/[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 1072 by chpe, Tue Oct 16 15:54:40 2012 UTC revision 1379 by ph10, Mon Oct 14 13:54:07 2013 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2012 University of Cambridge             Copyright (c) 1997-2013 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 80  to check them every time. */ Line 80  to check them every time. */
80  /* Definitions to allow mutual recursion */  /* Definitions to allow mutual recursion */
81    
82  static int  static int
83    add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,    add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84      const pcre_uint32 *, unsigned int);      const pcre_uint32 *, unsigned int);
85    
86  static BOOL  static BOOL
87    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL,    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
88      int, int, int *, int *, branch_chain *, compile_data *, int *);      pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
89        compile_data *, int *);
90    
91    
92    
# Line 114  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 121  overrun before it actually does run off Line 129  overrun before it actually does run off
129    
130  /* Private flags added to firstchar and reqchar. */  /* Private flags added to firstchar and reqchar. */
131    
132  #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */  #define REQ_CASELESS    (1 << 0)        /* Indicates caselessness */
133  #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */  #define REQ_VARY        (1 << 1)        /* Reqchar followed non-literal item */
134  #define REQ_MASK       (REQ_CASELESS | REQ_VARY)  /* Negative values for the firstchar and reqchar flags */
135    #define REQ_UNSET       (-2)
136    #define REQ_NONE        (-1)
137    
138  /* Repeated character flags. */  /* Repeated character flags. */
139    
# Line 452  static const char error_texts[] = Line 462  static const char error_texts[] =
462    "POSIX collating elements are not supported\0"    "POSIX collating elements are not supported\0"
463    "this version of PCRE is compiled without UTF support\0"    "this version of PCRE is compiled without UTF support\0"
464    "spare error\0"  /** DEAD **/    "spare error\0"  /** DEAD **/
465    "character value in \\x{...} sequence is too large\0"    "character value in \\x{} or \\o{} is too large\0"
466    /* 35 */    /* 35 */
467    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
468    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
# Line 484  static const char error_texts[] = Line 494  static const char error_texts[] =
494    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
495    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
496    /* 60 */    /* 60 */
497    "(*VERB) not recognized\0"    "(*VERB) not recognized or malformed\0"
498    "number is too big\0"    "number is too big\0"
499    "subpattern name expected\0"    "subpattern name expected\0"
500    "digit expected after (?+\0"    "digit expected after (?+\0"
# Line 505  static const char error_texts[] = Line 515  static const char error_texts[] =
515    "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"    "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
516    "character value in \\u.... sequence is too large\0"    "character value in \\u.... sequence is too large\0"
517    "invalid UTF-32 string\0"    "invalid UTF-32 string\0"
518      "setting UTF is disabled by the application\0"
519      "non-hex character in \\x{} (closing brace missing?)\0"
520      /* 80 */
521      "non-octal character in \\o{} (closing brace missing?)\0"
522      "missing opening brace after \\o\0"
523    ;    ;
524    
525  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 644  static const pcre_uint8 ebcdic_chartab[] Line 659  static const pcre_uint8 ebcdic_chartab[]
659  #endif  #endif
660    
661    
662    /* This table is used to check whether auto-possessification is possible
663    between adjacent character-type opcodes. The left-hand (repeated) opcode is
664    used to select the row, and the right-hand opcode is use to select the column.
665    A value of 1 means that auto-possessification is OK. For example, the second
666    value in the first row means that \D+\d can be turned into \D++\d.
