/[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 1360 by ph10, Tue Sep 3 10:25:39 2013 UTC revision 1365 by ph10, Sun Oct 6 18:33:56 2013 UTC
# Line 115  kicks in at the same number of forward r Line 115  kicks in at the same number of forward r
115  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
116  #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)  #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
117    
118  /* This value determines the size of the initial vector that is used for  /* This value determines the size of the initial vector that is used for
119  remembering named groups during the pre-compile. It is allocated on the stack,  remembering named groups during the pre-compile. It is allocated on the stack,
120  but if it is too small, it is expanded using malloc(), in a similar way to the  but if it is too small, it is expanded using malloc(), in a similar way to the
121  workspace. The value is the number of slots in the list. */  workspace. The value is the number of slots in the list. */
122    
123  #define NAMED_GROUP_LIST_SIZE  20  #define NAMED_GROUP_LIST_SIZE  20
# Line 655  static const pcre_uint8 ebcdic_chartab[] Line 655  static const pcre_uint8 ebcdic_chartab[]
655  #endif  #endif
656    
657    
658    /* This table is used to check whether auto-possessification is possible
659    between adjacent character-type opcodes. The left-hand (repeated) opcode is
660    used to select the row, and the right-hand opcode is use to select the column.
661    A value of 1 means that auto-possessification is OK. For example, the second
662    value in the first row means that \D+\d can be turned into \D++\d.
663    
664    The Unicode property types (\P and \p) have to be present to fill out the table
665    because of what their opcode values are, but the table values should always be
666    zero because property types are handled separately in the code. The last four
667    columns apply to items that cannot be repeated, so there is no need to have
668    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
669    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
670    
671    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
672    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
673    
674    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
675    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
676      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
677      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
678      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
679      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
680      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
681      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
682      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
683      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
684      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
685      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
686      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
687      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
688      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
689      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
690      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
691      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
692      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
693    };
694    
695    
696    /* This table is used to check whether auto-possessification is possible
697    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
698    left-hand (repeated) opcode is used to select the row, and the right-hand
699    opcode is used to select the column. The values are as follows:
700    
701      0   Always return FALSE (never auto-possessify)
702      1   Character groups are distinct (possessify if both are OP_PROP)
703      2   Check character categories in the same group (general or particular)
704      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
705    
706      4   Check left general category vs right particular category
707      5   Check right general category vs left particular category
708    
709      6   Left alphanum vs right general category
710      7   Left space vs right general category
711      8   Left word vs right general category
712    
713      9   Right alphanum vs left general category
714     10   Right space vs left general category
715     11   Right word vs left general category
716    
717     12   Left alphanum vs right particular category
718     13   Left space vs right particular category
719     14   Left word vs right particular category
720    
721     15   Right alphanum vs left particular category
722     16   Right space vs left particular category
723     17   Right word vs left particular category
724    */
725    
726    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
727    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
728      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
729      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
730      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
731      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
732      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
733      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
734      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
735      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
736      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
737      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
738      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
739    };
740    
741    /* This table is used to check whether auto-possessification is possible
742    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
743    specifies a general category and the other specifies a particular category. The
744    row is selected by the general category and the column by the particular
745    category. The value is 1 if the particular category is not part of the general
746    category. */
747    
748    static const pcre_uint8 catposstab[7][30] = {
749    /* Cc Cf Cn Co Cs Ll Lm Lo Lt Lu Mc Me Mn Nd Nl No Pc Pd Pe Pf Pi Po Ps Sc Sk Sm So Zl Zp Zs */
750      { 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* C */
751      { 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* L */
752      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* M */
753      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* N */
754      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1 },  /* P */
755      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1 },  /* S */
756      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0 }   /* Z */
757    };
758    
759    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
760    a general or particular category. The properties in each row are those
761    that apply to the character set in question. Duplication means that a little
762    unnecessary work is done when checking, but this keeps things much simpler
763    because they can all use the same code. For more details see the comment where
764    this table is used.
765    
766    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
767    "space", but from Perl 5.18 it's included, so both categories are treated the
768    same here. */
769    
770    static const pcre_uint8 posspropstab[3][4] = {
771      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
772      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
773      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
774    };
775    
776    
777    
778  /*************************************************  /*************************************************
779  *            Find an error text                  *  *            Find an error text                  *
# Line 682  return s; Line 801  return s;
801  }  }
802    
803    
804    
805  /*************************************************  /*************************************************
806  *           Expand the workspace                 *  *           Expand the workspace                 *
807  *************************************************/  *************************************************/
# Line 1199  if ((options & PCRE_UCP) != 0 && escape Line 1319  if ((options & PCRE_UCP) != 0 && escape
1319  return escape;  return escape;
1320  }  }
1321    
1322    
1323    
1324  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
1325  /*************************************************  /*************************************************
1326  *               Handle \P and \p                 *  *               Handle \P and \p                 *
# Line 1296  return FALSE; Line 1418  return FALSE;
1418    
1419    
1420    
   
1421  /*************************************************  /*************************************************
1422  *         Read repeat counts                     *  *         Read repeat counts                     *
1423  *************************************************/  *************************************************/
# Line 1403  for (;;) Line 1524  for (;;)
1524    
1525      case OP_CALLOUT:      case OP_CALLOUT:
1526      case OP_CREF:      case OP_CREF:
1527      case OP_NCREF:      case OP_DNCREF:
1528      case OP_RREF:      case OP_RREF:
1529      case OP_NRREF:      case OP_DNRREF:
1530      case OP_DEF:      case OP_DEF:
1531      code += PRIV(OP_lengths)[*code];      code += PRIV(OP_lengths)[*code];
1532      break;      break;
# Line 1419  for (;;) Line 1540  for (;;)
1540    
1541    
1542    
   
1543  /*************************************************  /*************************************************
1544  *        Find the fixed length of a branch       *  *        Find the fixed length of a branch       *
1545  *************************************************/  *************************************************/
# Line 1543  for (;;) Line 1663  for (;;)
1663      case OP_COMMIT:      case OP_COMMIT:
1664      case OP_CREF:      case OP_CREF:
1665      case OP_DEF:      case OP_DEF:
1666        case OP_DNCREF:
1667        case OP_DNRREF:
1668      case OP_DOLL:      case OP_DOLL:
1669      case OP_DOLLM:      case OP_DOLLM:
1670      case OP_EOD:      case OP_EOD:
1671      case OP_EODN:      case OP_EODN:
1672      case OP_FAIL:      case OP_FAIL:
     case OP_NCREF:  
     case OP_NRREF:  
1673      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1674      case OP_PRUNE:      case OP_PRUNE:
1675      case OP_REVERSE:      case OP_REVERSE:
# Line 1722  for (;;) Line 1842  for (;;)
1842      case OP_QUERYI:      case OP_QUERYI:
1843      case OP_REF:      case OP_REF:
1844      case OP_REFI:      case OP_REFI:
1845        case OP_DNREF:
1846        case OP_DNREFI:
1847      case OP_SBRA:      case OP_SBRA:
1848      case OP_SBRAPOS:      case OP_SBRAPOS:
1849      case OP_SCBRA:      case OP_SCBRA:
# Line 1758  for (;;) Line 1880  for (;;)
1880    
1881    
1882    
   
