/[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 1348 by ph10, Fri Jul 5 10:38:37 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
119    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
121    workspace. The value is the number of slots in the list. */
122    
123    #define NAMED_GROUP_LIST_SIZE  20
124    
125  /* The overrun tests check for a slightly smaller size so that they detect the  /* The overrun tests check for a slightly smaller size so that they detect the
126  overrun before it actually does run off the end of the data block. */  overrun before it actually does run off the end of the data block. */
127    
# Line 648  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 675  return s; Line 801  return s;
801  }  }
802    
803    
804    
805  /*************************************************  /*************************************************
806  *           Expand the workspace                 *  *           Expand the workspace                 *
807  *************************************************/  *************************************************/
# Line 1192  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 1289  return FALSE; Line 1418  return FALSE;
1418    
1419    
1420    
   
1421  /*************************************************  /*************************************************
1422  *         Read repeat counts                     *  *         Read repeat counts                     *
1423  *************************************************/  *************************************************/
# Line 1358  return p; Line 1486  return p;
1486    
1487    
1488  /*************************************************  /*************************************************
 *  Subroutine for finding forward reference      *  
 *************************************************/  
   
 /* This recursive function is called only from find_parens() below. The  
 top-level call starts at the beginning of the pattern. All other calls must  
 start at a parenthesis. It scans along a pattern's text looking for capturing  
 subpatterns, and counting them. If it finds a named pattern that matches the  
 name it is given, it returns its number. Alternatively, if the name is NULL, it  
 returns when it reaches a given numbered subpattern. Recursion is used to keep  
 track of subpatterns that reset the capturing group numbers - the (?| feature.  
   
 This function was originally called only from the second pass, in which we know  
 that if (?< or (?' or (?P< is encountered, the name will be correctly  
 terminated because that is checked in the first pass. There is now one call to  
 this function in the first pass, to check for a recursive back reference by  
 name (so that we can make the whole group atomic). In this case, we need check  
 only up to the current position in the pattern, and that is still OK because  
 and previous occurrences will have been checked. To make this work, the test  
 for "end of pattern" is a check against cd->end_pattern in the main loop,  
 instead of looking for a binary zero. This means that the special first-pass  
 call can adjust cd->end_pattern temporarily. (Checks for binary zero while  
 processing items within the loop are OK, because afterwards the main loop will  
 terminate.)  
   
 Arguments:  
   ptrptr       address of the current character pointer (updated)  
   cd           compile background data  
   name         name to seek, or NULL if seeking a numbered subpattern  
   lorn         name length, or subpattern number if name is NULL  
   xmode        TRUE if we are in /x mode  
   utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode  
   count        pointer to the current capturing subpattern number (updated)  
   
 Returns:       the number of the named subpattern, or -1 if not found  
 */  
   
 static int  
 find_parens_sub(pcre_uchar **ptrptr, compile_data *cd, const pcre_uchar *name, int lorn,  
   BOOL xmode, BOOL utf, int *count)  
 {  
 pcre_uchar *ptr = *ptrptr;  
 int start_count = *count;  
 int hwm_count = start_count;  
 BOOL dup_parens = FALSE;  
   
 /* If the first character is a parenthesis, check on the type of group we are  
 dealing with. The very first call may not start with a parenthesis. */  
   
 if (ptr[0] == CHAR_LEFT_PARENTHESIS)  
   {  
   /* Handle specials such as (*SKIP) or (*UTF8) etc. */  
   
   if (ptr[1] == CHAR_ASTERISK)  
     {  
     ptr += 2;  
     while (ptr < cd->end_pattern && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;  
     }  
   
   /* Handle a normal, unnamed capturing parenthesis. */  
   
   else if (ptr[1] != CHAR_QUESTION_MARK)  
     {  
     *count += 1;  
     if (name == NULL && *count == lorn) return *count;  
     ptr++;  
     }  
   
   /* All cases now have (? at the start. Remember when we are in a group  
   where the parenthesis numbers are duplicated. */  
   
   else if (ptr[2] == CHAR_VERTICAL_LINE)  
     {  
     ptr += 3;  
     dup_parens = TRUE;  
     }  
   
   /* Handle comments; all characters are allowed until a ket is reached. */  
   
   else if (ptr[2] == CHAR_NUMBER_SIGN)  
     {  
     for (ptr += 3; *ptr != CHAR_NULL; ptr++)  
       if (*ptr == CHAR_RIGHT_PARENTHESIS) break;  
     goto FAIL_EXIT;  
     }  
   
   /* Handle a condition. If it is an assertion, just carry on so that it  
   is processed as normal. If not, skip to the closing parenthesis of the  
   condition (there can't be any nested parens). */  
   
   else if (ptr[2] == CHAR_LEFT_PARENTHESIS)  
     {  
     ptr += 2;  
     if (ptr[1] != CHAR_QUESTION_MARK)  
       {  
       while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;  
       if (*ptr != CHAR_NULL) ptr++;  
       }  
     }  
   
   /* Start with (? but not a condition. */  
   
   else  
     {  
     ptr += 2;  
     if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */  
   
     /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */  
   
     if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&  
         ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)  
       {  
       pcre_uchar term;  
       const pcre_uchar *thisname;  
       *count += 1;  
       if (name == NULL && *count == lorn) return *count;  
       term = *ptr++;  
       if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;  
       thisname = ptr;  
       while (*ptr != term) ptr++;  
       if (name != NULL && lorn == (int)(ptr - thisname) &&  
           STRNCMP_UC_UC(name, thisname, (unsigned int)lorn) == 0)  
         return *count;  
       term++;  
       }  
     }  
   }  
   
 /* Past any initial parenthesis handling, scan for parentheses or vertical  
 bars. Stop if we get to cd->end_pattern. Note that this is important for the  
 first-pass call when this value is temporarily adjusted to stop at the current  
 position. So DO NOT change this to a test for binary zero. */  
   
 for (; ptr < cd->end_pattern; ptr++)  
   {  
   /* Skip over backslashed characters and also entire \Q...\E */  
   
   if (*ptr == CHAR_BACKSLASH)  
     {  
     if (*(++ptr) == CHAR_NULL) goto FAIL_EXIT;  
     if (*ptr == CHAR_Q) for (;;)  
       {  
       while (*(++ptr) != CHAR_NULL && *ptr != CHAR_BACKSLASH) {};  
       if (*ptr == CHAR_NULL) goto FAIL_EXIT;  
       if (*(++ptr) == CHAR_E) break;  
       }  
     continue;  
     }  
   
   /* Skip over character classes; this logic must be similar to the way they  
   are handled for real. If the first character is '^', skip it. Also, if the  
   first few characters (either before or after ^) are \Q\E or \E we skip them  
   too. This makes for compatibility with Perl. Note the use of STR macros to  
   encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */  
   
   if (*ptr == CHAR_LEFT_SQUARE_BRACKET)  
     {  
     BOOL negate_class = FALSE;  
     for (;;)  
       {  
       if (ptr[1] == CHAR_BACKSLASH)  
         {  
         if (ptr[2] == CHAR_E)  
           ptr+= 2;  
         else if (STRNCMP_UC_C8(ptr + 2,  
                  STR_Q STR_BACKSLASH STR_E, 3) == 0)  
           ptr += 4;  
         else  
           break;  
         }  
       else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)  
         {  
         negate_class = TRUE;  
         ptr++;  
         }  
       else break;  
       }  
   
     /* If the next character is ']', it is a data character that must be  
     skipped, except in JavaScript compatibility mode. */  
   
     if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&  
         (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)  
       ptr++;  
   
     while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)  
       {  
       if (*ptr == CHAR_NULL) return -1;  
       if (*ptr == CHAR_BACKSLASH)  
         {  
         if (*(++ptr) == CHAR_NULL) goto FAIL_EXIT;  
         if (*ptr == CHAR_Q) for (;;)  
           {  
           while (*(++ptr) != CHAR_NULL && *ptr != CHAR_BACKSLASH) {};  
           if (*ptr == CHAR_NULL) goto FAIL_EXIT;  
           if (*(++ptr) == CHAR_E) break;  
           }  
         continue;  
         }  
       }  
     continue;  
     }  
   