667    
668    The Unicode property types (\P and \p) have to be present to fill out the table
669    because of what their opcode values are, but the table values should always be
670    zero because property types are handled separately in the code. The last four
671    columns apply to items that cannot be repeated, so there is no need to have
672    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
673    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
674    
675    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
676    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
677    
678    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
679    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
680      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
681      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
682      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
683      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
684      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
685      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
686      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
687      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
688      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
689      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
690      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
691      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
692      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
693      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
694      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
695      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
696      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
697    };
698    
699    
700    /* This table is used to check whether auto-possessification is possible
701    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
702    left-hand (repeated) opcode is used to select the row, and the right-hand
703    opcode is used to select the column. The values are as follows:
704    
705      0   Always return FALSE (never auto-possessify)
706      1   Character groups are distinct (possessify if both are OP_PROP)
707      2   Check character categories in the same group (general or particular)
708      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
709    
710      4   Check left general category vs right particular category
711      5   Check right general category vs left particular category
712    
713      6   Left alphanum vs right general category
714      7   Left space vs right general category
715      8   Left word vs right general category
716    
717      9   Right alphanum vs left general category
718     10   Right space vs left general category
719     11   Right word vs left general category
720    
721     12   Left alphanum vs right particular category
722     13   Left space vs right particular category
723     14   Left word vs right particular category
724    
725     15   Right alphanum vs left particular category
726     16   Right space vs left particular category
727     17   Right word vs left particular category
728    */
729    
730    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
731    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
732      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
733      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
734      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
735      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
736      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
737      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
738      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
739      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
740      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
741      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
742      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
743    };
744    
745    /* This table is used to check whether auto-possessification is possible
746    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
747    specifies a general category and the other specifies a particular category. The
748    row is selected by the general category and the column by the particular
749    category. The value is 1 if the particular category is not part of the general
750    category. */
751    
752    static const pcre_uint8 catposstab[7][30] = {
753    /* 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 */
754      { 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 */
755      { 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 */
756      { 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 */
757      { 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 */
758      { 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 */
759      { 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 */
760      { 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 */
761    };
762    
763    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
764    a general or particular category. The properties in each row are those
765    that apply to the character set in question. Duplication means that a little
766    unnecessary work is done when checking, but this keeps things much simpler
767    because they can all use the same code. For more details see the comment where
768    this table is used.
769    
770    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
771    "space", but from Perl 5.18 it's included, so both categories are treated the
772    same here. */
773    
774    static const pcre_uint8 posspropstab[3][4] = {
775      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
776      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
777      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
778    };
779    
780    
781    
782  /*************************************************  /*************************************************
# Line 665  find_error_text(int n) Line 798  find_error_text(int n)
798  const char *s = error_texts;  const char *s = error_texts;
799  for (; n > 0; n--)  for (; n > 0; n--)
800    {    {
801    while (*s++ != 0) {};    while (*s++ != CHAR_NULL) {};
802    if (*s == 0) return "Error text not found (please report)";    if (*s == CHAR_NULL) return "Error text not found (please report)";
803    }    }
804  return s;  return s;
805  }  }
806    
807    
808    
809  /*************************************************  /*************************************************
810  *           Expand the workspace                 *  *           Expand the workspace                 *
811  *************************************************/  *************************************************/
# Line 749  return (*p == CHAR_RIGHT_CURLY_BRACKET); Line 883  return (*p == CHAR_RIGHT_CURLY_BRACKET);
883  *************************************************/  *************************************************/
884    
885  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
886  positive value for a simple escape such as \n, or 0 for a data character  positive value for a simple escape such as \n, or 0 for a data character which
887  which will be placed in chptr. A backreference to group n is returned as  will be placed in chptr. A backreference to group n is returned as negative n.
888  negative n. When UTF-8 is enabled, a positive value greater than 255 may  When UTF-8 is enabled, a positive value greater than 255 may be returned in
889  be returned in chptr.  chptr. On entry, ptr is pointing at the \. On exit, it is on the final
890  On entry,ptr is pointing at the \. On exit, it is on the final character of the  character of the escape sequence.
 escape sequence.  