1883  /*************************************************  /*************************************************
1884  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
1885  *************************************************/  *************************************************/
# Line 2460  return TRUE; Line 2581  return TRUE;
2581    
2582    
2583  /*************************************************  /*************************************************
2584  *           Check for POSIX class syntax         *  *        Base opcode of repeated opcodes         *
2585  *************************************************/  *************************************************/
2586    
2587  /* This function is called when the sequence "[:" or "[." or "[=" is  /* Returns the base opcode for repeated single character type opcodes. If the
2588  encountered in a character class. It checks whether this is followed by a  opcode is not a repeated character type, it returns with the original value.
 sequence of characters terminated by a matching ":]" or ".]" or "=]". If we  
 reach an unescaped ']' without the special preceding character, return FALSE.  
2589    
2590  Originally, this function only recognized a sequence of letters between the  Arguments:  c opcode
2591  terminators, but it seems that Perl recognizes any sequence of characters,  Returns:    base opcode for the type
2592  though of course unknown POSIX names are subsequently rejected. Perl gives an  */
 "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE  
 didn't consider this to be a POSIX class. Likewise for [:1234:].  
2593    
2594  The problem in trying to be exactly like Perl is in the handling of escapes. We  static pcre_uchar
2595  have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX  get_repeat_base(pcre_uchar c)
2596  class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code  {
2597  below handles the special case of \], but does not try to do any other escape  return (c > OP_TYPEPOSUPTO)? c :
2598  processing. This makes it different from Perl for cases such as [:l\ower:]         (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2599  where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize         (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2600  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,         (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2601  I think.         (c >= OP_STARI)?      OP_STARI :
2602                                 OP_STAR;
2603    }
2604    
 A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.  
 It seems that the appearance of a nested POSIX class supersedes an apparent  
 external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or  
 a digit.  
2605    
2606  In Perl, unescaped square brackets may also appear as part of class names. For  
2607  example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for  #ifdef SUPPORT_UCP
2608  [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not  /*************************************************
2609  seem right at all. PCRE does not allow closing square brackets in POSIX class  *        Check a character and a property        *
2610  names.  *************************************************/
2611    
2612    /* This function is called by check_auto_possessive() when a property item
2613    is adjacent to a fixed character.
2614    
2615  Arguments:  Arguments:
2616    ptr      pointer to the initial [    c            the character
2617    endptr   where to return the end pointer    ptype        the property type
2618      pdata        the data for the type
2619      negated      TRUE if it's a negated property (\P or \p{^)
2620    
2621  Returns:   TRUE or FALSE  Returns:       TRUE if auto-possessifying is OK
2622  */  */
2623    
2624  static BOOL  static BOOL
2625  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2626      BOOL negated)
2627  {  {
2628  pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */  const pcre_uint32 *p;
2629  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  const ucd_record *prop = GET_UCD(c);
2630  for (++ptr; *ptr != CHAR_NULL; ptr++)  
2631    switch(ptype)
2632    {    {
2633    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)    case PT_LAMP:
2634      ptr++;    return (prop->chartype == ucp_Lu ||
2635    else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;            prop->chartype == ucp_Ll ||
2636    else            prop->chartype == ucp_Lt) == negated;
2637    
2638      case PT_GC:
2639      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2640    
2641      case PT_PC:
2642      return (pdata == prop->chartype) == negated;
2643    
2644      case PT_SC:
2645      return (pdata == prop->script) == negated;
2646    
2647      /* These are specials */
2648    
2649      case PT_ALNUM:
2650      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2651              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2652    
2653      /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2654      means that Perl space and POSIX space are now identical. PCRE was changed
2655      at release 8.34. */
2656    
2657      case PT_SPACE:    /* Perl space */
2658      case PT_PXSPACE:  /* POSIX space */
2659      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2660              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2661              c == CHAR_FF || c == CHAR_CR)
2662              == negated;
2663    
2664      case PT_WORD:
2665      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2666              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2667              c == CHAR_UNDERSCORE) == negated;
2668    
2669      case PT_CLIST:
2670      p = PRIV(ucd_caseless_sets) + prop->caseset;
2671      for (;;)
2672      {      {
2673      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      if (c < *p) return !negated;
2674        {      if (c == *p++) return negated;
       *endptr = ptr;  
       return TRUE;  
       }  
     if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&  
          (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||  
           ptr[1] == CHAR_EQUALS_SIGN) &&  
         check_posix_syntax(ptr, endptr))  
       return FALSE;  
2675      }      }
2676      break;  /* Control never reaches here */
2677    }    }
2678    
2679  return FALSE;  return FALSE;
2680  }  }
2681    #endif  /* SUPPORT_UCP */
2682    
2683    
2684    
2685  /*************************************************  /*************************************************
2686  *          Check POSIX class name                *  *        Fill the character property list        *
2687  *************************************************/  *************************************************/
2688    
2689  /* This function is called to check the name given in a POSIX-style class entry  /* Checks whether the code points to an opcode that can take part in auto-
2690  such as [:alnum:].  possessification, and if so, fills a list with its properties.
2691    
2692  Arguments:  Arguments:
2693    ptr        points to the first letter    code        points to start of expression
2694    len        the length of the name    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2695      fcc         points to case-flipping table
2696      list        points to output list
2697                  list[0] will be filled with the opcode
2698                  list[1] will be non-zero if this opcode
2699                    can match an empty character string
2700                  list[2..7] depends on the opcode
2701    
2702  Returns:     a value representing the name, or -1 if unknown  Returns:      points to the start of the next opcode if *code is accepted
2703                  NULL if *code is not accepted
2704  */  */
2705    
2706  static int  static const pcre_uchar *
2707  check_posix_name(const pcre_uchar *ptr, int len)  get_chr_property_list(const pcre_uchar *code, BOOL utf,
2708      const pcre_uint8 *fcc, pcre_uint32 *list)
2709  {  {
2710  const char *pn = posix_names;  pcre_uchar c = *code;
2711  register int yield = 0;  const pcre_uchar *end;
2712  while (posix_name_lengths[yield] != 0)  const pcre_uint32 *clist_src;
2713    pcre_uint32 *clist_dest;
2714    pcre_uint32 chr;
2715    pcre_uchar base;
2716    
2717    list[0] = c;
2718    list[1] = FALSE;
2719    code++;
2720    
2721    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2722    {    {
2723    if (len == posix_name_lengths[yield] &&    base = get_repeat_base(c);
2724      STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;    c -= (base - OP_STAR);
   pn += posix_name_lengths[yield] + 1;  
   yield++;  
   }  
 return -1;  
 }  
2725    
2726      if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2727        code += IMM2_SIZE;
2728    
2729  /*************************************************    list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
 *    Adjust OP_RECURSE items in repeated group   *  
 *************************************************/  
2730    
2731  /* OP_RECURSE items contain an offset from the start of the regex to the group    switch(base)
2732  that is referenced. This means that groups can be replicated for fixed      {
2733  repetition simply by copying (because the recursion is allowed to refer to      case OP_STAR:
2734  earlier groups that are outside the current group). However, when a group is      list[0] = OP_CHAR;
2735  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is      break;
 inserted before it, after it has been compiled. This means that any OP_RECURSE  
 items within it that refer to the group itself or any contained groups have to  
 have their offsets adjusted. That one of the jobs of this function. Before it  
 is called, the partially compiled regex must be temporarily terminated with  
 OP_END.  
2736    
2737  This function has been extended with the possibility of forward references for      case OP_STARI:
2738  recursions and subroutine calls. It must also check the list of such references      list[0] = OP_CHARI;
2739  for the group we are dealing with. If it finds that one of the recursions in      break;
 the current group is on this list, it adjusts the offset in the list, not the  
 value in the reference (which is a group number).  
2740    
2741  Arguments:      case OP_NOTSTAR:
2742    group      points to the start of the group      list[0] = OP_NOT;
2743    adjust     the amount by which the group is to be moved      break;
   utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode  
   cd         contains pointers to tables etc.  
   save_hwm   the hwm forward reference pointer at the start of the group  
2744    
2745  Returns:     nothing      case OP_NOTSTARI:
2746  */      list[0] = OP_NOTI;
2747        break;
2748    
2749  static void      case OP_TYPESTAR:
2750  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,      list[0] = *code;
2751    pcre_uchar *save_hwm)      code++;
2752  {      break;
2753  pcre_uchar *ptr = group;      }
2754      c = list[0];
2755      }
2756    
2757  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)  switch(c)
2758    {    {
2759    int offset;    case OP_NOT_DIGIT:
2760    pcre_uchar *hc;    case OP_DIGIT:
2761      case OP_NOT_WHITESPACE:
2762      case OP_WHITESPACE:
2763      case OP_NOT_WORDCHAR:
2764      case OP_WORDCHAR:
2765      case OP_ANY:
2766      case OP_ALLANY:
2767      case OP_ANYNL:
2768      case OP_NOT_HSPACE:
2769      case OP_HSPACE:
2770      case OP_NOT_VSPACE:
2771      case OP_VSPACE:
2772      case OP_EXTUNI:
2773      case OP_EODN:
2774      case OP_EOD:
2775      case OP_DOLL:
2776      case OP_DOLLM:
2777      return code;
2778    