   /* Skip comments in /x mode */  
   
   if (xmode && *ptr == CHAR_NUMBER_SIGN)  
     {  
     ptr++;  
     while (*ptr != CHAR_NULL)  
       {  
       if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }  
       ptr++;  
 #ifdef SUPPORT_UTF  
       if (utf) FORWARDCHAR(ptr);  
 #endif  
       }  
     if (*ptr == CHAR_NULL) goto FAIL_EXIT;  
     continue;  
     }  
   
   /* Check for the special metacharacters */  
   
   if (*ptr == CHAR_LEFT_PARENTHESIS)  
     {  
     int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);  
     if (rc > 0) return rc;  
     if (*ptr == CHAR_NULL) goto FAIL_EXIT;  
     }  
   
   else if (*ptr == CHAR_RIGHT_PARENTHESIS)  
     {  
     if (dup_parens && *count < hwm_count) *count = hwm_count;  
     goto FAIL_EXIT;  
     }  
   
   else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)  
     {  
     if (*count > hwm_count) hwm_count = *count;  
     *count = start_count;  
     }  
   }  
   
 FAIL_EXIT:  
 *ptrptr = ptr;  
 return -1;  
 }  
   
   
   
   
 /*************************************************  
 *       Find forward referenced subpattern       *  
 *************************************************/  
   
 /* This function scans along a pattern's text looking for capturing  
 subpatterns, and counting them. If it finds a named pattern that matches the  
 name it is given, it returns its number. Alternatively, if the name is NULL, it  
 returns when it reaches a given numbered subpattern. This is used for forward  
 references to subpatterns. We used to be able to start this scan from the  
 current compiling point, using the current count value from cd->bracount, and  
 do it all in a single loop, but the addition of the possibility of duplicate  
 subpattern numbers means that we have to scan from the very start, in order to  
 take account of such duplicates, and to use a recursive function to keep track  
 of the different types of group.  
   
 Arguments:  
   cd           compile background data  
   name         name to seek, or NULL if seeking a numbered subpattern  
   lorn         name length, or subpattern number if name is NULL  
   xmode        TRUE if we are in /x mode  
   utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode  
   
 Returns:       the number of the found subpattern, or -1 if not found  
 */  
   
 static int  
 find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,  
   BOOL utf)  
 {  
 pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;  
 int count = 0;  
 int rc;  
   
 /* If the pattern does not start with an opening parenthesis, the first call  
 to find_parens_sub() will scan right to the end (if necessary). However, if it  
 does start with a parenthesis, find_parens_sub() will return when it hits the  
 matching closing parens. That is why we have to have a loop. */  
   
 for (;;)  
   {  
   rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);  
   if (rc > 0 || *ptr++ == CHAR_NULL) break;  
   }  
   
 return rc;  
 }  
   
   
   
   
 /*************************************************  
1489  *      Find first significant op code            *  *      Find first significant op code            *
1490  *************************************************/  *************************************************/
1491    
# Line 1696  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 1712  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 1836  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 2015  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 2051  for (;;) Line 1880  for (;;)
1880    
1881    
1882    
   