891    
892  Arguments:  Arguments:
893    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
894    chptr          points to the data character    chptr          points to a returned data character
895    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
896    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
897    options        the options bits    options        the options bits
# Line 771  Returns:         zero => a data characte Line 904  Returns:         zero => a data characte
904  */  */
905    
906  static int  static int
907  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
908    int bracount, int options, BOOL isclass)    int bracount, int options, BOOL isclass)
909  {  {
910  /* PCRE_UTF16 has the same value as PCRE_UTF8. */  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
# Line 786  ptr--;                            /* Set Line 919  ptr--;                            /* Set
919    
920  /* If backslash is at the end of the pattern, it's an error. */  /* If backslash is at the end of the pattern, it's an error. */
921    
922  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
923    
924  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
925  in a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
# Line 795  Otherwise further processing may be requ Line 928  Otherwise further processing may be requ
928  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
929  /* Not alphanumeric */  /* Not alphanumeric */
930  else if (c < CHAR_0 || c > CHAR_z) {}  else if (c < CHAR_0 || c > CHAR_z) {}
931  else if ((i = escapes[c - CHAR_0]) != 0) { if (i > 0) c = (pcre_uint32)i; else escape = -i; }  else if ((i = escapes[c - CHAR_0]) != 0)
932      { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
933    
934  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
935  /* Not alphanumeric */  /* Not alphanumeric */
# Line 845  else Line 979  else
979            }            }
980    
981  #if defined COMPILE_PCRE8  #if defined COMPILE_PCRE8
982          if (c > (utf ? 0x10ffff : 0xff))          if (c > (utf ? 0x10ffffU : 0xffU))
983  #elif defined COMPILE_PCRE16  #elif defined COMPILE_PCRE16
984          if (c > (utf ? 0x10ffff : 0xffff))          if (c > (utf ? 0x10ffffU : 0xffffU))
985  #elif defined COMPILE_PCRE32  #elif defined COMPILE_PCRE32
986          if (utf && c > 0x10ffff)          if (utf && c > 0x10ffffU)
987  #endif  #endif
988            {            {
989            *errorcodeptr = ERR76;            *errorcodeptr = ERR76;
# Line 896  else Line 1030  else
1030      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1031        {        {
1032        const pcre_uchar *p;        const pcre_uchar *p;
1033        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)        for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
1034          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1035        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)        if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1036          {          {
1037          escape = ESC_k;          escape = ESC_k;
1038          break;          break;
# Line 961  else Line 1095  else
1095      break;      break;
1096    
1097      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1098      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. Perl has changed
1099      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1100        recommended to avoid the ambiguities in the old syntax.
1101    
1102      Outside a character class, the digits are read as a decimal number. If the      Outside a character class, the digits are read as a decimal number. If the
1103      number is less than 10, or if there are that many previous extracting      number is less than 8 (used to be 10), or if there are that many previous
1104      left brackets, then it is a back reference. Otherwise, up to three octal      extracting left brackets, then it is a back reference. Otherwise, up to
1105      digits are read to form an escaped byte. Thus \123 is likely to be octal      three octal digits are read to form an escaped byte. Thus \123 is likely to
1106      123 (cf \0123, which is octal 012 followed by the literal 3). If the octal      be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1107      value is greater than 377, the least significant 8 bits are taken. Inside a      the octal value is greater than 377, the least significant 8 bits are
1108      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1109    
1110        Inside a character class, \ followed by a digit is always either a literal
1111        8 or 9 or an octal number. */
1112    
1113      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1114      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
# Line 997  else Line 1135  else
1135          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1136          break;          break;
1137          }          }
1138        if (s < 10 || s <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1139          {          {
1140          escape = -s;          escape = -s;
1141          break;          break;
# Line 1005  else Line 1143  else
1143        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1144        }        }
1145    
1146      /* Handle an octal number following \. If the first digit is 8 or 9, Perl      /* Handle a digit following \ when the number is not a back reference. If
1147      generates a binary zero byte and treats the digit as a following literal.      the first digit is 8 or 9, Perl used to generate a binary zero byte and
1148      Thus we have to pull back the pointer by one. */      then treat the digit as a following literal. At least by Perl 5.18 this
1149        changed so as not to insert the binary zero. */
1150    
1151      if ((c = *ptr) >= CHAR_8)      if ((c = *ptr) >= CHAR_8) break;
1152        {  
1153        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1154    
1155      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
1156      larger first octal digit. The original code used just to take the least      larger first octal digit. The original code used just to take the least
# Line 1031  else Line 1167  else
1167  #endif  #endif
1168      break;      break;
1169    
1170      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \o is a relatively new Perl feature, supporting a more general way of
1171      than 0xff in utf or non-8bit mode, but only if the ddd are hex digits.      specifying character codes in octal. The only supported form is \o{ddd}. */
1172      If not, { is treated as a data character. */  
1173        case CHAR_o:
1174        if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1175          {
1176          ptr += 2;
1177          c = 0;
1178          overflow = FALSE;
1179          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1180            {
1181            register pcre_uint32 cc = *ptr++;
1182            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1183    #ifdef COMPILE_PCRE32
1184            if (c >= 0x20000000l) { overflow = TRUE; break; }
1185    #endif
1186            c = (c << 3) + cc - CHAR_0 ;
1187    #if defined COMPILE_PCRE8
1188            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1189    #elif defined COMPILE_PCRE16
1190            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1191    #elif defined COMPILE_PCRE32
1192            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1193    #endif
1194            }
1195          if (overflow)
1196            {
1197            while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1198            *errorcodeptr = ERR34;
1199            }
1200          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1201            {
1202            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1203            }
1204          else *errorcodeptr = ERR80;
1205          }
1206        break;
1207    
1208        /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1209        numbers. Otherwise it is a lowercase x letter. */
1210    
1211      case CHAR_x:      case CHAR_x:
1212      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1213        {        {
       /* In JavaScript, \x must be followed by two hexadecimal numbers.  