2779    /* See if this recursion is on the forward reference list. If so, adjust the    case OP_CHAR:
2780    reference. */    case OP_NOT:
2781      GETCHARINCTEST(chr, code);
2782      list[2] = chr;
2783      list[3] = NOTACHAR;
2784      return code;
2785    
2786    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    case OP_CHARI:
2787      case OP_NOTI:
2788      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2789      GETCHARINCTEST(chr, code);
2790      list[2] = chr;
2791    
2792    #ifdef SUPPORT_UCP
2793      if (chr < 128 || (chr < 256 && !utf))
2794        list[3] = fcc[chr];
2795      else
2796        list[3] = UCD_OTHERCASE(chr);
2797    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2798      list[3] = (chr < 256) ? fcc[chr] : chr;
2799    #else
2800      list[3] = fcc[chr];
2801    #endif
2802    
2803      /* The othercase might be the same value. */
2804    
2805      if (chr == list[3])
2806        list[3] = NOTACHAR;
2807      else
2808        list[4] = NOTACHAR;
2809      return code;
2810    
2811    #ifdef SUPPORT_UCP
2812      case OP_PROP:
2813      case OP_NOTPROP:
2814      if (code[0] != PT_CLIST)
2815      {      {
2816      offset = (int)GET(hc, 0);      list[2] = code[0];
2817      if (cd->start_code + offset == ptr + 1)      list[3] = code[1];
2818        {      return code + 2;
       PUT(hc, 0, offset + adjust);  
       break;  
       }  
2819      }      }
2820    
2821    /* Otherwise, adjust the recursion offset if it's after the start of this    /* Convert only if we have anough space. */
   group. */  
2822    
2823    if (hc >= cd->hwm)    clist_src = PRIV(ucd_caseless_sets) + code[1];
2824      clist_dest = list + 2;
2825      code += 2;
2826    
2827      do {
2828         /* Early return if there is not enough space. */
2829         if (clist_dest >= list + 8)
2830           {
2831           list[2] = code[0];
2832           list[3] = code[1];
2833           return code;
2834           }
2835         *clist_dest++ = *clist_src;
2836         }
2837       while(*clist_src++ != NOTACHAR);
2838    
2839      /* Enough space to store all characters. */
2840    
2841      list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
2842      return code;
2843    #endif
2844    
2845      case OP_NCLASS:
2846      case OP_CLASS:
2847    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2848      case OP_XCLASS:
2849    
2850      if (c == OP_XCLASS)
2851        end = code + GET(code, 0);
2852      else
2853    #endif
2854        end = code + 32 / sizeof(pcre_uchar);
2855    
2856      switch(*end)
2857      {      {
2858      offset = (int)GET(ptr, 1);      case OP_CRSTAR:
2859      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      case OP_CRMINSTAR:
2860      }      case OP_CRQUERY:
2861        case OP_CRMINQUERY:
2862        list[1] = TRUE;
2863        end++;
2864        break;
2865    
2866    ptr += 1 + LINK_SIZE;      case OP_CRRANGE:
2867        case OP_CRMINRANGE:
2868        list[1] = (GET2(end, 1) == 0);
2869        end += 1 + 2 * IMM2_SIZE;
2870        break;
2871        }
2872      list[2] = end - code;
2873      return end;
2874    }    }
2875    return NULL;    /* Opcode not accepted */
2876  }  }
2877    
2878    
2879    
2880  /*************************************************  /*************************************************
2881  *        Insert an automatic callout point       *  *    Scan further character sets for match       *
2882  *************************************************/  *************************************************/
2883    
2884  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert  /* Checks whether the base and the current opcode have a common character, in
2885  callout points before each pattern item.  which case the base cannot be possessified.
2886    
2887  Arguments:  Arguments:
2888    code           current code pointer    code        points to the byte code
2889    ptr            current pattern pointer    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2890    cd             pointers to tables etc    cd          static compile data
2891      base_list   the data list of the base opcode
2892    
2893  Returns:         new code pointer  Returns:      TRUE if the auto-possessification is possible
2894  */  */
2895    
2896  static pcre_uchar *  static BOOL
2897  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)  compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
2898      const pcre_uint32* base_list)
2899  {  {
2900  *code++ = OP_CALLOUT;  pcre_uchar c;
2901  *code++ = 255;  pcre_uint32 list[8];
2902  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */  const pcre_uint32* chr_ptr;
2903  PUT(code, LINK_SIZE, 0);                       /* Default length */  const pcre_uint32* ochr_ptr;
2904  return code + 2 * LINK_SIZE;  const pcre_uint32* list_ptr;
2905  }  pcre_uint32 chr;
2906    
2907    for(;;)
2908      {
2909      c = *code;
2910    
2911      /* Skip over callouts */
2912    
2913  /*************************************************    if (c == OP_CALLOUT)
2914  *         Complete a callout item                *      {
2915  *************************************************/      code += PRIV(OP_lengths)[c];
2916        continue;
2917        }
2918    
2919  /* A callout item contains the length of the next item in the pattern, which    if (c == OP_ALT)
2920  we can't fill in till after we have reached the relevant point. This is used      {
2921  for both automatic and manual callouts.      do code += GET(code, 1); while (*code == OP_ALT);
2922        c = *code;
2923        }
2924    
2925  Arguments:    switch(c)
2926    previous_callout   points to previous callout item      {
2927    ptr                current pattern pointer      case OP_END:
2928    cd                 pointers to tables etc      /* TRUE only in greedy case. The non-greedy case could be replaced by an
2929        OP_EXACT, but it is probably not worth it. (And note that OP_EXACT uses
2930        more memory, which we cannot get at this stage.) */
2931    
2932  Returns:             nothing      return base_list[1] != 0;
 */  
2933    
2934  static void      case OP_KET:
2935  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)      /* If the bracket is capturing, and referenced by an OP_RECURSE, the
2936  {      non-greedy case cannot be converted to a possessive form. We do not test
2937  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));      the bracket type at the moment, but we might do it in the future to improve
2938  PUT(previous_callout, 2 + LINK_SIZE, length);      this condition. (But note that recursive calls are always atomic.) */
 }  
2939    
2940        if (base_list[1] == 0) return FALSE;
2941        code += PRIV(OP_lengths)[c];
2942        continue;
2943        }
2944    
2945      /* Check for a supported opcode, and load its properties. */
2946    
2947  #ifdef SUPPORT_UCP    code = get_chr_property_list(code, utf, cd->fcc, list);
2948  /*************************************************    if (code == NULL) return FALSE;    /* Unsupported */
 *           Get othercase range                  *  
 *************************************************/  
2949    
2950  /* This function is passed the start and end of a class range, in UTF-8 mode    /* If either opcode is a small character list, set pointers for comparing
2951  with UCP support. It searches up the characters, looking for ranges of    characters from that list with another list, or with a property. */
 characters in the "other" case. Each call returns the next one, updating the  
 start address. A character with multiple other cases is returned on its own  
 with a special return value.  
2952    
2953  Arguments:    if (base_list[0] == OP_CHAR)
2954    cptr        points to starting character value; updated      {
2955    d           end value      chr_ptr = base_list + 2;
2956    ocptr       where to put start of othercase range      list_ptr = list;
2957    odptr       where to put end of othercase range      }
2958      else if (list[0] == OP_CHAR)
2959        {
2960        chr_ptr = list + 2;
2961        list_ptr = base_list;
2962        }
2963    
2964  Yield:        -1 when no more    /* Some property combinations also acceptable. Unicode property opcodes are
2965                 0 when a range is returned    processed specially; the rest can be handled with a lookup table. */
               >0 the CASESET offset for char with multiple other cases  
                 in this case, ocptr contains the original  
 */  
2966    
2967  static int    else
2968  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,      {
2969    pcre_uint32 *odptr)      pcre_uint32 leftop, rightop;
 {  
 pcre_uint32 c, othercase, next;  
 unsigned int co;  
2970    
2971  /* Find the first character that has an other case. If it has multiple other      if (list[1] != 0) return FALSE;   /* Must match at least one character */
2972  cases, return its case offset value. */      leftop = base_list[0];
2973        rightop = list[0];
2974    
2975  for (c = *cptr; c <= d; c++)  #ifdef SUPPORT_UCP
2976    {      if (leftop == OP_PROP || leftop == OP_NOTPROP)
2977    if ((co = UCD_CASESET(c)) != 0)        {
2978      {        if (rightop == OP_EOD) return TRUE;
2979      *ocptr = c++;   /* Character that has the set */        if (rightop == OP_PROP || rightop == OP_NOTPROP)
2980      *cptr = c;      /* Rest of input range */          {
2981      return (int)co;          int n;
2982            const pcre_uint8 *p;
2983            BOOL same = leftop == rightop;
2984            BOOL lisprop = leftop == OP_PROP;
2985            BOOL risprop = rightop == OP_PROP;
2986            BOOL bothprop = lisprop && risprop;
2987    
2988            /* There's a table that specifies how each combination is to be
2989            processed:
2990              0   Always return FALSE (never auto-possessify)
2991              1   Character groups are distinct (possessify if both are OP_PROP)
2992              2   Check character categories in the same group (general or particular)
2993              3   Return TRUE if the two opcodes are not the same
2994              ... see comments below
2995            */
2996    
2997            n = propposstab[base_list[2]][list[2]];
2998            switch(n)
2999              {
3000              case 0: return FALSE;
3001              case 1: return bothprop;
3002              case 2: return (base_list[3] == list[3]) != same;
3003              case 3: return !same;
3004    
3005              case 4:  /* Left general category, right particular category */
3006              return risprop && catposstab[base_list[3]][list[3]] == same;
3007    
3008              case 5:  /* Right general category, left particular category */
3009              return lisprop && catposstab[list[3]][base_list[3]] == same;
3010    
3011              /* This code is logically tricky. Think hard before fiddling with it.
3012              The posspropstab table has four entries per row. Each row relates to
3013              one of PCRE's special properties such as ALNUM or SPACE or WORD.
3014              Only WORD actually needs all four entries, but using repeats for the
3015              others means they can all use the same code below.
3016    
3017              The first two entries in each row are Unicode general categories, and
3018              apply always, because all the characters they include are part of the
3019              PCRE character set. The third and fourth entries are a general and a
3020              particular category, respectively, that include one or more relevant
3021              characters. One or the other is used, depending on whether the check