1883  /*************************************************  /*************************************************
1884  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
1885  *************************************************/  *************************************************/
# Line 2361  Returns:      TRUE if what is matched co Line 2189  Returns:      TRUE if what is matched co
2189  typedef struct recurse_check {  typedef struct recurse_check {
2190    struct recurse_check *prev;    struct recurse_check *prev;
2191    const pcre_uchar *group;    const pcre_uchar *group;
2192  } recurse_check;  } recurse_check;
2193    
2194  static BOOL  static BOOL
2195  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
# Line 2377  for (code = first_significant_code(code Line 2205  for (code = first_significant_code(code
2205    const pcre_uchar *ccode;    const pcre_uchar *ccode;
2206    
2207    c = *code;    c = *code;
2208    
2209    /* Skip over forward assertions; the other assertions are skipped by    /* Skip over forward assertions; the other assertions are skipped by
2210    first_significant_code() with a TRUE final argument. */    first_significant_code() with a TRUE final argument. */
2211    
# Line 2405  for (code = first_significant_code(code Line 2233  for (code = first_significant_code(code
2233      NULL. */      NULL. */
2234    
2235      if (cd->start_workspace != NULL)      if (cd->start_workspace != NULL)
2236        {        {
2237        const pcre_uchar *tcode;        const pcre_uchar *tcode;
2238        for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)        for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2239          if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;          if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2240        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2241        }        }
2242    
2243      /* If we are scanning a completed pattern, there are no forward references      /* If we are scanning a completed pattern, there are no forward references
2244      and all groups are complete. We need to detect whether this is a recursive      and all groups are complete. We need to detect whether this is a recursive
2245      call, as otherwise there will be an infinite loop. If it is a recursion,      call, as otherwise there will be an infinite loop. If it is a recursion,
2246      just skip over it. Simple recursions are easily detected. For mutual      just skip over it. Simple recursions are easily detected. For mutual
2247      recursions we keep a chain on the stack. */      recursions we keep a chain on the stack. */
2248    
2249      else      else
2250        {        {
2251        recurse_check *r = recurses;        recurse_check *r = recurses;
2252        const pcre_uchar *endgroup = scode;        const pcre_uchar *endgroup = scode;
2253    
2254        do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);        do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2255        if (code >= scode && code <= endgroup) continue;  /* Simple recursion */        if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2256    
2257        for (r = recurses; r != NULL; r = r->prev)        for (r = recurses; r != NULL; r = r->prev)
2258          if (r->group == scode) break;          if (r->group == scode) break;
2259        if (r != NULL) continue;   /* Mutual recursion */        if (r != NULL) continue;   /* Mutual recursion */
# Line 2436  for (code = first_significant_code(code Line 2264  for (code = first_significant_code(code
2264    
2265      empty_branch = FALSE;      empty_branch = FALSE;
2266      this_recurse.prev = recurses;      this_recurse.prev = recurses;
2267      this_recurse.group = scode;      this_recurse.group = scode;
2268    
2269      do      do
2270        {        {
2271        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
# Line 2557  for (code = first_significant_code(code Line 2385  for (code = first_significant_code(code
2385      case OP_ANY:      case OP_ANY:
2386      case OP_ALLANY:      case OP_ALLANY:
2387      case OP_ANYBYTE:      case OP_ANYBYTE:
2388    
2389      case OP_PROP:      case OP_PROP:
2390      case OP_NOTPROP:      case OP_NOTPROP:
2391      case OP_ANYNL:      case OP_ANYNL:
2392    
2393      case OP_NOT_HSPACE:      case OP_NOT_HSPACE:
2394      case OP_HSPACE:      case OP_HSPACE:
2395      case OP_NOT_VSPACE:      case OP_NOT_VSPACE:
2396      case OP_VSPACE:      case OP_VSPACE:
2397      case OP_EXTUNI:      case OP_EXTUNI:
2398    
2399      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2400      case OP_DIGIT:      case OP_DIGIT:
2401      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2402      case OP_WHITESPACE:      case OP_WHITESPACE:
2403      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2404      case OP_WORDCHAR:      case OP_WORDCHAR:
2405    
2406      case OP_CHAR:      case OP_CHAR:
2407      case OP_CHARI:      case OP_CHARI:
2408      case OP_NOT:      case OP_NOT:
2409      case OP_NOTI:      case OP_NOTI:
2410    
2411      case OP_PLUS:      case OP_PLUS:
2412      case OP_PLUSI:      case OP_PLUSI:
2413      case OP_MINPLUS:      case OP_MINPLUS:
# Line 2589  for (code = first_significant_code(code Line 2417  for (code = first_significant_code(code
2417      case OP_NOTPLUSI:      case OP_NOTPLUSI:
2418      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2419      case OP_NOTMINPLUSI:      case OP_NOTMINPLUSI:
2420    
2421      case OP_POSPLUS:      case OP_POSPLUS:
2422      case OP_POSPLUSI:      case OP_POSPLUSI:
2423      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2424      case OP_NOTPOSPLUSI:      case OP_NOTPOSPLUSI:
2425    
2426      case OP_EXACT:      case OP_EXACT:
2427      case OP_EXACTI:      case OP_EXACTI:
2428      case OP_NOTEXACT:      case OP_NOTEXACT:
2429      case OP_NOTEXACTI:      case OP_NOTEXACTI:
2430    
2431      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2432      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2433      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2434      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2435    
2436      return FALSE;      return FALSE;
2437    
2438      /* These are going to continue, as they may be empty, but we have to      /* These are going to continue, as they may be empty, but we have to
# Line 2644  for (code = first_significant_code(code Line 2472  for (code = first_significant_code(code
2472  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2473      case OP_STAR:      case OP_STAR:
2474      case OP_STARI:      case OP_STARI:
2475      case OP_NOTSTAR:      case OP_NOTSTAR:
2476      case OP_NOTSTARI:      case OP_NOTSTARI:
2477    
2478      case OP_MINSTAR:      case OP_MINSTAR:
2479      case OP_MINSTARI:      case OP_MINSTARI:
2480      case OP_NOTMINSTAR:      case OP_NOTMINSTAR:
2481      case OP_NOTMINSTARI:      case OP_NOTMINSTARI:
2482    
2483      case OP_POSSTAR:      case OP_POSSTAR:
2484      case OP_POSSTARI:      case OP_POSSTARI:
2485      case OP_NOTPOSSTAR:      case OP_NOTPOSSTAR:
2486      case OP_NOTPOSSTARI:      case OP_NOTPOSSTARI:
2487    
2488      case OP_QUERY:      case OP_QUERY:
2489      case OP_QUERYI:      case OP_QUERYI:
2490      case OP_NOTQUERY:      case OP_NOTQUERY:
2491      case OP_NOTQUERYI:      case OP_NOTQUERYI:
2492    
2493      case OP_MINQUERY:      case OP_MINQUERY:
2494      case OP_MINQUERYI:      case OP_MINQUERYI:
2495      case OP_NOTMINQUERY:      case OP_NOTMINQUERY:
2496      case OP_NOTMINQUERYI:      case OP_NOTMINQUERYI:
2497    
2498      case OP_POSQUERY:      case OP_POSQUERY:
2499      case OP_POSQUERYI:      case OP_POSQUERYI:
2500      case OP_NOTPOSQUERY:      case OP_NOTPOSQUERY:
2501      case OP_NOTPOSQUERYI:      case OP_NOTPOSQUERYI:
2502    
2503      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2504      break;      break;
2505    
2506      case OP_UPTO:      case OP_UPTO:
2507      case OP_UPTOI:      case OP_UPTOI:
2508      case OP_NOTUPTO:      case OP_NOTUPTO:
2509      case OP_NOTUPTOI:      case OP_NOTUPTOI:
2510    
2511      case OP_MINUPTO:      case OP_MINUPTO:
2512      case OP_MINUPTOI:      case OP_MINUPTOI:
2513      case OP_NOTMINUPTO:      case OP_NOTMINUPTO:
2514      case OP_NOTMINUPTOI:      case OP_NOTMINUPTOI:
2515    
2516      case OP_POSUPTO:      case OP_POSUPTO:
2517      case OP_POSUPTOI:      case OP_POSUPTOI:
2518      case OP_NOTPOSUPTO:      case OP_NOTPOSUPTO:
2519      case OP_NOTPOSUPTOI:      case OP_NOTPOSUPTOI:
2520    
2521      if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);      if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2522      break;      break;
2523  #endif  #endif
# Line 2753  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;
     {  
     if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)  
       {  
       *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;  
     }  
   }  
 return FALSE;  
 }  
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 */
 *          Check POSIX class name                *  
 *************************************************/  
2648    
2649  /* This function is called to check the name given in a POSIX-style class entry    case PT_ALNUM:
2650  such as [:alnum:].    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2651              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2652    
2653  Arguments:    /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2654    ptr        points to the first letter    means that Perl space and POSIX space are now identical. PCRE was changed
2655    len        the length of the name    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  Returns:     a value representing the name, or -1 if unknown    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  static int    case PT_CLIST:
2670  check_posix_name(const pcre_uchar *ptr, int len)    p = PRIV(ucd_caseless_sets) + prop->caseset;
2671  {    for (;;)
2672  const char *pn = posix_names;      {
2673  register int yield = 0;      if (c < *p) return !negated;
2674  while (posix_name_lengths[yield] != 0)      if (c == *p++) return negated;
2675    {      }
2676    if (len == posix_name_lengths[yield] &&    break;  /* Control never reaches here */
     STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;  
   pn += posix_name_lengths[yield] + 1;  
   yield++;  
2677    }    }
2678  return -1;  
2679    return FALSE;
2680  }  }
2681    #endif  /* SUPPORT_UCP */
2682    
2683    
2684    
2685  /*************************************************  /*************************************************
2686  *    Adjust OP_RECURSE items in repeated group   *  *        Fill the character property list        *
2687  *************************************************/  *************************************************/
2688    
2689  /* OP_RECURSE items contain an offset from the start of the regex to the group  /* Checks whether the code points to an opcode that can take part in auto-
2690  that is referenced. This means that groups can be replicated for fixed  possessification, and if so, fills a list with its properties.
 repetition simply by copying (because the recursion is allowed to refer to  
 earlier groups that are outside the current group). However, when a group is  
 optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is  
 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.  
   