       Otherwise it is a lowercase x letter. */  
1214        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1215          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1216          {          {
# Line 1056  else Line 1227  else
1227  #endif  #endif
1228            }            }
1229          }          }
1230        break;        }    /* End JavaScript handling */
       }  
1231    
1232      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1233        {      greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1234        const pcre_uchar *pt = ptr + 2;      digits. If not, { used to be treated as a data character. However, Perl
1235        seems to read hex digits up to the first non-such, and ignore the rest, so
1236        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1237        now gives an error. */
1238    
1239        c = 0;      else
1240        overflow = FALSE;        {
1241        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1242          {          {
1243          register pcre_uint32 cc = *pt++;          ptr += 2;
1244          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          c = 0;
1245            overflow = FALSE;
1246            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1247              {
1248              register pcre_uint32 cc = *ptr++;
1249              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1250    
1251  #ifdef COMPILE_PCRE32  #ifdef COMPILE_PCRE32
1252          if (c >= 0x10000000l) { overflow = TRUE; break; }            if (c >= 0x10000000l) { overflow = TRUE; break; }
1253  #endif  #endif
1254    
1255  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1256          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1257          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1258  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1259          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */            if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1260          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1261  #endif  #endif
1262    
1263  #if defined COMPILE_PCRE8  #if defined COMPILE_PCRE8
1264          if (c > (utf ? 0x10ffff : 0xff)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1265  #elif defined COMPILE_PCRE16  #elif defined COMPILE_PCRE16
1266          if (c > (utf ? 0x10ffff : 0xffff)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1267  #elif defined COMPILE_PCRE32  #elif defined COMPILE_PCRE32
1268          if (utf && c > 0x10ffff) { overflow = TRUE; break; }            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1269  #endif  #endif
1270          }            }
1271    
1272        if (overflow)          if (overflow)
1273          {            {
1274          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1275          *errorcodeptr = ERR34;            *errorcodeptr = ERR34;
1276          }            }
1277    
1278        if (*pt == CHAR_RIGHT_CURLY_BRACKET)          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1279          {            {
1280          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1281          ptr = pt;            }
         break;  
         }  
1282    
1283        /* If the sequence of hex digits does not end with '}', then we don't          /* If the sequence of hex digits does not end with '}', give an error.