3022              is for a general or a particular category. However, in both cases the
3023              category contains more characters than the specials that are defined
3024              for the property being tested against. Therefore, it cannot be used
3025              in a NOTPROP case.
3026    
3027              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3028              Underscore is covered by ucp_P or ucp_Po. */
3029    
3030              case 6:  /* Left alphanum vs right general category */
3031              case 7:  /* Left space vs right general category */
3032              case 8:  /* Left word vs right general category */
3033              p = posspropstab[n-6];
3034              return risprop && lisprop ==
3035                (list[3] != p[0] &&
3036                 list[3] != p[1] &&
3037                (list[3] != p[2] || !lisprop));
3038    
3039              case 9:   /* Right alphanum vs left general category */
3040              case 10:  /* Right space vs left general category */
3041              case 11:  /* Right word vs left general category */
3042              p = posspropstab[n-9];
3043              return lisprop && risprop ==
3044                (base_list[3] != p[0] &&
3045                 base_list[3] != p[1] &&
3046                (base_list[3] != p[2] || !risprop));
3047    
3048              case 12:  /* Left alphanum vs right particular category */
3049              case 13:  /* Left space vs right particular category */
3050              case 14:  /* Left word vs right particular category */
3051              p = posspropstab[n-12];
3052              return risprop && lisprop ==
3053                (catposstab[p[0]][list[3]] &&
3054                 catposstab[p[1]][list[3]] &&
3055                (list[3] != p[3] || !lisprop));
3056    
3057              case 15:  /* Right alphanum vs left particular category */
3058              case 16:  /* Right space vs left particular category */
3059              case 17:  /* Right word vs left particular category */
3060              p = posspropstab[n-15];
3061              return lisprop && risprop ==
3062                (catposstab[p[0]][base_list[3]] &&
3063                 catposstab[p[1]][base_list[3]] &&
3064                (base_list[3] != p[3] || !risprop));
3065              }
3066            }
3067          return FALSE;
3068          }
3069    
3070        else
3071    #endif  /* SUPPORT_UCP */
3072    
3073        return leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3074               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3075               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3076      }      }
   if ((othercase = UCD_OTHERCASE(c)) != c) break;  
   }  
3077    
3078  if (c > d) return -1;  /* Reached end of range */    /* Control reaches here only if one of the items is a small character list.
3079      All characters are checked against the other side. */
3080    
3081  *ocptr = othercase;    do
3082  next = othercase + 1;      {
3083        chr = *chr_ptr;
3084    
3085  for (++c; c <= d; c++)      switch(list_ptr[0])
3086    {        {
3087    if (UCD_OTHERCASE(c) != next) break;        case OP_CHAR:
3088    next++;        ochr_ptr = list_ptr + 2;
3089    }        do
3090            {
3091            if (chr == *ochr_ptr) return FALSE;
3092            ochr_ptr++;
3093            }
3094          while(*ochr_ptr != NOTACHAR);
3095          break;
3096    
3097  *odptr = next - 1;     /* End of othercase range */        case OP_NOT:
3098  *cptr = c;             /* Rest of input range */        ochr_ptr = list_ptr + 2;
3099  return 0;        do
3100  }          {
3101            if (chr == *ochr_ptr)
3102              break;
3103            ochr_ptr++;
3104            }
3105          while(*ochr_ptr != NOTACHAR);
3106          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3107          break;
3108    
3109          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3110          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3111    
3112          case OP_DIGIT:
3113          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3114          break;
3115    
3116  /*************************************************        case OP_NOT_DIGIT:
3117  *        Check a character and a property        *        if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3118  *************************************************/        break;
3119    
3120  /* This function is called by check_auto_possessive() when a property item        case OP_WHITESPACE:
3121  is adjacent to a fixed character.        if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3122          break;
3123    
3124  Arguments:        case OP_NOT_WHITESPACE:
3125    c            the character        if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3126    ptype        the property type        break;
   pdata        the data for the type  
   negated      TRUE if it's a negated property (\P or \p{^)  
3127    
3128  Returns:       TRUE if auto-possessifying is OK        case OP_WORDCHAR:
3129  */        if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3130          break;
3131    
3132  static BOOL        case OP_NOT_WORDCHAR:
3133  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata, BOOL negated)        if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3134  {        break;
 #ifdef SUPPORT_UCP  
 const pcre_uint32 *p;  
 #endif  
3135    
3136  const ucd_record *prop = GET_UCD(c);        case OP_HSPACE:
3137          switch(chr)
3138            {
3139            HSPACE_CASES: return FALSE;
3140            default: break;
3141            }
3142          break;
3143    
3144  switch(ptype)        case OP_NOT_HSPACE:
3145    {        switch(chr)
3146    case PT_LAMP:          {
3147    return (prop->chartype == ucp_Lu ||          HSPACE_CASES: break;
3148            prop->chartype == ucp_Ll ||          default: return FALSE;
3149            prop->chartype == ucp_Lt) == negated;          }
3150          break;
3151    
3152    case PT_GC:        case OP_ANYNL:
3153    return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;        case OP_VSPACE:
3154          switch(chr)
3155            {
3156            VSPACE_CASES: return FALSE;
3157            default: break;
3158            }
3159          break;
3160    
3161    case PT_PC:        case OP_NOT_VSPACE:
3162    return (pdata == prop->chartype) == negated;        switch(chr)
3163            {
3164            VSPACE_CASES: break;
3165            default: return FALSE;
3166            }
3167          break;
3168    
3169    case PT_SC:        case OP_DOLL:
3170    return (pdata == prop->script) == negated;        case OP_EODN:
3171          switch (chr)
3172            {
3173            case CHAR_CR:
3174            case CHAR_LF:
3175            case CHAR_VT:
3176            case CHAR_FF:
3177            case CHAR_NEL:
3178    #ifndef EBCDIC
3179            case 0x2028:
3180            case 0x2029:
3181    #endif  /* Not EBCDIC */
3182            return FALSE;
3183            }
3184          break;
3185    
3186    /* These are specials */        case OP_EOD:    /* Can always possessify before \z */
3187          break;
3188    
3189    case PT_ALNUM:        case OP_PROP:
3190    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||        case OP_NOTPROP:
3191            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;        if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3192                list_ptr[0] == OP_NOTPROP))
3193            return FALSE;
3194          break;
3195    
3196    case PT_SPACE:    /* Perl space */        /* The class comparisons work only when the class is the second item
3197    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||        of the pair, because there are at present no possessive forms of the
3198            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)        class opcodes. Note also that the "code" variable that is used below
3199            == negated;        points after the second item, and that the pointer for the first item
3200          is not available, so even if there were possessive forms of the class
3201          opcodes, the correct comparison could not be done. */
3202    
3203          case OP_NCLASS:
3204          if (chr > 255) return FALSE;
3205          /* Fall through */
3206    
3207          case OP_CLASS:
3208          if (list_ptr != list) return FALSE;   /* Class is first opcode */
3209          if (chr > 255) break;
3210          if ((((pcre_uint8 *)(code - list_ptr[2] + 1))[chr >> 3] & (1 << (chr & 7))) != 0)
3211            return FALSE;
3212          break;
3213    
3214    case PT_PXSPACE:  /* POSIX space */  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3215    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||        case OP_XCLASS:
3216            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||        if (list_ptr != list) return FALSE;   /* Class is first opcode */
3217            c == CHAR_FF || c == CHAR_CR)        if (PRIV(xclass)(chr, code - list_ptr[2] + 1 + LINK_SIZE, utf))
3218            == negated;          return FALSE;
3219          break;
3220    #endif
3221    
3222    case PT_WORD:        default:
3223    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||        return FALSE;
3224            PRIV(ucp_gentype)[prop->chartype] == ucp_N ||        }
           c == CHAR_UNDERSCORE) == negated;  
3225    
3226  #ifdef SUPPORT_UCP      chr_ptr++;
   case PT_CLIST:  
   p = PRIV(ucd_caseless_sets) + prop->caseset;  
   for (;;)  
     {  
     if (c < *p) return !negated;  
     if (c == *p++) return negated;  
3227      }      }
3228    break;  /* Control never reaches here */    while(*chr_ptr != NOTACHAR);
3229  #endif  
3230      /* At least one character must be matched from this opcode. */
3231    
3232      if (list[1] == 0) return TRUE;
3233    }    }
3234    
3235  return FALSE;  return FALSE;
3236  }  }
 #endif  /* SUPPORT_UCP */  
3237    
3238    
3239    
3240  /*************************************************  /*************************************************
3241  *     Check if auto-possessifying is possible    *  *    Scan compiled regex for auto-possession     *
3242  *************************************************/  *************************************************/
3243    
3244  /* This function is called for unlimited repeats of certain items, to see  /* Replaces single character iterations with their possessive alternatives
3245  whether the next thing could possibly match the repeated item. If not, it makes  if appropriate. This function modifies the compiled opcode!
 sense to automatically possessify the repeated item.  
3246    
3247  Arguments:  Arguments:
3248    previous      pointer to the repeated opcode    code        points to start of the byte code
3249    utf           TRUE in UTF-8 / UTF-16 / UTF-32 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3250    ptr           next character in pattern    cd          static compile data
   options       options bits  
   cd            contains pointers to tables etc.  
3251    
3252  Returns:        TRUE if possessifying is wanted  Returns:      nothing
3253  */  */
3254    
3255  static BOOL  static void
3256  check_auto_possessive(const pcre_uchar *previous, BOOL utf,  auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
   const pcre_uchar *ptr, int options, compile_data *cd)  
3257  {  {
3258  pcre_uint32 c = NOTACHAR;  register pcre_uchar c;
3259  pcre_uint32 next;  const pcre_uchar *end;
3260  int escape;  pcre_uint32 list[8];
 pcre_uchar op_code = *previous++;  
   