 This function has been extended with the possibility of forward references for  
 recursions and subroutine calls. It must also check the list of such references  
 for the group we are dealing with. If it finds that one of the recursions in  
 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).  
2691    
2692  Arguments:  Arguments:
2693    group      points to the start of the group    code        points to start of expression
2694    adjust     the amount by which the group is to be moved    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2695    utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode    fcc         points to case-flipping table
2696    cd         contains pointers to tables etc.    list        points to output list
2697    save_hwm   the hwm forward reference pointer at the start of the group                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:     nothing  Returns:      points to the start of the next opcode if *code is accepted
2703                  NULL if *code is not accepted
2704  */  */
2705    
2706  static void  static const pcre_uchar *
2707  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,  get_chr_property_list(const pcre_uchar *code, BOOL utf,
2708    pcre_uchar *save_hwm)    const pcre_uint8 *fcc, pcre_uint32 *list)
2709  {  {
2710  pcre_uchar *ptr = group;  pcre_uchar c = *code;
2711    const pcre_uchar *end;
2712    const pcre_uint32 *clist_src;
2713    pcre_uint32 *clist_dest;
2714    pcre_uint32 chr;
2715    pcre_uchar base;
2716    
2717  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)  list[0] = c;
2718    list[1] = FALSE;
2719    code++;
2720    
2721    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2722    {    {
2723    int offset;    base = get_repeat_base(c);
2724    pcre_uchar *hc;    c -= (base - OP_STAR);
2725    
2726    /* See if this recursion is on the forward reference list. If so, adjust the    if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2727    reference. */      code += IMM2_SIZE;
2728    
2729    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2730    
2731      switch(base)
2732      {      {
2733      offset = (int)GET(hc, 0);      case OP_STAR:
2734      if (cd->start_code + offset == ptr + 1)      list[0] = OP_CHAR;
2735        {      break;
       PUT(hc, 0, offset + adjust);  
       break;  
       }  
     }  
2736    
2737    /* Otherwise, adjust the recursion offset if it's after the start of this      case OP_STARI:
2738    group. */      list[0] = OP_CHARI;
2739        break;
2740    
2741    if (hc >= cd->hwm)      case OP_NOTSTAR:
2742      {      list[0] = OP_NOT;
2743      offset = (int)GET(ptr, 1);      break;
     if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);  
     }  
2744    
2745    ptr += 1 + LINK_SIZE;      case OP_NOTSTARI:
2746        list[0] = OP_NOTI;
2747        break;
2748    
2749        case OP_TYPESTAR:
2750        list[0] = *code;
2751        code++;
2752        break;
2753        }
2754      c = list[0];
2755    }    }
 }  
2756    
2757    switch(c)
2758      {
2759      case OP_NOT_DIGIT:
2760      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      case OP_CHAR:
2780      case OP_NOT:
2781      GETCHARINCTEST(chr, code);
2782      list[2] = chr;
2783      list[3] = NOTACHAR;
2784      return code;
2785    
2786  /*************************************************    case OP_CHARI:
2787  *        Insert an automatic callout point       *    case OP_NOTI:
2788  *************************************************/    list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2789      GETCHARINCTEST(chr, code);
2790      list[2] = chr;
2791    
2792  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert  #ifdef SUPPORT_UCP
2793  callout points before each pattern item.    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  Arguments:    /* The othercase might be the same value. */
   code           current code pointer  
   ptr            current pattern pointer  
   cd             pointers to tables etc  
2804    
2805  Returns:         new code pointer    if (chr == list[3])
2806  */      list[3] = NOTACHAR;
2807      else
2808        list[4] = NOTACHAR;
2809      return code;
2810    
2811  static pcre_uchar *  #ifdef SUPPORT_UCP
2812  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)    case OP_PROP:
2813  {    case OP_NOTPROP:
2814  *code++ = OP_CALLOUT;    if (code[0] != PT_CLIST)
2815  *code++ = 255;      {
2816  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */      list[2] = code[0];
2817  PUT(code, LINK_SIZE, 0);                       /* Default length */      list[3] = code[1];
2818  return code + 2 * LINK_SIZE;      return code + 2;
2819  }      }
2820    
2821      /* Convert only if we have anough space. */
2822    
2823      clist_src = PRIV(ucd_caseless_sets) + code[1];
2824      clist_dest = list + 2;
2825      code += 2;
2826    
2827  /*************************************************    do {
2828  *         Complete a callout item                *       /* 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  /* A callout item contains the length of the next item in the pattern, which    /* Enough space to store all characters. */
 we can't fill in till after we have reached the relevant point. This is used  
 for both automatic and manual callouts.  
2840    
2841  Arguments:    list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
2842    previous_callout   points to previous callout item    return code;
2843    ptr                current pattern pointer  #endif
   cd                 pointers to tables etc  
2844    
2845  Returns:             nothing    case OP_NCLASS:
2846  */    case OP_CLASS:
2847    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2848      case OP_XCLASS:
2849    
2850  static void    if (c == OP_XCLASS)
2851  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)      end = code + GET(code, 0);
2852  {    else
2853  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));  #endif
2854  PUT(previous_callout, 2 + LINK_SIZE, length);      end = code + 32 / sizeof(pcre_uchar);
2855    
2856      switch(*end)
2857        {
2858        case OP_CRSTAR:
2859        case OP_CRMINSTAR:
2860        case OP_CRQUERY:
2861        case OP_CRMINQUERY:
2862        list[1] = TRUE;
2863        end++;
2864        break;
2865    
2866        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    
 #ifdef SUPPORT_UCP  
2880  /*************************************************  /*************************************************
2881  *           Get othercase range                  *  *    Scan further character sets for match       *
2882  *************************************************/  *************************************************/
2883    
2884  /* This function is passed the start and end of a class range, in UTF-8 mode  /* Checks whether the base and the current opcode have a common character, in
2885  with UCP support. It searches up the characters, looking for ranges of  which case the base cannot be possessified.
 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.  
2886    
2887  Arguments:  Arguments:
2888    cptr        points to starting character value; updated    code        points to the byte code
2889    d           end value    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2890    ocptr       where to put start of othercase range    cd          static compile data
2891    odptr       where to put end of othercase range    base_list   the data list of the base opcode
2892    
2893  Yield:        -1 when no more  Returns:      TRUE if the auto-possessification is possible
                0 when a range is returned  
               >0 the CASESET offset for char with multiple other cases  
                 in this case, ocptr contains the original  
2894  */  */
2895    
2896  static int  static BOOL
2897  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,  compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
2898    pcre_uint32 *odptr)    const pcre_uint32* base_list)
2899  {  {
2900  pcre_uint32 c, othercase, next;  pcre_uchar c;
2901  unsigned int co;  pcre_uint32 list[8];
2902    const pcre_uint32* chr_ptr;
2903  /* Find the first character that has an other case. If it has multiple other  const pcre_uint32* ochr_ptr;
2904  cases, return its case offset value. */  const pcre_uint32* list_ptr;
2905    pcre_uint32 chr;
2906    
2907  for (c = *cptr; c <= d; c++)  for(;;)
2908    {    {
2909    if ((co = UCD_CASESET(c)) != 0)    c = *code;
2910    
2911      /* Skip over callouts */
2912    
2913      if (c == OP_CALLOUT)
2914      {      {
2915      *ocptr = c++;   /* Character that has the set */      code += PRIV(OP_lengths)[c];
2916      *cptr = c;      /* Rest of input range */      continue;
     return (int)co;  
2917      }      }
   if ((othercase = UCD_OTHERCASE(c)) != c) break;  
   }  
2918    
2919  if (c > d) return -1;  /* Reached end of range */    if (c == OP_ALT)
2920        {
2921        do code += GET(code, 1); while (*code == OP_ALT);
2922        c = *code;
2923        }
2924    
2925  *ocptr = othercase;    switch(c)
2926  next = othercase + 1;      {
2927        case OP_END:
2928        /* 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  for (++c; c <= d; c++)      return base_list[1] != 0;
   {  
   if (UCD_OTHERCASE(c) != next) break;  
   next++;  
   }  
2933    
2934  *odptr = next - 1;     /* End of othercase range */      case OP_KET:
2935  *cptr = c;             /* Rest of input range */      /* If the bracket is capturing, and referenced by an OP_RECURSE, the
2936  return 0;      non-greedy case cannot be converted to a possessive form. We do not test
2937  }      the bracket type at the moment, but we might do it in the future to improve
2938        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  /*************************************************    code = get_chr_property_list(code, utf, cd->fcc, list);
2948  *        Check a character and a property        *    if (code == NULL) return FALSE;    /* Unsupported */
 *************************************************/  
2949    
2950  /* This function is called by check_auto_possessive() when a property item    /* If either opcode is a small character list, set pointers for comparing
