: ERR25;
6876          *ptrptr = ptr;          *ptrptr = ptr;
6877          return FALSE;          return FALSE;
6878          }          }
# Line 6726  do { Line 7047  do {
7047    
7048     /* Other brackets */     /* Other brackets */
7049    
7050     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC ||
7051                op == OP_COND)
7052       {       {
7053       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
7054       }       }
# Line 6830  do { Line 7152  do {
7152    
7153     /* Other brackets */     /* Other brackets */
7154    
7155     else if (op == OP_ASSERT || op == OP_ONCE)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC)
7156       {       {
7157       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
7158       }       }
# Line 6900  do { Line 7222  do {
7222       case OP_SCBRAPOS:       case OP_SCBRAPOS:
7223       case OP_ASSERT:       case OP_ASSERT:
7224       case OP_ONCE:       case OP_ONCE:
7225         case OP_ONCE_NC:
7226       case OP_COND:       case OP_COND:
7227       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)
7228         return -1;         return -1;
# Line 6992  compile_data *cd = &compile_block; Line 7315  compile_data *cd = &compile_block;
7315  computing the amount of memory that is needed. Compiled items are thrown away  computing the amount of memory that is needed. Compiled items are thrown away
7316  as soon as possible, so that a fairly large buffer should be sufficient for  as soon as possible, so that a fairly large buffer should be sufficient for
7317  this purpose. The same space is used in the second phase for remembering where  this purpose. The same space is used in the second phase for remembering where
7318  to fill in forward references to subpatterns. */  to fill in forward references to subpatterns. That may overflow, in which case
7319    new memory is obtained from malloc(). */
7320    
7321  uschar cworkspace[COMPILE_WORK_SIZE];  uschar cworkspace[COMPILE_WORK_SIZE];
7322    
# Line 7182  cd->bracount = cd->final_bracount = 0; Line 7506  cd->bracount = cd->final_bracount = 0;
7506  cd->names_found = 0;  cd->names_found = 0;
7507  cd->name_entry_size = 0;  cd->name_entry_size = 0;
7508  cd->name_table = NULL;  cd->name_table = NULL;
 cd->start_workspace = cworkspace;  
7509  cd->start_code = cworkspace;  cd->start_code = cworkspace;
7510  cd->hwm = cworkspace;  cd->hwm = cworkspace;
7511    cd->start_workspace = cworkspace;
7512    cd->workspace_size = COMPILE_WORK_SIZE;
7513  cd->start_pattern = (const uschar *)pattern;  cd->start_pattern = (const uschar *)pattern;
7514  cd->end_pattern = (const uschar *)(pattern + strlen(pattern));  cd->end_pattern = (const uschar *)(pattern + strlen(pattern));
7515  cd->req_varyopt = 0;  cd->req_varyopt = 0;
# Line 7219  externally provided function. Integer ov Line 7544  externally provided function. Integer ov
7544  because nowadays we limit the maximum value of cd->names_found and  because nowadays we limit the maximum value of cd->names_found and
7545  cd->name_entry_size. */  cd->name_entry_size. */
7546    
7547  size = length + sizeof(real_pcre) + cd->names_found * (cd->name_entry_size + 3);  size = length + sizeof(real_pcre) + cd->names_found * cd->name_entry_size;
7548  re = (real_pcre *)(pcre_malloc)(size);  re = (real_pcre *)(pcre_malloc)(size);
7549    
7550  if (re == NULL)  if (re == NULL)
# Line 7262  cd->names_found = 0; Line 7587  cd->names_found = 0;
7587  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
7588  codestart = cd->name_table + re->name_entry_size * re->name_count;  codestart = cd->name_table + re->name_entry_size * re->name_count;
7589  cd->start_code = codestart;  cd->start_code = codestart;
7590  cd->hwm = cworkspace;  cd->hwm = (uschar *)(cd->start_workspace);
7591  cd->req_varyopt = 0;  cd->req_varyopt = 0;
7592  cd->had_accept = FALSE;  cd->had_accept = FALSE;
7593  cd->check_lookbehind = FALSE;  cd->check_lookbehind = FALSE;
# Line 7296  if debugging, leave the test till after Line 7621  if debugging, leave the test till after
7621  if (code - codestart > length) errorcode = ERR23;  if (code - codestart > length) errorcode = ERR23;
7622  #endif  #endif
7623    
7624  /* Fill in any forward references that are required. */  /* Fill in any forward references that are required. There may be repeated
7625    references; optimize for them, as searching a large regex takes time. */
7626    
7627  while (errorcode == 0 && cd->hwm > cworkspace)  if (cd->hwm > cd->start_workspace)
7628    {    {
7629    int offset, recno;    int prev_recno = -1;
7630    const uschar *groupptr;    const uschar *groupptr = NULL;
7631    cd->hwm -= LINK_SIZE;    while (errorcode == 0 && cd->hwm > cd->start_workspace)
7632    offset = GET(cd->hwm, 0);      {
7633    recno = GET(codestart, offset);      int offset, recno;
7634    groupptr = _pcre_find_bracket(codestart, utf8, recno);      cd->hwm -= LINK_SIZE;
7635    if (groupptr == NULL) errorcode = ERR53;      offset = GET(cd->hwm, 0);
7636      else PUT(((uschar *)codestart), offset, (int)(groupptr - codestart));      recno = GET(codestart, offset);
7637        if (recno != prev_recno)
7638          {
7639          groupptr = _pcre_find_bracket(codestart, utf8, recno);
7640          prev_recno = recno;
7641          }
7642        if (groupptr == NULL) errorcode = ERR53;
7643          else PUT(((uschar *)codestart), offset, (int)(groupptr - codestart));
7644        }
7645    }    }
7646    
7647    /* If the workspace had to be expanded, free the new memory. */
7648    
7649    if (cd->workspace_size > COMPILE_WORK_SIZE)
7650      (pcre_free)((void *)cd->start_workspace);
7651    
7652  /* Give an error if there's back reference to a non-existent capturing  /* Give an error if there's back reference to a non-existent capturing
7653  subpattern. */  subpattern. */
7654    
# Line 7348  if (cd->check_lookbehind) Line 7687  if (cd->check_lookbehind)
7687        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
7688        if (fixed_length < 0)        if (fixed_length < 0)
7689          {          {
7690          errorcode = (fixed_length == -2)? ERR36 : ERR25;          errorcode = (fixed_length == -2)? ERR36 :
7691                        (fixed_length == -4)? ERR70 : ERR25;
7692          break;          break;
7693          }          }
7694        PUT(cc, 1, fixed_length);        PUT(cc, 1, fixed_length);

Legend:
Removed from v.721  
changed lines
  Added in v.807

  ViewVC Help
Powered by ViewVC 1.1.5