 /* Skip whitespace and comments in extended mode */  
3261    
3262  if ((options & PCRE_EXTENDED) != 0)  for (;;)
3263    {    {
3264    for (;;)    c = *code;
3265    
3266      if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3267      {      {
3268      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      c -= get_repeat_base(c) - OP_STAR;
3269      if (*ptr == CHAR_NUMBER_SIGN)      end = (c <= OP_MINUPTO) ?
3270          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3271        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3272    
3273        if (end != NULL && compare_opcodes(end, utf, cd, list))
3274        {        {
3275        ptr++;        switch(c)
       while (*ptr != CHAR_NULL)  
3276          {          {
3277          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          case OP_STAR:
3278          ptr++;          *code += OP_POSSTAR - OP_STAR;
3279  #ifdef SUPPORT_UTF          break;
3280          if (utf) FORWARDCHAR(ptr);  
3281  #endif          case OP_MINSTAR:
3282            *code += OP_POSSTAR - OP_MINSTAR;
3283            break;
3284    
3285            case OP_PLUS:
3286            *code += OP_POSPLUS - OP_PLUS;
3287            break;
3288    
3289            case OP_MINPLUS:
3290            *code += OP_POSPLUS - OP_MINPLUS;
3291            break;
3292    
3293            case OP_QUERY:
3294            *code += OP_POSQUERY - OP_QUERY;
3295            break;
3296    
3297            case OP_MINQUERY:
3298            *code += OP_POSQUERY - OP_MINQUERY;
3299            break;
3300    
3301            case OP_UPTO:
3302            *code += OP_POSUPTO - OP_UPTO;
3303            break;
3304    
3305            case OP_MINUPTO:
3306            *code += OP_MINUPTO - OP_UPTO;
3307            break;
3308          }          }
3309        }        }
3310      else break;      c = *code;
3311      }      }
   }  
3312    
3313  /* If the next item is one that we can handle, get its value. A non-negative    switch(c)
3314  value is a character, a negative value is an escape value. */      {
3315        case OP_END:
3316        return;
3317    
3318  if (*ptr == CHAR_BACKSLASH)      case OP_TYPESTAR:
3319    {      case OP_TYPEMINSTAR:
3320    int temperrorcode = 0;      case OP_TYPEPLUS:
3321    escape = check_escape(&ptr, &next, &temperrorcode, cd->bracount, options,      case OP_TYPEMINPLUS:
3322      FALSE);      case OP_TYPEQUERY:
3323    if (temperrorcode != 0) return FALSE;      case OP_TYPEMINQUERY:
3324    ptr++;    /* Point after the escape sequence */      case OP_TYPEPOSSTAR:
3325    }      case OP_TYPEPOSPLUS:
3326  else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)      case OP_TYPEPOSQUERY:
3327    {      if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3328    escape = 0;      break;
 #ifdef SUPPORT_UTF  
   if (utf) { GETCHARINC(next, ptr); } else  
 #endif  
   next = *ptr++;  
   }  
 else return FALSE;  
3329    
3330  /* Skip whitespace and comments in extended mode */      case OP_TYPEUPTO:
3331        case OP_TYPEMINUPTO:
3332        case OP_TYPEEXACT:
3333        case OP_TYPEPOSUPTO:
3334        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3335          code += 2;
3336        break;
3337    
3338  if ((options & PCRE_EXTENDED) != 0)      case OP_XCLASS:
3339    {      code += GET(code, 1);
3340    for (;;)      break;
3341      {  
3342      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      case OP_MARK:
3343      if (*ptr == CHAR_NUMBER_SIGN)      case OP_PRUNE_ARG:
3344        {      case OP_SKIP_ARG:
3345        ptr++;      case OP_THEN_ARG:
3346        while (*ptr != CHAR_NULL)      code += code[1];
3347          {      break;
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
       }  
     else break;  
3348      }      }
   }  
3349    
3350  /* If the next thing is itself optional, we have to give up. */    /* Add in the fixed length from the table */
3351    
3352  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||    code += PRIV(OP_lengths)[c];
   STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)  
     return FALSE;  
3353    
3354  /* If the previous item is a character, get its value. */    /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3355      a multi-byte character. The length in the table is a minimum, so we have to
3356      arrange to skip the extra bytes. */
3357    
3358  if (op_code == OP_CHAR || op_code == OP_CHARI ||  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3359      op_code == OP_NOT || op_code == OP_NOTI)    if (utf) switch(c)
3360    {      {
3361  #ifdef SUPPORT_UTF      case OP_CHAR:
3362    GETCHARTEST(c, previous);      case OP_CHARI:
3363        case OP_NOT:
3364        case OP_NOTI:
3365        case OP_STAR:
3366        case OP_MINSTAR:
3367        case OP_PLUS:
3368        case OP_MINPLUS:
3369        case OP_QUERY:
3370        case OP_MINQUERY:
3371        case OP_UPTO:
3372        case OP_MINUPTO:
3373        case OP_EXACT:
3374        case OP_POSSTAR:
3375        case OP_POSPLUS:
3376        case OP_POSQUERY:
3377        case OP_POSUPTO:
3378        case OP_STARI:
3379        case OP_MINSTARI:
3380        case OP_PLUSI:
3381        case OP_MINPLUSI:
3382        case OP_QUERYI:
3383        case OP_MINQUERYI:
3384        case OP_UPTOI:
3385        case OP_MINUPTOI:
3386        case OP_EXACTI:
3387        case OP_POSSTARI:
3388        case OP_POSPLUSI:
3389        case OP_POSQUERYI:
3390        case OP_POSUPTOI:
3391        case OP_NOTSTAR:
3392        case OP_NOTMINSTAR:
3393        case OP_NOTPLUS:
3394        case OP_NOTMINPLUS:
3395        case OP_NOTQUERY:
3396        case OP_NOTMINQUERY:
3397        case OP_NOTUPTO:
3398        case OP_NOTMINUPTO:
3399        case OP_NOTEXACT:
3400        case OP_NOTPOSSTAR:
3401        case OP_NOTPOSPLUS:
3402        case OP_NOTPOSQUERY:
3403        case OP_NOTPOSUPTO:
3404        case OP_NOTSTARI:
3405        case OP_NOTMINSTARI:
3406        case OP_NOTPLUSI:
3407        case OP_NOTMINPLUSI:
3408        case OP_NOTQUERYI:
3409        case OP_NOTMINQUERYI:
3410        case OP_NOTUPTOI:
3411        case OP_NOTMINUPTOI:
3412        case OP_NOTEXACTI:
3413        case OP_NOTPOSSTARI:
3414        case OP_NOTPOSPLUSI:
3415        case OP_NOTPOSQUERYI:
3416        case OP_NOTPOSUPTOI:
3417        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3418        break;
3419        }
3420  #else  #else
3421    c = *previous;    (void)(utf);  /* Keep compiler happy by referencing function argument */
3422  #endif  #endif
3423    }    }
3424    }
3425    
 /* Now compare the next item with the previous opcode. First, handle cases when  
 the next item is a character. */  
3426    
 if (escape == 0)  
   {  
   /* For a caseless UTF match, the next character may have more than one other  
   case, which maps to the special PT_CLIST property. Check this first. */  
3427    
3428  #ifdef SUPPORT_UCP  /*************************************************
3429    if (utf && c != NOTACHAR && (options & PCRE_CASELESS) != 0)  *           Check for POSIX class syntax         *
3430      {  *************************************************/
     unsigned int ocs = UCD_CASESET(next);  
     if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);  
     }  
 #endif  
3431    
3432    switch(op_code)  /* This function is called when the sequence "[:" or "[." or "[=" is
3433      {  encountered in a character class. It checks whether this is followed by a
3434      case OP_CHAR:  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3435      return c != next;  reach an unescaped ']' without the special preceding character, return FALSE.
3436    
3437      /* For CHARI (caseless character) we must check the other case. If we have  Originally, this function only recognized a sequence of letters between the
3438      Unicode property support, we can use it to test the other case of  terminators, but it seems that Perl recognizes any sequence of characters,
3439      high-valued characters. We know that next can have only one other case,  though of course unknown POSIX names are subsequently rejected. Perl gives an
3440      because multi-other-case characters are dealt with above. */  "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3441    didn't consider this to be a POSIX class. Likewise for [:1234:].
3442    
3443      case OP_CHARI:  The problem in trying to be exactly like Perl is in the handling of escapes. We
3444      if (c == next) return FALSE;  have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3445  #ifdef SUPPORT_UTF  class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3446      if (utf)  below handles the special case of \], but does not try to do any other escape
3447        {  processing. This makes it different from Perl for cases such as [:l\ower:]
3448        pcre_uint32 othercase;  where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3449        if (next < 128) othercase = cd->fcc[next]; else  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
3450  #ifdef SUPPORT_UCP  I think.
       othercase = UCD_OTHERCASE(next);  
 #else  
       othercase = NOTACHAR;  
 #endif  
       return c != othercase;  
       }  
     else  
 #endif  /* SUPPORT_UTF */  
     return (c != TABLE_GET(next, cd->fcc, next));  /* Not UTF */  
3451    
3452      case OP_NOT:  A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3453      return c == next;  It seems that the appearance of a nested POSIX class supersedes an apparent
3454    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3455    a digit.
3456    
3457      case OP_NOTI:  In Perl, unescaped square brackets may also appear as part of class names. For
3458      if (c == next) return TRUE;  example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3459  #ifdef SUPPORT_UTF  [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3460      if (utf)  seem right at all. PCRE does not allow closing square brackets in POSIX class
3461    names.
3462    
3463    Arguments:
3464      ptr      pointer to the initial [
3465      endptr   where to return the end pointer
3466    
3467    Returns:   TRUE or FALSE
3468    */
3469    
3470    static BOOL
3471    check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3472    {
3473    pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
3474    terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
3475    for (++ptr; *ptr != CHAR_NULL; ptr++)
3476      {
3477      if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3478        ptr++;
3479      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3480      else
3481        {
3482        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3483        {        {
3484        pcre_uint32 othercase;        *endptr = ptr;
3485        if (next < 128) othercase = cd->fcc[next]; else        return TRUE;
 #ifdef SUPPORT_UCP  
       othercase = UCD_OTHERCASE(next);  
 #else  
       othercase = NOTACHAR;  
 #endif  
       return c == othercase;  
3486        }        }
3487      else      if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3488  #endif  /* SUPPORT_UTF */           (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3489      return (c == TABLE_GET(next, cd->fcc, next));  /* Not UTF */            ptr[1] == CHAR_EQUALS_SIGN) &&
3490            check_posix_syntax(ptr, endptr))
3491          return FALSE;
3492        }
3493      }
3494    return FALSE;
3495    }
3496    
     /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.  
     When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */  
3497    
     case OP_DIGIT:  
     return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;  
3498    
     case OP_NOT_DIGIT:  
     return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;  
3499    
3500      case OP_WHITESPACE:  /*************************************************
3501      return next > 255 || (cd->ctypes[next] & ctype_space) == 0;  *          Check POSIX class name                *
3502    *************************************************/
3503    
3504    /* This function is called to check the name given in a POSIX-style class entry
3505    such as [:alnum:].
3506    
3507    Arguments:
3508      ptr        points to the first letter
3509      len        the length of the name
3510    
3511    Returns:     a value representing the name, or -1 if unknown
3512    */
3513    
3514    static int
3515    check_posix_name(const pcre_uchar *ptr, int len)
3516    {
3517    const char *pn = posix_names;
3518    register int yield = 0;
3519    while (posix_name_lengths[yield] != 0)
3520      {
3521      if (len == posix_name_lengths[yield] &&
3522        STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3523      pn += posix_name_lengths[yield] + 1;
3524      yield++;
3525      }
3526    return -1;
3527    }
3528    
3529    
3530    /*************************************************
3531    *    Adjust OP_RECURSE items in repeated group   *
3532    *************************************************/
3533    
3534    /* OP_RECURSE items contain an offset from the start of the regex to the group
3535    that is referenced. This means that groups can be replicated for fixed
3536    repetition simply by copying (because the recursion is allowed to refer to
3537    earlier groups that are outside the current group). However, when a group is
3538    optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3539    inserted before it, after it has been compiled. This means that any OP_RECURSE
3540    items within it that refer to the group itself or any contained groups have to
3541    have their offsets adjusted. That one of the jobs of this function. Before it
3542    is called, the partially compiled regex must be temporarily terminated with
3543    OP_END.
3544    
3545    This function has been extended with the possibility of forward references for
3546    recursions and subroutine calls. It must also check the list of such references
3547    for the group we are dealing with. If it finds that one of the recursions in
3548    the current group is on this list, it adjusts the offset in the list, not the
3549    value in the reference (which is a group number).
3550    
3551      case OP_NOT_WHITESPACE:  Arguments:
3552      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;    group      points to the start of the group
3553      adjust     the amount by which the group is to be moved
3554      utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
3555      cd         contains pointers to tables etc.
3556      save_hwm   the hwm forward reference pointer at the start of the group
3557    
3558      case OP_WORDCHAR:  Returns:     nothing
3559      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;  */
3560    
3561      case OP_NOT_WORDCHAR:  static void
3562      return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3563      pcre_uchar *save_hwm)
3564    {
3565    pcre_uchar *ptr = group;
3566    
3567      case OP_HSPACE:  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3568      case OP_NOT_HSPACE:    {
3569      switch(next)    int offset;
3570        {    pcre_uchar *hc;
       HSPACE_CASES:  
       return op_code == OP_NOT_HSPACE;  
3571    
3572        default:    /* See if this recursion is on the forward reference list. If so, adjust the
3573        return op_code != OP_NOT_HSPACE;    reference. */
       }  
3574    
3575      case OP_ANYNL:    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3576      case OP_VSPACE:      {
3577      case OP_NOT_VSPACE:      offset = (int)GET(hc, 0);
3578      switch(next)      if (cd->start_code + offset == ptr + 1)
3579        {        {
3580        VSPACE_CASES:        PUT(hc, 0, offset + adjust);
3581        return op_code == OP_NOT_VSPACE;        break;
   