2951  is adjacent to a fixed character.    characters from that list with another list, or with a property. */
2952    
2953  Arguments:    if (base_list[0] == OP_CHAR)
2954    c            the character      {
2955    ptype        the property type      chr_ptr = base_list + 2;
2956    pdata        the data for the type      list_ptr = list;
2957    negated      TRUE if it's a negated property (\P or \p{^)      }
2958      else if (list[0] == OP_CHAR)
2959        {
2960        chr_ptr = list + 2;
2961        list_ptr = base_list;
2962        }
2963    
2964  Returns:       TRUE if auto-possessifying is OK    /* Some property combinations also acceptable. Unicode property opcodes are
2965  */    processed specially; the rest can be handled with a lookup table. */
2966    
2967      else
2968        {
2969        pcre_uint32 leftop, rightop;
2970    
2971        if (list[1] != 0) return FALSE;   /* Must match at least one character */
2972        leftop = base_list[0];
2973        rightop = list[0];
2974    
 static BOOL  
 check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata, BOOL negated)  
 {  
2975  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2976  const pcre_uint32 *p;      if (leftop == OP_PROP || leftop == OP_NOTPROP)
2977  #endif        {
2978          if (rightop == OP_EOD) return TRUE;
2979          if (rightop == OP_PROP || rightop == OP_NOTPROP)
2980            {
2981            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  const ucd_record *prop = GET_UCD(c);      else
3071    #endif  /* SUPPORT_UCP */
3072    
3073  switch(ptype)      return leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3074    {             rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3075    case PT_LAMP:             autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3076    return (prop->chartype == ucp_Lu ||      }
           prop->chartype == ucp_Ll ||  
           prop->chartype == ucp_Lt) == negated;  
3077    
3078    case PT_GC:    /* Control reaches here only if one of the items is a small character list.
3079    return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;    All characters are checked against the other side. */
3080    
3081    case PT_PC:    do
3082    return (pdata == prop->chartype) == negated;      {
3083        chr = *chr_ptr;
3084    
3085    case PT_SC:      switch(list_ptr[0])
3086    return (pdata == prop->script) == negated;        {
3087          case OP_CHAR:
3088          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    /* These are specials */        case OP_NOT:
3098          ochr_ptr = list_ptr + 2;
3099          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    case PT_ALNUM:        /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3110    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||        set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
           PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;  
3111    
3112    case PT_SPACE:    /* Perl space */        case OP_DIGIT:
3113    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||        if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3114            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)        break;
           == negated;  
3115    
3116    case PT_PXSPACE:  /* POSIX space */        case OP_NOT_DIGIT:
3117    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||        if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3118            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||        break;
           c == CHAR_FF || c == CHAR_CR)  
           == negated;  
3119    
3120    case PT_WORD:        case OP_WHITESPACE:
3121    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||        if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3122            PRIV(ucp_gentype)[prop->chartype] == ucp_N ||        break;
           c == CHAR_UNDERSCORE) == negated;  
3123    
3124  #ifdef SUPPORT_UCP        case OP_NOT_WHITESPACE:
3125    case PT_CLIST:        if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3126    p = PRIV(ucd_caseless_sets) + prop->caseset;        break;
3127    for (;;)  
3128      {        case OP_WORDCHAR:
3129      if (c < *p) return !negated;        if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3130      if (c == *p++) return negated;        break;
3131      }  
3132    break;  /* Control never reaches here */        case OP_NOT_WORDCHAR:
3133          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3134          break;
3135    
3136          case OP_HSPACE:
3137          switch(chr)
3138            {
3139            HSPACE_CASES: return FALSE;
3140            default: break;
3141            }
3142          break;
3143    
3144          case OP_NOT_HSPACE:
3145          switch(chr)
3146            {
3147            HSPACE_CASES: break;
3148            default: return FALSE;
3149            }
3150          break;
3151    
3152          case OP_ANYNL:
3153          case OP_VSPACE:
3154          switch(chr)
3155            {
3156            VSPACE_CASES: return FALSE;
3157            default: break;
3158            }
3159          break;
3160    
3161          case OP_NOT_VSPACE:
3162          switch(chr)
3163            {
3164            VSPACE_CASES: break;
3165            default: return FALSE;
3166            }
3167          break;
3168    
3169          case OP_DOLL:
3170          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          case OP_EOD:    /* Can always possessify before \z */
3187          break;
3188    
3189          case OP_PROP:
3190          case OP_NOTPROP:
3191          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3192                list_ptr[0] == OP_NOTPROP))
3193            return FALSE;
3194          break;
3195    
3196          /* The class comparisons work only when the class is the second item
3197          of the pair, because there are at present no possessive forms of the
3198          class opcodes. Note also that the "code" variable that is used below
3199          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    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3215          case OP_XCLASS:
3216          if (list_ptr != list) return FALSE;   /* Class is first opcode */
3217          if (PRIV(xclass)(chr, code - list_ptr[2] + 1 + LINK_SIZE, utf))
3218            return FALSE;
3219          break;
3220  #endif  #endif
3221    
3222          default:
3223          return FALSE;
3224          }
3225    
3226        chr_ptr++;
3227        }
3228      while(*chr_ptr != NOTACHAR);
3229    
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        case OP_TYPESTAR:
3319        case OP_TYPEMINSTAR:
3320        case OP_TYPEPLUS:
3321        case OP_TYPEMINPLUS:
3322        case OP_TYPEQUERY:
3323        case OP_TYPEMINQUERY:
3324        case OP_TYPEPOSSTAR:
3325        case OP_TYPEPOSPLUS:
3326        case OP_TYPEPOSQUERY:
3327        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3328        break;
3329    
3330  if (*ptr == CHAR_BACKSLASH)      case OP_TYPEUPTO:
3331    {      case OP_TYPEMINUPTO:
3332    int temperrorcode = 0;      case OP_TYPEEXACT:
3333    escape = check_escape(&ptr, &next, &temperrorcode, cd->bracount, options,      case OP_TYPEPOSUPTO:
3334      FALSE);      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3335    if (temperrorcode != 0) return FALSE;        code += 2;
3336    ptr++;    /* Point after the escape sequence */      break;
   }  
 else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)  
   {  
   escape = 0;  
 #ifdef SUPPORT_UTF  
   if (utf) { GETCHARINC(next, ptr); } else  
 #endif  
   next = *ptr++;  
   }  
 else return FALSE;  
3337    
3338  /* Skip whitespace and comments in extended mode */      case OP_XCLASS:
3339        code += GET(code, 1);
3340        break;
3341    
3342  if ((options & PCRE_EXTENDED) != 0)      case OP_MARK:
3343    {      case OP_PRUNE_ARG:
3344    for (;;)      case OP_SKIP_ARG:
3345      {      case OP_THEN_ARG:
3346      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      code += code[1];
3347      if (*ptr == CHAR_NUMBER_SIGN)      break;
       {  
       ptr++;  
       while (*ptr != CHAR_NULL)  
         {  
         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      case OP_NOT_WHITESPACE:  /* This function is called to check the name given in a POSIX-style class entry
3505      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;  such as [:alnum:].
3506    
3507      case OP_WORDCHAR:  Arguments:
3508      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;    ptr        points to the first letter
3509      len        the length of the name
3510    
3511      case OP_NOT_WORDCHAR:  Returns:     a value representing the name, or -1 if unknown
3512      return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;  */
3513    
3514      case OP_HSPACE:  static int
3515      case OP_NOT_HSPACE:  check_posix_name(const pcre_uchar *ptr, int len)
3516      switch(next)  {
3517        {  const char *pn = posix_names;
3518        HSPACE_CASES:  register int yield = 0;
3519        return op_code == OP_NOT_HSPACE;  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    
       default:  
       return op_code != OP_NOT_HSPACE;  
       }  
3529    
3530      case OP_ANYNL:  /*************************************************
3531      case OP_VSPACE:  *    Adjust OP_RECURSE items in repeated group   *
3532      case OP_NOT_VSPACE:  *************************************************/
     switch(next)  
       {  
       VSPACE_CASES:  
       return op_code == OP_NOT_VSPACE;  
3533    
3534        default:  /* OP_RECURSE items contain an offset from the start of the regex to the group
3535        return op_code != OP_NOT_VSPACE;  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  #ifdef SUPPORT_UCP  This function has been extended with the possibility of forward references for
3546      case OP_PROP:  recursions and subroutine calls. It must also check the list of such references
3547      return check_char_prop(next, previous[0], previous[1], FALSE);  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_NOTPROP:  Arguments:
3552      return check_char_prop(next, previous[0], previous[1], TRUE);    group      points to the start of the group