       default:  
       return op_code != OP_NOT_VSPACE;  
3582        }        }
   
 #ifdef SUPPORT_UCP  
     case OP_PROP:  
     return check_char_prop(next, previous[0], previous[1], FALSE);  
   
     case OP_NOTPROP:  
     return check_char_prop(next, previous[0], previous[1], TRUE);  
 #endif  
   
     default:  
     return FALSE;  
3583      }      }
   }  
3584    
3585  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP    /* Otherwise, adjust the recursion offset if it's after the start of this
3586  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are    group. */
 generated only when PCRE_UCP is *not* set, that is, when only ASCII  
 characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are  
 replaced by OP_PROP codes when PCRE_UCP is set. */  
3587    
3588  switch(op_code)    if (hc >= cd->hwm)
   {  
   case OP_CHAR:  
   case OP_CHARI:  
   switch(escape)  
3589      {      {
3590      case ESC_d:      offset = (int)GET(ptr, 1);
3591      return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3592        }
     case ESC_D:  
     return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;  
   
     case ESC_s:  
     return c > 255 || (cd->ctypes[c] & ctype_space) == 0;  
   
     case ESC_S:  
     return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;  
3593    
3594      case ESC_w:    ptr += 1 + LINK_SIZE;
3595      return c > 255 || (cd->ctypes[c] & ctype_word) == 0;    }
3596    }
3597    
     case ESC_W:  
     return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;  
3598    
     case ESC_h:  
     case ESC_H:  
     switch(c)  
       {  
       HSPACE_CASES:  
       return escape != ESC_h;  
3599    
3600        default:  /*************************************************
3601        return escape == ESC_h;  *        Insert an automatic callout point       *
3602        }  *************************************************/
3603    
3604      case ESC_v:  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
3605      case ESC_V:  callout points before each pattern item.
     switch(c)  
       {  
       VSPACE_CASES:  
       return escape != ESC_v;  
3606    
3607        default:  Arguments:
3608        return escape == ESC_v;    code           current code pointer
3609        }    ptr            current pattern pointer
3610      cd             pointers to tables etc
3611    
3612      /* When PCRE_UCP is set, these values get generated for \d etc. Find  Returns:         new code pointer
3613      their substitutions and process them. The result will always be either  */
     ESC_p or ESC_P. Then fall through to process those values. */  
3614    
3615  #ifdef SUPPORT_UCP  static pcre_uchar *
3616      case ESC_du:  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
3617      case ESC_DU:  {
3618      case ESC_wu:  *code++ = OP_CALLOUT;
3619      case ESC_WU:  *code++ = 255;
3620      case ESC_su:  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
3621      case ESC_SU:  PUT(code, LINK_SIZE, 0);                       /* Default length */
3622        {  return code + 2 * LINK_SIZE;
3623        int temperrorcode = 0;  }
       ptr = substitutes[escape - ESC_DU];  
       escape = check_escape(&ptr, &next, &temperrorcode, 0, options, FALSE);  
       if (temperrorcode != 0) return FALSE;  
       ptr++;    /* For compatibility */  
       }  
     /* Fall through */  
3624    
     case ESC_p:  
     case ESC_P:  
       {  
       unsigned int ptype = 0, pdata = 0;  
       int errorcodeptr;  
       BOOL negated;  
3625    
       ptr--;      /* Make ptr point at the p or P */  
       if (!get_ucp(&ptr, &negated, &ptype, &pdata, &errorcodeptr))  
         return FALSE;  
       ptr++;      /* Point past the final curly ket */  
3626    
3627        /* If the property item is optional, we have to give up. (When generated  /*************************************************
3628        from \d etc by PCRE_UCP, this test will have been applied much earlier,  *         Complete a callout item                *
3629        to the original \d etc. At this point, ptr will point to a zero byte. */  *************************************************/
3630    
3631        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  /* A callout item contains the length of the next item in the pattern, which
3632          STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)  we can't fill in till after we have reached the relevant point. This is used
3633            return FALSE;  for both automatic and manual callouts.
3634    
3635        /* Do the property check. */  Arguments:
3636      previous_callout   points to previous callout item
3637      ptr                current pattern pointer
3638      cd                 pointers to tables etc
3639    
3640        return check_char_prop(c, ptype, pdata, (escape == ESC_P) != negated);  Returns:             nothing
3641        }  */
 #endif  
3642    
3643      default:  static void
3644      return FALSE;  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
3645      }  {
3646    int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
3647    PUT(previous_callout, 2 + LINK_SIZE, length);
3648    }
3649    
   /* In principle, support for Unicode properties should be integrated here as  
   well. It means re-organizing the above code so as to get hold of the property  
   values before switching on the op-code. However, I wonder how many patterns  
   combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,  
   these op-codes are never generated.) */  
3650    
   case OP_DIGIT:  
   return escape == ESC_D || escape == ESC_s || escape == ESC_W ||  
          escape == ESC_h || escape == ESC_v || escape == ESC_R;  
3651    
3652    case OP_NOT_DIGIT:  #ifdef SUPPORT_UCP
3653    return escape == ESC_d;  /*************************************************
3654    *           Get othercase range                  *
3655    *************************************************/
3656    
3657    case OP_WHITESPACE:  /* This function is passed the start and end of a class range, in UTF-8 mode
3658    return escape == ESC_S || escape == ESC_d || escape == ESC_w;  with UCP support. It searches up the characters, looking for ranges of
3659    characters in the "other" case. Each call returns the next one, updating the
3660    start address. A character with multiple other cases is returned on its own
3661    with a special return value.
3662    
3663    case OP_NOT_WHITESPACE:  Arguments:
3664    return escape == ESC_s || escape == ESC_h || escape == ESC_v || escape == ESC_R;    cptr        points to starting character value; updated
3665      d           end value
3666      ocptr       where to put start of othercase range
3667      odptr       where to put end of othercase range
3668    
3669    case OP_HSPACE:  Yield:        -1 when no more
3670    return escape == ESC_S || escape == ESC_H || escape == ESC_d ||                 0 when a range is returned
3671           escape == ESC_w || escape == ESC_v || escape == ESC_R;                >0 the CASESET offset for char with multiple other cases
3672                    in this case, ocptr contains the original
3673    */
3674    
3675    case OP_NOT_HSPACE:  static int
3676    return escape == ESC_h;  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
3677      pcre_uint32 *odptr)
3678    {
3679    pcre_uint32 c, othercase, next;
3680    unsigned int co;
3681    
3682    /* Can't have \S in here because VT matches \S (Perl anomaly) */  /* Find the first character that has an other case. If it has multiple other
3683    case OP_ANYNL:  cases, return its case offset value. */
   case OP_VSPACE:  
   return escape == ESC_V || escape == ESC_d || escape == ESC_w;  
3684    
3685    case OP_NOT_VSPACE:  for (c = *cptr; c <= d; c++)
3686    return escape == ESC_v || escape == ESC_R;    {
3687      if ((co = UCD_CASESET(c)) != 0)
3688        {
3689        *ocptr = c++;   /* Character that has the set */
3690        *cptr = c;      /* Rest of input range */
3691        return (int)co;
3692        }
3693      if ((othercase = UCD_OTHERCASE(c)) != c) break;
3694      }
3695    
3696    case OP_WORDCHAR:  if (c > d) return -1;  /* Reached end of range */
   return escape == ESC_W || escape == ESC_s || escape == ESC_h ||  
          escape == ESC_v || escape == ESC_R;  
3697    
3698    case OP_NOT_WORDCHAR:  *ocptr = othercase;
3699    return escape == ESC_w || escape == ESC_d;  next = othercase + 1;
3700    
3701    default:  for (++c; c <= d; c++)
3702    return FALSE;    {
3703      if (UCD_OTHERCASE(c) != next) break;
3704      next++;
3705    }    }
3706    
3707  /* Control does not reach here */  *odptr = next - 1;     /* End of othercase range */
3708    *cptr = c;             /* Rest of input range */
3709    return 0;
3710  }  }
3711    #endif  /* SUPPORT_UCP */
3712    
3713    
3714    
# Line 4127  for (;; ptr++) Line 4627  for (;; ptr++)
4627              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
4628              continue;              continue;
4629    
4630              /* Perl 5.004 onwards omits VT from \s, but we must preserve it              /* Perl 5.004 onwards omitted VT from \s, but restored it at Perl
4631              if it was previously set by something earlier in the character              5.18. Before PCRE 8.34, we had to preserve the VT bit if it was
4632              class. Luckily, the value of CHAR_VT is 0x0b in both ASCII and              previously set by something earlier in the character class.
4633              EBCDIC, so we lazily just adjust the appropriate bit. */              Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
4634                we could just adjust the appropriate bit. From PCRE 8.34 we no
4635                longer treat \s and \S specially. */
4636    
4637              case ESC_s:              case ESC_s:
4638              classbits[0] |= cbits[cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
             classbits[1] |= cbits[cbit_space+1] & ~0x08;  
             for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];  
4639              continue;              continue;
4640    
4641              case ESC_S:              case ESC_S:
4642              should_flip_negation = TRUE;              should_flip_negation = TRUE;
4643              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
             classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */  
4644              continue;              continue;
4645    
4646              /* The rest apply in both UCP and non-UCP cases. */              /* The rest apply in both UCP and non-UCP cases. */
# Line 4640  for (;; ptr++) Line 5139  for (;; ptr++)
5139            }            }
5140          }          }
5141    
       /* If the repetition is unlimited, it pays to see if the next thing on  
       the line is something that cannot possibly match this character. If so,  
       automatically possessifying this item gains some performance in the case  
       where the match fails. */  
   
       if (!possessive_quantifier &&  
           repeat_max < 0 &&  
           check_auto_possessive(previous, utf, ptr + 1, options, cd))  
         {  
         repeat_type = 0;    /* Force greedy */  
         possessive_quantifier = TRUE;  
         }  
   
5142        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
5143        }        }
5144    
# Line 4670  for (;; ptr++) Line 5156  for (;; ptr++)
5156        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
5157        c = *previous;        c = *previous;
5158    
       if (!possessive_quantifier &&  
           repeat_max < 0 &&  
           check_auto_possessive(previous, utf, ptr + 1, options, cd))  
         {  
         repeat_type = 0;    /* Force greedy */  
         possessive_quantifier = TRUE;  
         }  
   
5159        OUTPUT_SINGLE_REPEAT:        OUTPUT_SINGLE_REPEAT:
5160        if (*previous == OP_PROP || *previous == OP_NOTPROP)        if (*previous == OP_PROP || *previous == OP_NOTPROP)
5161          {          {
# Line 4826  for (;; ptr++) Line 5304  for (;; ptr++)
5304      /* If previous was a character class or a back reference, we put the repeat      /* If previous was a character class or a back reference, we put the repeat
5305      stuff after it, but just skip the item if the repeat was {0,0}. */      stuff after it, but just skip the item if the repeat was {0,0}. */
5306    
5307      else if (*previous == OP_CLASS ||      else if (*previous == OP_CLASS || *previous == OP_NCLASS ||
              *previous == OP_NCLASS ||  
5308  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5309               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
5310  #endif  #endif
5311               *previous == OP_REF ||               *previous == OP_REF   || *previous == OP_REFI ||
5312               *previous == OP_REFI)               *previous == OP_DNREF || *previous == OP_DNREFI)
5313        {        {
5314        if (repeat_max == 0)        if (repeat_max == 0)
5315          {          {
# Line 5553  for (;; ptr++) Line 6030  for (;; ptr++)
6030                 tempptr[2] == CHAR_LESS_THAN_SIGN))                 tempptr[2] == CHAR_LESS_THAN_SIGN))
6031            break;            break;
6032    
6033          /* Most other conditions use OP_CREF (a couple change to OP_RREF          /* Other conditions use OP_CREF/OP_DNCREF/OP_RREF/OP_DNRREF, and all
6034          below), and all need to skip 1+IMM2_SIZE bytes at the start of the group. */          need to skip at least 1+IMM2_SIZE bytes at the start of the group. */
6035    
6036          code[1+LINK_SIZE] = OP_CREF;          code[1+LINK_SIZE] = OP_CREF;
6037          skipbytes = 1+IMM2_SIZE;          skipbytes = 1+IMM2_SIZE;
# Line 5570  for (;; ptr++) Line 6047  for (;; ptr++)
6047            }            }
6048    
6049          /* Check for a test for a named group's having been set, using the Perl          /* Check for a test for a named group's having been set, using the Perl
6050          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name'), and also allow for the original PCRE
6051            syntax of (?(name) or for (?(+n), (?(-n), and just (?(n). As names may
6052            consist entirely of digits, there is scope for ambiguity. */
6053    
6054          else if (ptr[1] == CHAR_LESS_THAN_SIGN)          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
6055            {            {
# Line 5587  for (;; ptr++) Line 6066  for (;; ptr++)
6066            terminator = CHAR_NULL;            terminator = CHAR_NULL;
6067            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
6068            }            }
6069    
6070            /* When a name is one of a number of duplicates, a different opcode is
6071            used and it needs more memory. Unfortunately we cannot tell whether a
6072            name is a duplicate in the first pass, so we have to allow for more
6073            memory except when we know it is a relative numerical reference. */
6074    
6075            if (refsign < 0 && lengthptr != NULL) *lengthptr += IMM2_SIZE;
6076    
6077          /* We now expect to read a name; any thing else is an error */          /* We now expect to read a name (possibly all digits); any thing else
6078            is an error. In the case of all digits, also get it as a number. */
6079    
6080          if (!MAX_255(ptr[1]) || (cd->ctypes[ptr[1]] & ctype_word) == 0)          if (!MAX_255(ptr[1]) || (cd->ctypes[ptr[1]] & ctype_word) == 0)
6081            {            {
# Line 5597  for (;; ptr++) Line 6084  for (;; ptr++)
6084            goto FAILED;            goto FAILED;
6085            }            }
6086    
         /* Read the name, but also get it as a number if it's all digits */  
   
6087          recno = 0;          recno = 0;
6088          name = ++ptr;          name = ++ptr;
6089          while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0)          while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0)
# Line 5609  for (;; ptr++) Line 6094  for (;; ptr++)
6094            }            }
6095          namelen = (int)(ptr - name);          namelen = (int)(ptr - name);
6096    
6097            /* Check the terminator */
6098    
6099          if ((terminator > 0 && *ptr++ != (pcre_uchar)terminator) ||          if ((terminator > 0 && *ptr++ != (pcre_uchar)terminator) ||
6100              *ptr++ != CHAR_RIGHT_PARENTHESIS)              *ptr++ != CHAR_RIGHT_PARENTHESIS)
6101            {            {
# Line 5644  for (;; ptr++) Line 6131  for (;; ptr++)
6131            }            }
6132    
6133          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
6134          name. If we find a name, add one to the opcode to change OP_CREF or          name. */
6135          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. */  
   
6136          slot = cd->name_table;          slot = cd->name_table;
6137          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
6138            {            {
# Line 5656  for (;; ptr++) Line 6140  for (;; ptr++)
6140            slot += cd->name_entry_size;            slot += cd->name_entry_size;
6141            }            }
6142    
6143          /* Found the named subpattern */          /* Found the named subpattern. If the name is duplicated, add one to
6144            the opcode to change CREF/RREF into DNCREF/DNRREF and insert
6145            appropriate data values. Otherwise, just insert the unique subpattern
6146            number. */
6147    
6148          if (i < cd->names_found)          if (i < cd->names_found)
6149            {            {
6150            recno = GET2(slot, 0);            int offset = i++;
6151            PUT2(code, 2+LINK_SIZE, recno);            int count = 1;
6152            code[1+LINK_SIZE]++;            recno = GET2(slot, 0);   /* Number from first found */
6153              for (; i < cd->names_found; i++)
6154                {
6155                slot += cd->name_entry_size;
6156                if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) != 0) break;
6157                count++;
6158                }
6159              if (count > 1)
6160                {
6161                PUT2(code, 2+LINK_SIZE, offset);
6162                PUT2(code, 2+LINK_SIZE+IMM2_SIZE, count);
6163                skipbytes += IMM2_SIZE;
6164                code[1+LINK_SIZE]++;
6165                }
6166              else  /* Not a duplicated name */
6167                {
6168                PUT2(code, 2+LINK_SIZE, recno);
6169                }
6170            }            }
6171    
6172          /* If terminator == CHAR_NULL it means that the name followed directly          /* If terminator == CHAR_NULL it means that the name followed directly
# Line 5837  for (;; ptr++) Line 6341  for (;; ptr++)
6341    
6342          /* In the pre-compile phase, do a syntax check, remember the longest          /* In the pre-compile phase, do a syntax check, remember the longest
6343          name, and then remember the group in a vector, expanding it if          name, and then remember the group in a vector, expanding it if
6344          necessary. Duplicates for the same number are skipped; other duplicates          necessary. Duplicates for the same number are skipped; other duplicates
6345          are checked for validity. In the actual compile, there is nothing to          are checked for validity. In the actual compile, there is nothing to
6346          do. */          do. */
6347    
6348          if (lengthptr != NULL)          if (lengthptr != NULL)
6349            {            {
6350            named_group *ng;            named_group *ng;
6351            pcre_uint32 number = cd->bracount + 1;            pcre_uint32 number = cd->bracount + 1;
6352    
6353            if (*ptr != (pcre_uchar)terminator)            if (*ptr != (pcre_uchar)terminator)
6354              {              {
6355              *errorcodeptr = ERR42;              *errorcodeptr = ERR42;
6356              goto FAILED;              goto FAILED;
6357              }              }
6358    
6359            if (cd->names_found >= MAX_NAME_COUNT)            if (cd->names_found >= MAX_NAME_COUNT)
6360              {              {
6361              *errorcodeptr = ERR49;              *errorcodeptr = ERR49;
6362              goto FAILED;              goto FAILED;
6363              }              }
6364    
6365            if (namelen + IMM2_SIZE + 1 > cd->name_entry_size)            if (namelen + IMM2_SIZE + 1 > cd->name_entry_size)
6366              {              {
6367              cd->name_entry_size = namelen + IMM2_SIZE + 1;              cd->name_entry_size = namelen + IMM2_SIZE + 1;
# Line 5867  for (;; ptr++) Line 6371  for (;; ptr++)
6371                goto FAILED;                goto FAILED;
6372                }                }
6373              }              }
6374    
6375            /* Scan the list to check for duplicates. For duplicate names, if the            /* Scan the list to check for duplicates. For duplicate names, if the
6376            number is the same, break the loop, which causes the name to be            number is the same, break the loop, which causes the name to be
6377            discarded; otherwise, if DUPNAMES is not set, give an error.            discarded; otherwise, if DUPNAMES is not set, give an error.
6378            If it is set, allow the name with a different number, but continue            If it is set, allow the name with a different number, but continue
6379            scanning in case this is a duplicate with the same number. For            scanning in case this is a duplicate with the same number. For
6380            non-duplicate names, give an error if the number is duplicated. */            non-duplicate names, give an error if the number is duplicated. */
6381    
6382            ng = cd->named_groups;            ng = cd->named_groups;
6383            for (i = 0; i < cd->names_found; i++, ng++)            for (i = 0; i < cd->names_found; i++, ng++)
6384              {              {
6385              if (namelen == ng->length &&              if (namelen == ng->length &&
6386                  STRNCMP_UC_UC(name, ng->name, namelen) == 0)                  STRNCMP_UC_UC(name, ng->name, namelen) == 0)
6387                {                {
6388                if (ng->number == number) break;                if (ng->number == number) break;
6389                if ((options & PCRE_DUPNAMES) == 0)                if ((options & PCRE_DUPNAMES) == 0)
6390                  {                  {
6391                  *errorcodeptr = ERR43;                  *errorcodeptr = ERR43;
6392                  goto FAILED;                  goto FAILED;
6393                  }                  }
6394                }                cd->dupnames = TRUE;  /* Duplicate names exist */
6395                  }
6396              else if (ng->number == number)              else if (ng->number == number)
6397                {                {
6398                *errorcodeptr = ERR65;                *errorcodeptr = ERR65;
6399                goto FAILED;                goto FAILED;
6400                }                }
6401              }              }
6402    
6403            if (i >= cd->names_found)     /* Not a duplicate with same number */            if (i >= cd->names_found)     /* Not a duplicate with same number */
6404              {              {
6405              /* Increase the list size if necessary */              /* Increase the list size if necessary */
6406    
6407              if (cd->names_found >= cd->named_group_list_size)              if (cd->names_found >= cd->named_group_list_size)
6408                {                {
6409                int newsize = cd->named_group_list_size * 2;                int newsize = cd->named_group_list_size * 2;
6410                named_group *newspace = (PUBL(malloc))                named_group *newspace = (PUBL(malloc))
6411                  (newsize * sizeof(named_group));                  (newsize * sizeof(named_group));
6412    
6413                if (newspace == NULL)                if (newspace == NULL)
6414                  {                  {
6415                  *errorcodeptr = ERR21;                  *errorcodeptr = ERR21;
6416                  goto FAILED;                  goto FAILED;
6417                  }                  }
6418    
6419                memcpy(newspace, cd->named_groups,                memcpy(newspace, cd->named_groups,
6420                  cd->named_group_list_size * sizeof(named_group));                  cd->named_group_list_size * sizeof(named_group));
6421                if (cd->named_group_list_size > NAMED_GROUP_LIST_SIZE)                if (cd->named_group_list_size > NAMED_GROUP_LIST_SIZE)
6422                  (PUBL(free))((void *)cd->named_groups);                  (PUBL(free))((void *)cd->named_groups);
6423                cd->named_groups = newspace;                cd->named_groups = newspace;
6424                cd->named_group_list_size = newsize;                cd->named_group_list_size = newsize;
6425                }                }
6426    
6427              cd->named_groups[cd->names_found].name = name;              cd->named_groups[cd->names_found].name = name;
6428              cd->named_groups[cd->names_found].length = namelen;              cd->named_groups[cd->names_found].length = namelen;
6429              cd->named_groups[cd->names_found].number = number;              cd->named_groups[cd->names_found].number = number;
# Line 5957  for (;; ptr++) Line 6462  for (;; ptr++)
6462          if (lengthptr != NULL)          if (lengthptr != NULL)
6463            {            {
6464            named_group *ng;            named_group *ng;
6465    
6466            if (namelen == 0)            if (namelen == 0)
6467              {              {
6468              *errorcodeptr = ERR62;              *errorcodeptr = ERR62;
# Line 5975  for (;; ptr++) Line 6480  for (;; ptr++)
6480              }              }
6481    
6482            /* The name table does not exist in the first pass; instead we must            /* The name table does not exist in the first pass; instead we must
6483            scan the list of names encountered so far in order to get the            scan the list of names encountered so far in order to get the
6484            number. If the name is not found, set the value to 0 for a forward            number. If the name is not found, set the value to 0 for a forward
6485            reference. */            reference. */
6486    
# Line 5985  for (;; ptr++) Line 6490  for (;; ptr++)
6490              if (namelen == ng->length &&              if (namelen == ng->length &&
6491                  STRNCMP_UC_UC(name, ng->name, namelen) == 0)                  STRNCMP_UC_UC(name, ng->name, namelen) == 0)
6492                break;                break;
6493              }              }
6494            recno = (i < cd->names_found)? ng->number : 0;            recno = (i < cd->names_found)? ng->number : 0;
6495    
6496              /* Count named back references. */
6497    
6498              if (!is_recurse) cd->namedrefcount++;
6499            }            }
6500    