3553  #endif    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      default:  Returns:     nothing
3559      return FALSE;  */
     }  
   }  
3560    
3561  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP  static void
3562  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3563  generated only when PCRE_UCP is *not* set, that is, when only ASCII    pcre_uchar *save_hwm)
3564  characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are  {
3565  replaced by OP_PROP codes when PCRE_UCP is set. */  pcre_uchar *ptr = group;
3566    
3567  switch(op_code)  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3568    {    {
3569    case OP_CHAR:    int offset;
3570    case OP_CHARI:    pcre_uchar *hc;
3571    switch(escape)  
3572      {    /* See if this recursion is on the forward reference list. If so, adjust the
3573      case ESC_d:    reference. */
     return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;  
3574    
3575      case ESC_D:    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3576      return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;      {
3577        offset = (int)GET(hc, 0);
3578        if (cd->start_code + offset == ptr + 1)
3579          {
3580          PUT(hc, 0, offset + adjust);
3581          break;
3582          }
3583        }
3584    
3585      case ESC_s:    /* Otherwise, adjust the recursion offset if it's after the start of this
3586      return c > 255 || (cd->ctypes[c] & ctype_space) == 0;    group. */
3587    
3588      case ESC_S:    if (hc >= cd->hwm)
3589      return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;      {
3590        offset = (int)GET(ptr, 1);
3591        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3592        }
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 3754  to find out the amount of memory needed, Line 3961  to find out the amount of memory needed,
3961  phase. The value of lengthptr distinguishes the two phases.  phase. The value of lengthptr distinguishes the two phases.
3962    
3963  Arguments:  Arguments:
3964    optionsptr     pointer to the option bits    optionsptr        pointer to the option bits
3965    codeptr        points to the pointer to the current code point    codeptr           points to the pointer to the current code point
3966    ptrptr         points to the current pattern pointer    ptrptr            points to the current pattern pointer
3967    errorcodeptr   points to error code variable    errorcodeptr      points to error code variable
3968    firstcharptr    place to put the first required character    firstcharptr      place to put the first required character
3969    firstcharflagsptr place to put the first character flags, or a negative number    firstcharflagsptr place to put the first character flags, or a negative number
3970    reqcharptr     place to put the last required character    reqcharptr        place to put the last required character
3971    reqcharflagsptr place to put the last required character flags, or a negative number    reqcharflagsptr   place to put the last required character flags, or a negative number
3972    bcptr          points to current branch chain    bcptr             points to current branch chain
3973    cond_depth     conditional nesting depth    cond_depth        conditional nesting depth
3974    cd             contains pointers to tables etc.    cd                contains pointers to tables etc.
3975    lengthptr      NULL during the real compile phase    lengthptr         NULL during the real compile phase
3976                   points to length accumulator during pre-compile phase                      points to length accumulator during pre-compile phase
3977    
3978  Returns:         TRUE on success  Returns:            TRUE on success
3979                   FALSE, with *errorcodeptr set non-zero on error                      FALSE, with *errorcodeptr set non-zero on error
3980  */  */
3981    
3982  static BOOL  static BOOL
# Line 4420  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 4933  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 4963  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 5119  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 5846  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 5863  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 5880  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 5890  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 5902  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 5937  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 5949  for (;; ptr++) Line 6140  for (;; ptr++)
6140            slot += cd->name_entry_size;            slot += cd->name_entry_size;
6141            }            }
6142    
6143          /* Found a previous 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          /* Search the pattern for a forward reference */              slot += cd->name_entry_size;
6156                if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) != 0) break;
6157          else if ((i = find_parens(cd, name, namelen,              count++;
6158                          (options & PCRE_EXTENDED) != 0, utf)) > 0)              }
6159            {            if (count > 1)
6160            PUT2(code, 2+LINK_SIZE, i);              {
6161            code[1+LINK_SIZE]++;              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 6130  for (;; ptr++) Line 6332  for (;; ptr++)
6332          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
6333          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
6334          case CHAR_APOSTROPHE:          case CHAR_APOSTROPHE:
6335            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
6336              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
6337            name = ++ptr;
6338    
6339            while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
6340            namelen = (int)(ptr - name);
6341    
6342            /* 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
6344            necessary. Duplicates for the same number are skipped; other duplicates
6345            are checked for validity. In the actual compile, there is nothing to
6346            do. */
6347    
6348            if (lengthptr != NULL)
6349            {            {
6350            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?            named_group *ng;
6351              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;            pcre_uint32 number = cd->bracount + 1;
           name = ++ptr;  
6352    
6353            while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            if (*ptr != (pcre_uchar)terminator)
6354            namelen = (int)(ptr - name);              {
6355                *errorcodeptr = ERR42;
6356                goto FAILED;
6357                }
6358    
6359            /* In the pre-compile phase, just do a syntax check. */            if (cd->names_found >= MAX_NAME_COUNT)
6360                {
6361                *errorcodeptr = ERR49;
6362                goto FAILED;
6363                }
6364    
6365            if (lengthptr != NULL)            if (namelen + IMM2_SIZE + 1 > cd->name_entry_size)
6366              {              {
6367              if (*ptr != (pcre_uchar)terminator)              cd->name_entry_size = namelen + IMM2_SIZE + 1;
6368                {              if (namelen > MAX_NAME_SIZE)
               *errorcodeptr = ERR42;  
               goto FAILED;  
               }  
             if (cd->names_found >= MAX_NAME_COUNT)  
6369                {                {
6370                *errorcodeptr = ERR49;                *errorcodeptr = ERR48;
6371                goto FAILED;                goto FAILED;
6372                }                }
6373              if (namelen + IMM2_SIZE + 1 > cd->name_entry_size)              }
6374    
6375              /* 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
6377              discarded; otherwise, if DUPNAMES is not set, give an error.
6378              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
6380              non-duplicate names, give an error if the number is duplicated. */
6381    
6382              ng = cd->named_groups;
6383              for (i = 0; i < cd->names_found; i++, ng++)
6384                {
6385                if (namelen == ng->length &&
6386                    STRNCMP_UC_UC(name, ng->name, namelen) == 0)
6387                {                {
6388                cd->name_entry_size = namelen + IMM2_SIZE + 1;                if (ng->number == number) break;
6389                if (namelen > MAX_NAME_SIZE)                if ((options & PCRE_DUPNAMES) == 0)
6390                  {                  {
6391                  *errorcodeptr = ERR48;                  *errorcodeptr = ERR43;
6392                  goto FAILED;                  goto FAILED;
6393                  }                  }
6394                  cd->dupnames = TRUE;  /* Duplicate names exist */
6395                  }
6396                else if (ng->number == number)
6397                  {
6398                  *errorcodeptr = ERR65;
6399                  goto FAILED;
6400                }                }
6401              }              }
6402    
6403            /* In the real compile, create the entry in the table, maintaining            if (i >= cd->names_found)     /* Not a duplicate with same number */
           alphabetical order. Duplicate names for different numbers are  
           permitted only if PCRE_DUPNAMES is set. Duplicate names for the same  
           number are always OK. (An existing number can be re-used if (?|  
           appears in the pattern.) In either event, a duplicate name results in  
           a duplicate entry in the table, even if the number is the same. This  
           is because the number of names, and hence the table size, is computed  
           in the pre-compile, and it affects various numbers and pointers which  
           would all have to be modified, and the compiled code moved down, if  
           duplicates with the same number were omitted from the table. This  
           doesn't seem worth the hassle. However, *different* names for the  
           same number are not permitted. */  
   
           else  
6404              {              {
6405              BOOL dupname = FALSE;              /* Increase the list size if necessary */
             slot = cd->name_table;  
6406    
6407              for (i = 0; i < cd->names_found; i++)              if (cd->names_found >= cd->named_group_list_size)
6408                {                {
6409                int crc = memcmp(name, slot+IMM2_SIZE, IN_UCHARS(namelen));                int newsize = cd->named_group_list_size * 2;
6410                if (crc == 0)                named_group *newspace = (PUBL(malloc))
6411                  {                  (newsize * sizeof(named_group));
                 if (slot[IMM2_SIZE+namelen] == 0)  
                   {  
                   if (GET2(slot, 0) != cd->bracount + 1 &&  
                       (options & PCRE_DUPNAMES) == 0)  
                     {  
                     *errorcodeptr = ERR43;  
                     goto FAILED;  
                     }  
                   else dupname = TRUE;  
                   }  
                 else crc = -1;      /* Current name is a substring */  
                 }  
   
               /* Make space in the table and break the loop for an earlier  
               name. For a duplicate or later name, carry on. We do this for  
               duplicates so that in the simple case (when ?(| is not used) they  
               are in order of their numbers. */  
6412    
6413                if (crc < 0)                if (newspace == NULL)
6414                  {                  {
6415                  memmove(slot + cd->name_entry_size, slot,                  *errorcodeptr = ERR21;
6416                    IN_UCHARS((cd->names_found - i) * cd->name_entry_size));                  goto FAILED;
                 break;  
6417                  }                  }
6418    
6419                /* Continue the loop for a later or duplicate name */                memcpy(newspace, cd->named_groups,
6420                    cd->named_group_list_size * sizeof(named_group));
6421                slot += cd->name_entry_size;                if (cd->named_group_list_size > NAMED_GROUP_LIST_SIZE)
6422                }                  (PUBL(free))((void *)cd->named_groups);
6423                  cd->named_groups = newspace;
6424              /* For non-duplicate names, check for a duplicate number before                cd->named_group_list_size = newsize;
             adding the new name. */  
   