6501          /* In the real compile, search the name table. We check the name          /* In the real compile, search the name table. We check the name
# Line 6009  for (;; ptr++) Line 6518  for (;; ptr++)
6518              {              {
6519              recno = GET2(slot, 0);              recno = GET2(slot, 0);
6520              }              }
6521            else            else
6522              {              {
6523              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
6524              goto FAILED;              goto FAILED;
6525              }              }
6526            }            }
6527    
6528          /* In both phases, we can now go to the code than handles numerical          /* In both phases, for recursions, we can now go to the code than
6529          recursion or backreferences. */          handles numerical recursion. */
6530    
6531          if (is_recurse) goto HANDLE_RECURSION;          if (is_recurse) goto HANDLE_RECURSION;
6532            else goto HANDLE_REFERENCE;  
6533            /* In the second pass we must see if the name is duplicated. If so, we
6534            generate a different opcode. */
6535    
6536            if (lengthptr == NULL && cd->dupnames)
6537              {
6538              int count = 1;
6539              unsigned int index = i;
6540              pcre_uchar *cslot = slot + cd->name_entry_size;
6541    
6542              for (i++; i < cd->names_found; i++)
6543                {
6544                if (STRCMP_UC_UC(slot + IMM2_SIZE, cslot + IMM2_SIZE) != 0) break;
6545                count++;
6546                cslot += cd->name_entry_size;
6547                }
6548    
6549              if (count > 1)
6550                {
6551                if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
6552                previous = code;
6553                *code++ = ((options & PCRE_CASELESS) != 0)? OP_DNREFI : OP_DNREF;
6554                PUT2INC(code, 0, index);
6555                PUT2INC(code, 0, count);
6556    
6557                /* Process each potentially referenced group. */
6558    
6559                for (; slot < cslot; slot += cd->name_entry_size)
6560                  {
6561                  open_capitem *oc;
6562                  recno = GET2(slot, 0);
6563                  cd->backref_map |= (recno < 32)? (1 << recno) : 1;
6564                  if (recno > cd->top_backref) cd->top_backref = recno;
6565    
6566                  /* Check to see if this back reference is recursive, that it, it
6567                  is inside the group that it references. A flag is set so that the
6568                  group can be made atomic. */
6569    
6570                  for (oc = cd->open_caps; oc != NULL; oc = oc->next)
6571                    {
6572                    if (oc->number == recno)
6573                      {
6574                      oc->flag = TRUE;
6575                      break;
6576                      }
6577                    }
6578                  }
6579    
6580                continue;  /* End of back ref handling */
6581                }
6582              }
6583    
6584            /* First pass, or a non-duplicated name. */
6585    
6586            goto HANDLE_REFERENCE;
6587    
6588    
6589          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
# Line 6603  for (;; ptr++) Line 7166  for (;; ptr++)
7166          open_capitem *oc;          open_capitem *oc;
7167          recno = -escape;          recno = -escape;
7168    
7169          HANDLE_REFERENCE:    /* Come here from named backref handling */          /* Come here from named backref handling when the reference is to a
7170            single group (i.e. not to a duplicated name. */
7171    
7172            HANDLE_REFERENCE:
7173          if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;          if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
7174          previous = code;          previous = code;
7175          *code++ = ((options & PCRE_CASELESS) != 0)? OP_REFI : OP_REF;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_REFI : OP_REF;
# Line 7290  do { Line 7856  do {
7856       switch (*scode)       switch (*scode)
7857         {         {
7858         case OP_CREF:         case OP_CREF:
7859         case OP_NCREF:         case OP_DNCREF:
7860         case OP_RREF:         case OP_RREF:
7861         case OP_NRREF:         case OP_DNRREF:
7862         case OP_DEF:         case OP_DEF:
7863         return FALSE;         return FALSE;
7864    
# Line 7468  return c; Line 8034  return c;
8034  *************************************************/  *************************************************/
8035    
8036  /* This function is called between compiling passes to add an entry to the  /* This function is called between compiling passes to add an entry to the
8037  name/number table, maintaining alphabetical order. Checking for permitted  name/number table, maintaining alphabetical order. Checking for permitted
8038  and forbidden duplicates has already been done.  and forbidden duplicates has already been done.
8039    
8040  Arguments:  Arguments:
8041    cd           the compile data block    cd           the compile data block
8042    name         the name to add    name         the name to add
8043    length       the length of the name    length       the length of the name
8044    groupno      the group number    groupno      the group number
8045    
8046  Returns:       nothing  Returns:       nothing
8047  */  */
8048    
8049  static void  static void
8050  add_name(compile_data *cd, const pcre_uchar *name, int length,  add_name(compile_data *cd, const pcre_uchar *name, int length,
8051    unsigned int groupno)    unsigned int groupno)
8052  {  {
8053  int i;  int i;
# Line 7490  pcre_uchar *slot = cd->name_table; Line 8056  pcre_uchar *slot = cd->name_table;
8056  for (i = 0; i < cd->names_found; i++)  for (i = 0; i < cd->names_found; i++)
8057    {    {
8058    int crc = memcmp(name, slot+IMM2_SIZE, IN_UCHARS(length));    int crc = memcmp(name, slot+IMM2_SIZE, IN_UCHARS(length));
8059    if (crc == 0 && slot[IMM2_SIZE+length] != 0)    if (crc == 0 && slot[IMM2_SIZE+length] != 0)
8060      crc = -1; /* Current name is a substring */      crc = -1; /* Current name is a substring */
8061    
8062    /* Make space in the table and break the loop for an earlier name. For a    /* Make space in the table and break the loop for an earlier name. For a
# Line 7605  new memory is obtained from malloc(). */ Line 8171  new memory is obtained from malloc(). */
8171    
8172  pcre_uchar cworkspace[COMPILE_WORK_SIZE];  pcre_uchar cworkspace[COMPILE_WORK_SIZE];
8173    
8174  /* This vector is used for remembering name groups during the pre-compile. In a  /* This vector is used for remembering name groups during the pre-compile. In a
8175  similar way to cworkspace, it can be expanded using malloc() if necessary. */  similar way to cworkspace, it can be expanded using malloc() if necessary. */
8176    
8177  named_group named_groups[NAMED_GROUP_LIST_SIZE];  named_group named_groups[NAMED_GROUP_LIST_SIZE];
# Line 7872  cd->bracount = cd->final_bracount = 0; Line 8438  cd->bracount = cd->final_bracount = 0;
8438  cd->names_found = 0;  cd->names_found = 0;
8439  cd->name_entry_size = 0;  cd->name_entry_size = 0;
8440  cd->name_table = NULL;  cd->name_table = NULL;
8441    cd->dupnames = FALSE;
8442    cd->namedrefcount = 0;
8443  cd->start_code = cworkspace;  cd->start_code = cworkspace;
8444  cd->hwm = cworkspace;  cd->hwm = cworkspace;
8445  cd->start_workspace = cworkspace;  cd->start_workspace = cworkspace;
# Line 7909  if (length > MAX_PATTERN_SIZE) Line 8477  if (length > MAX_PATTERN_SIZE)
8477    goto PCRE_EARLY_ERROR_RETURN;    goto PCRE_EARLY_ERROR_RETURN;
8478    }    }
8479    
8480  /* Compute the size of data block needed and get it, either from malloc or  /* If there are groups with duplicate names and there are also references by
8481  externally provided function. Integer overflow should no longer be possible  name, we must allow for the possibility of named references to duplicated
8482  because nowadays we limit the maximum value of cd->names_found and  groups. These require an extra data item each. */
 cd->name_entry_size. */  
8483    
8484  size = sizeof(REAL_PCRE) + (length + cd->names_found * cd->name_entry_size) * sizeof(pcre_uchar);  if (cd->dupnames && cd->namedrefcount > 0)
8485  re = (REAL_PCRE *)(PUBL(malloc))(size);    length += cd->namedrefcount * IMM2_SIZE * sizeof(pcre_uchar);
8486    
8487    /* Compute the size of the data block for storing the compiled pattern. Integer
8488    overflow should no longer be possible because nowadays we limit the maximum
8489    value of cd->names_found and cd->name_entry_size. */
8490    
8491    size = sizeof(REAL_PCRE) +
8492      (length + cd->names_found * cd->name_entry_size) * sizeof(pcre_uchar);
8493    
8494    /* Get the memory. */
8495    
8496    re = (REAL_PCRE *)(PUBL(malloc))(size);
8497  if (re == NULL)  if (re == NULL)
8498    {    {
8499    errorcode = ERR21;    errorcode = ERR21;
# Line 7970  cd->had_pruneorskip = FALSE; Line 8547  cd->had_pruneorskip = FALSE;
8547  cd->check_lookbehind = FALSE;  cd->check_lookbehind = FALSE;
8548  cd->open_caps = NULL;  cd->open_caps = NULL;
8549    
8550  /* If any named groups were found, create the name/number table from the list  /* If any named groups were found, create the name/number table from the list
8551  created in the first pass. */  created in the first pass. */
8552    
8553  if (cd->names_found > 0)  if (cd->names_found > 0)
8554    {    {
8555    int i = cd->names_found;    int i = cd->names_found;
8556    named_group *ng = cd->named_groups;    named_group *ng = cd->named_groups;
8557    cd->names_found = 0;    cd->names_found = 0;
8558    for (; i > 0; i--, ng++)    for (; i > 0; i--, ng++)
8559      add_name(cd, ng->name, ng->length, ng->number);      add_name(cd, ng->name, ng->length, ng->number);
8560    if (cd->named_group_list_size > NAMED_GROUP_LIST_SIZE)    if (cd->named_group_list_size > NAMED_GROUP_LIST_SIZE)
8561      (PUBL(free))((void *)cd->named_groups);      (PUBL(free))((void *)cd->named_groups);
8562    }    }
8563    
8564  /* Set up a starting, non-extracting bracket, then compile the expression. On  /* Set up a starting, non-extracting bracket, then compile the expression. On
8565  error, errorcode will be set non-zero, so we don't need to look at the result  error, errorcode will be set non-zero, so we don't need to look at the result
# Line 8059  subpattern. */ Line 8636  subpattern. */
8636    
8637  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;
8638    
8639    /* Unless disabled, check whether single character iterators can be
8640    auto-possessified. The function overwrites the appropriate opcode values. */
8641    
8642    if ((options & PCRE_NO_AUTO_POSSESSIFY) == 0)
8643      auto_possessify((pcre_uchar *)codestart, utf, cd);
8644    
8645  /* If there were any lookbehind assertions that contained OP_RECURSE  /* If there were any lookbehind assertions that contained OP_RECURSE
8646  (recursions or subroutine calls), a flag is set for them to be checked here,  (recursions or subroutine calls), a flag is set for them to be checked here,
8647  because they may contain forward references. Actual recursions cannot be fixed  because they may contain forward references. Actual recursions cannot be fixed

Legend:
Removed from v.1360  
changed lines
  Added in v.1365

  ViewVC Help
Powered by ViewVC 1.1.5