             if (!dupname)  
               {  
               pcre_uchar *cslot = cd->name_table;  
               for (i = 0; i < cd->names_found; i++)  
                 {  
                 if (cslot != slot)  
                   {  
                   if (GET2(cslot, 0) == cd->bracount + 1)  
                     {  
                     *errorcodeptr = ERR65;  
                     goto FAILED;  
                     }  
                   }  
                 else i--;  
                 cslot += cd->name_entry_size;  
                 }  
6425                }                }
6426    
6427              PUT2(slot, 0, cd->bracount + 1);              cd->named_groups[cd->names_found].name = name;
6428              memcpy(slot + IMM2_SIZE, name, IN_UCHARS(namelen));              cd->named_groups[cd->names_found].length = namelen;
6429              slot[IMM2_SIZE + namelen] = 0;              cd->named_groups[cd->names_found].number = number;
6430                cd->names_found++;
6431              }              }
6432            }            }
6433    
6434          /* In both pre-compile and compile, count the number of names we've          ptr++;                    /* Move past > or ' in both passes. */
         encountered. */  
   
         cd->names_found++;  
         ptr++;                    /* Move past > or ' */  
6435          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
6436    
6437    
# Line 6277  for (;; ptr++) Line 6461  for (;; ptr++)
6461    
6462          if (lengthptr != NULL)          if (lengthptr != NULL)
6463            {            {
6464            const pcre_uchar *temp;            named_group *ng;
6465    
6466            if (namelen == 0)            if (namelen == 0)
6467              {              {
# Line 6295  for (;; ptr++) Line 6479  for (;; ptr++)
6479              goto FAILED;              goto FAILED;
6480              }              }
6481    
6482            /* The name table does not exist in the first pass, so we cannot            /* The name table does not exist in the first pass; instead we must
6483            do a simple search as in the code below. Instead, we have to scan the            scan the list of names encountered so far in order to get the
6484            pattern to find the number. It is important that we scan it only as            number. If the name is not found, set the value to 0 for a forward
6485            far as we have got because the syntax of named subpatterns has not            reference. */
6486            been checked for the rest of the pattern, and find_parens() assumes  
6487            correct syntax. In any case, it's a waste of resources to scan            ng = cd->named_groups;
6488            further. We stop the scan at the current point by temporarily            for (i = 0; i < cd->names_found; i++, ng++)
6489            adjusting the value of cd->endpattern. */              {
6490                if (namelen == ng->length &&
6491            temp = cd->end_pattern;                  STRNCMP_UC_UC(name, ng->name, namelen) == 0)
6492            cd->end_pattern = ptr;                break;
6493            recno = find_parens(cd, name, namelen,              }
6494              (options & PCRE_EXTENDED) != 0, utf);            recno = (i < cd->names_found)? ng->number : 0;
6495            cd->end_pattern = temp;  
6496            if (recno < 0) recno = 0;    /* Forward ref; set dummy number */            /* Count named back references. */
6497    
6498              if (!is_recurse) cd->namedrefcount++;
6499            }            }
6500    
6501          /* In the real compile, seek the name in the table. We check the name          /* In the real compile, search the name table. We check the name
6502          first, and then check that we have reached the end of the name in the          first, and then check that we have reached the end of the name in the
6503          table. That way, if the name that is longer than any in the table,          table. That way, if the name is longer than any in the table, the
6504          the comparison will fail without reading beyond the table entry. */          comparison will fail without reading beyond the table entry. */
6505    
6506          else          else
6507            {            {
# Line 6328  for (;; ptr++) Line 6514  for (;; ptr++)
6514              slot += cd->name_entry_size;              slot += cd->name_entry_size;
6515              }              }
6516    
6517            if (i < cd->names_found)         /* Back reference */            if (i < cd->names_found)
6518              {              {
6519              recno = GET2(slot, 0);              recno = GET2(slot, 0);
6520              }              }
6521            else if ((recno =                /* Forward back reference */            else
                     find_parens(cd, name, namelen,  
                       (options & PCRE_EXTENDED) != 0, utf)) <= 0)  
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 6444  for (;; ptr++) Line 6682  for (;; ptr++)
6682    
6683              if (called == NULL)              if (called == NULL)
6684                {                {
6685                if (find_parens(cd, NULL, recno,                if (recno > cd->final_bracount)
                     (options & PCRE_EXTENDED) != 0, utf) < 0)  
6686                  {                  {
6687                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
6688                  goto FAILED;                  goto FAILED;
# Line 6929  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 7058  for (;; ptr++) Line 7298  for (;; ptr++)
7298          *code++ = OP_PROP;          *code++ = OP_PROP;
7299          *code++ = PT_CLIST;          *code++ = PT_CLIST;
7300          *code++ = c;          *code++ = c;
7301          if (firstcharflags == REQ_UNSET) firstcharflags = zerofirstcharflags = REQ_NONE;          if (firstcharflags == REQ_UNSET)
7302              firstcharflags = zerofirstcharflags = REQ_NONE;
7303          break;          break;
7304          }          }
7305        }        }
# Line 7147  out the amount of memory needed, as well Line 7388  out the amount of memory needed, as well
7388  value of lengthptr distinguishes the two phases.  value of lengthptr distinguishes the two phases.
7389    
7390  Arguments:  Arguments:
7391    options        option bits, including any changes for this subpattern    options           option bits, including any changes for this subpattern
7392    codeptr        -> the address of the current code pointer    codeptr           -> the address of the current code pointer
7393    ptrptr         -> the address of the current pattern pointer    ptrptr            -> the address of the current pattern pointer
7394    errorcodeptr   -> pointer to error code variable    errorcodeptr      -> pointer to error code variable
7395    lookbehind     TRUE if this is a lookbehind assertion    lookbehind        TRUE if this is a lookbehind assertion
7396    reset_bracount TRUE to reset the count for each branch    reset_bracount    TRUE to reset the count for each branch
7397    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes         skip this many bytes at start (for brackets and OP_COND)
7398    cond_depth     depth of nesting for conditional subpatterns    cond_depth        depth of nesting for conditional subpatterns
7399    firstcharptr    place to put the first required character    firstcharptr      place to put the first required character
7400    firstcharflagsptr place to put the first character flags, or a negative number    firstcharflagsptr place to put the first character flags, or a negative number
7401    reqcharptr     place to put the last required character    reqcharptr        place to put the last required character
7402    reqcharflagsptr place to put the last required character flags, or a negative number    reqcharflagsptr   place to put the last required character flags, or a negative number
7403    bcptr          pointer to the chain of currently open branches    bcptr             pointer to the chain of currently open branches
7404    cd             points to the data block with tables pointers etc.    cd                points to the data block with tables pointers etc.
7405    lengthptr      NULL during the real compile phase    lengthptr         NULL during the real compile phase
7406                   points to length accumulator during pre-compile phase                      points to length accumulator during pre-compile phase
7407    
7408  Returns:         TRUE on success  Returns:            TRUE on success
7409  */  */
7410    
7411  static BOOL  static BOOL
# Line 7615  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 7701  return TRUE; Line 7942  return TRUE;
7942  discarded, because they can cause conflicts with actual literals that follow.  discarded, because they can cause conflicts with actual literals that follow.
7943  However, if we end up without a first char setting for an unanchored pattern,  However, if we end up without a first char setting for an unanchored pattern,
7944  it is worth scanning the regex to see if there is an initial asserted first  it is worth scanning the regex to see if there is an initial asserted first
7945  char. If all branches start with the same asserted char, or with a bracket all  char. If all branches start with the same asserted char, or with a
7946  of whose alternatives start with the same asserted char (recurse ad lib), then  non-conditional bracket all of whose alternatives start with the same asserted
7947  we return that char, otherwise -1.  char (recurse ad lib), then we return that char, with the flags set to zero or
7948    REQ_CASELESS; otherwise return zero with REQ_NONE in the flags.
7949    
7950  Arguments:  Arguments:
7951    code       points to start of expression (the bracket)    code       points to start of expression (the bracket)
7952    flags       points to the first char flags, or to REQ_NONE    flags      points to the first char flags, or to REQ_NONE
7953    inassert   TRUE if in an assertion    inassert   TRUE if in an assertion
7954    
7955  Returns:     the fixed first char, or 0 with REQ_NONE in flags  Returns:     the fixed first char, or 0 with REQ_NONE in flags
# Line 7744  do { Line 7986  do {
7986       case OP_ASSERT:       case OP_ASSERT:
7987       case OP_ONCE:       case OP_ONCE:
7988       case OP_ONCE_NC:       case OP_ONCE_NC:
      case OP_COND:  
7989       d = find_firstassertedchar(scode, &dflags, op == OP_ASSERT);       d = find_firstassertedchar(scode, &dflags, op == OP_ASSERT);
7990       if (dflags < 0)       if (dflags < 0)
7991         return 0;         return 0;
# Line 7789  return c; Line 8030  return c;
8030    
8031    
8032  /*************************************************  /*************************************************
8033    *     Add an entry to the name/number table      *
8034    *************************************************/
8035    
8036    /* This function is called between compiling passes to add an entry to the
8037    name/number table, maintaining alphabetical order. Checking for permitted
8038    and forbidden duplicates has already been done.
8039    
8040    Arguments:
8041      cd           the compile data block
8042      name         the name to add
8043      length       the length of the name
8044      groupno      the group number
8045    
8046    Returns:       nothing
8047    */
8048    
8049    static void
8050    add_name(compile_data *cd, const pcre_uchar *name, int length,
8051      unsigned int groupno)
8052    {
8053    int i;
8054    pcre_uchar *slot = cd->name_table;
8055    
8056    for (i = 0; i < cd->names_found; i++)
8057      {
8058      int crc = memcmp(name, slot+IMM2_SIZE, IN_UCHARS(length));
8059      if (crc == 0 && slot[IMM2_SIZE+length] != 0)
8060        crc = -1; /* Current name is a substring */
8061    
8062      /* Make space in the table and break the loop for an earlier name. For a
8063      duplicate or later name, carry on. We do this for duplicates so that in the
8064      simple case (when ?(| is not used) they are in order of their numbers. In all
8065      cases they are in the order in which they appear in the pattern. */
8066    
8067      if (crc < 0)
8068        {
8069        memmove(slot + cd->name_entry_size, slot,
8070          IN_UCHARS((cd->names_found - i) * cd->name_entry_size));
8071        break;
8072        }
8073    
8074      /* Continue the loop for a later or duplicate name */
8075    
8076      slot += cd->name_entry_size;
8077      }
8078    
8079    PUT2(slot, 0, groupno);
8080    memcpy(slot + IMM2_SIZE, name, IN_UCHARS(length));
8081    slot[IMM2_SIZE + length] = 0;
8082    cd->names_found++;
8083    }
8084    
8085    
8086    
8087    /*************************************************
8088  *        Compile a Regular Expression            *  *        Compile a Regular Expression            *
8089  *************************************************/  *************************************************/
8090    
# Line 7875  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
8175    similar way to cworkspace, it can be expanded using malloc() if necessary. */
8176    
8177    named_group named_groups[NAMED_GROUP_LIST_SIZE];
8178    
8179  /* Set this early so that early errors get offset 0. */  /* Set this early so that early errors get offset 0. */
8180    
8181  ptr = (const pcre_uchar *)pattern;  ptr = (const pcre_uchar *)pattern;
# Line 8137  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;
8446  cd->workspace_size = COMPILE_WORK_SIZE;  cd->workspace_size = COMPILE_WORK_SIZE;
8447    cd->named_groups = named_groups;
8448    cd->named_group_list_size = NAMED_GROUP_LIST_SIZE;
8449  cd->start_pattern = (const pcre_uchar *)pattern;  cd->start_pattern = (const pcre_uchar *)pattern;
8450  cd->end_pattern = (const pcre_uchar *)(pattern + STRLEN_UC((const pcre_uchar *)pattern));  cd->end_pattern = (const pcre_uchar *)(pattern + STRLEN_UC((const pcre_uchar *)pattern));
8451  cd->req_varyopt = 0;  cd->req_varyopt = 0;
# Line 8172  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 8223  cd->final_bracount = cd->bracount;  /* S Line 8537  cd->final_bracount = cd->bracount;  /* S
8537  cd->assert_depth = 0;  cd->assert_depth = 0;
8538  cd->bracount = 0;  cd->bracount = 0;
8539  cd->max_lookbehind = 0;  cd->max_lookbehind = 0;
 cd->names_found = 0;  
8540  cd->name_table = (pcre_uchar *)re + re->name_table_offset;  cd->name_table = (pcre_uchar *)re + re->name_table_offset;
8541  codestart = cd->name_table + re->name_entry_size * re->name_count;  codestart = cd->name_table + re->name_entry_size * re->name_count;
8542  cd->start_code = codestart;  cd->start_code = codestart;
# Line 8234  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
8551    created in the first pass. */
8552    
8553    if (cd->names_found > 0)
8554      {
8555      int i = cd->names_found;
8556      named_group *ng = cd->named_groups;
8557      cd->names_found = 0;
8558      for (; i > 0; i--, ng++)
8559        add_name(cd, ng->name, ng->length, ng->number);
8560      if (cd->named_group_list_size > NAMED_GROUP_LIST_SIZE)
8561        (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
8566  of the function here. */  of the function here. */
# Line 8297  if (cd->hwm > cd->start_workspace) Line 8624  if (cd->hwm > cd->start_workspace)
8624      }      }
8625    }    }
8626    
8627  /* If the workspace had to be expanded, free the new memory. Set the pointer to  /* If the workspace had to be expanded, free the new memory. Set the pointer to
8628  NULL to indicate that forward references have been filled in. */  NULL to indicate that forward references have been filled in. */
8629    
8630  if (cd->workspace_size > COMPILE_WORK_SIZE)  if (cd->workspace_size > COMPILE_WORK_SIZE)
8631    (PUBL(free))((void *)cd->start_workspace);    (PUBL(free))((void *)cd->start_workspace);
8632  cd->start_workspace = NULL;  cd->start_workspace = NULL;
8633    
8634  /* Give an error if there's back reference to a non-existent capturing  /* Give an error if there's back reference to a non-existent capturing
8635  subpattern. */  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
# Line 8506  if (code - codestart > length) Line 8839  if (code - codestart > length)
8839    }    }
8840  #endif   /* PCRE_DEBUG */  #endif   /* PCRE_DEBUG */
8841    
8842  /* Check for a pattern than can match an empty string, so that this information  /* Check for a pattern than can match an empty string, so that this information
8843  can be provided to applications. */  can be provided to applications. */
8844    
8845  do  do
# Line 8515  do Line 8848  do
8848      {      {
8849      re->flags |= PCRE_MATCH_EMPTY;      re->flags |= PCRE_MATCH_EMPTY;
8850      break;      break;
8851      }      }
8852    codestart += GET(codestart, 1);    codestart += GET(codestart, 1);
8853    }    }
8854  while (*codestart == OP_ALT);  while (*codestart == OP_ALT);

Legend:
Removed from v.1348  
changed lines
  Added in v.1365

  ViewVC Help
Powered by ViewVC 1.1.5