1284        recognize this construct; fall through to the normal \x handling. */          We used just to recognize this construct and fall through to the normal
1285        }          \x handling, but nowadays Perl gives an error, which seems much more
1286            sensible, so we do too. */
1287    
1288      /* Read just a single-byte hex-defined char */          else *errorcodeptr = ERR79;
1289            }   /* End of \x{} processing */
1290    
1291      c = 0;        /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1292      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)  
1293        {        else
1294        pcre_uint32 cc;                          /* Some compilers don't like */          {
1295        cc = *(++ptr);                           /* ++ in initializers */          c = 0;
1296            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1297              {
1298              pcre_uint32 cc;                          /* Some compilers don't like */
1299              cc = *(++ptr);                           /* ++ in initializers */
1300  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1301        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1302        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1303  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1304        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */            if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1305        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1306  #endif  #endif
1307        }            }
1308            }     /* End of \xdd handling */
1309          }       /* End of Perl-style \x handling */
1310      break;      break;
1311    
1312      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
# Line 1132  else Line 1316  else
1316    
1317      case CHAR_c:      case CHAR_c:
1318      c = *(++ptr);      c = *(++ptr);
1319      if (c == 0)      if (c == CHAR_NULL)
1320        {        {
1321        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1322        break;        break;
# Line 1188  if ((options & PCRE_UCP) != 0 && escape Line 1372  if ((options & PCRE_UCP) != 0 && escape
1372  return escape;  return escape;
1373  }  }
1374    
1375    
1376    
1377  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
1378  /*************************************************  /*************************************************
1379  *               Handle \P and \p                 *  *               Handle \P and \p                 *
# Line 1201  escape sequence. Line 1387  escape sequence.
1387  Argument:  Argument:
1388    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1389    negptr         points to a boolean that is set TRUE for negation else FALSE    negptr         points to a boolean that is set TRUE for negation else FALSE
1390    dptr           points to an int that is set to the detailed property value    ptypeptr       points to an unsigned int that is set to the type value
1391      pdataptr       points to an unsigned int that is set to the detailed property value
1392    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1393    
1394  Returns:         type value from ucp_type_table, or -1 for an invalid type  Returns:         TRUE if the type value was found, or FALSE for an invalid type
1395  */  */
1396    
1397  static int  static BOOL
1398  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1399      unsigned int *pdataptr, int *errorcodeptr)
1400  {  {
1401  pcre_uchar c;  pcre_uchar c;
1402  int i, bot, top;  int i, bot, top;
# Line 1216  const pcre_uchar *ptr = *ptrptr; Line 1404  const pcre_uchar *ptr = *ptrptr;
1404  pcre_uchar name[32];  pcre_uchar name[32];
1405    
1406  c = *(++ptr);  c = *(++ptr);
1407  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1408    
1409  *negptr = FALSE;  *negptr = FALSE;
1410    
# Line 1233  if (c == CHAR_LEFT_CURLY_BRACKET) Line 1421  if (c == CHAR_LEFT_CURLY_BRACKET)
1421    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1422      {      {
1423      c = *(++ptr);      c = *(++ptr);
1424      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1425      if (c == CHAR_RIGHT_CURLY_BRACKET) break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1426      name[i] = c;      name[i] = c;
1427      }      }
# Line 1263  while (bot < top) Line 1451  while (bot < top)
1451    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1452    if (r == 0)    if (r == 0)
1453      {      {
1454      *dptr = PRIV(utt)[i].value;      *ptypeptr = PRIV(utt)[i].type;
1455      return PRIV(utt)[i].type;      *pdataptr = PRIV(utt)[i].value;
1456        return TRUE;
1457      }      }
1458    if (r > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1459    }    }
1460    
1461  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1462  *ptrptr = ptr;  *ptrptr = ptr;
1463  return -1;  return FALSE;
1464    
1465  ERROR_RETURN:  ERROR_RETURN:
1466  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1467  *ptrptr = ptr;  *ptrptr = ptr;
1468  return -1;  return FALSE;
1469  }  }
1470  #endif  #endif
1471    
1472    
1473    
   
1474  /*************************************************  /*************************************************
1475  *         Read repeat counts                     *  *         Read repeat counts                     *
1476  *************************************************/  *************************************************/
# Line 1351  return p; Line 1539  return p;
1539    
1540    
1541  /*************************************************  /*************************************************
 *  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;  
   
   /* 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 != 0; 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 != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;  
       if (*ptr != 0) 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) == 0) goto FAIL_EXIT;  
     if (*ptr == CHAR_Q) for (;;)  
       {  
       while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};  
       if (*ptr == 0) 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 == 0) return -1;  
       if (*ptr == CHAR_BACKSLASH)  
         {  
         if (*(++ptr) == 0) goto FAIL_EXIT;  
         if (*ptr == CHAR_Q) for (;;)  
           {  
           while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};  
           if (*ptr == 0) goto FAIL_EXIT;  
           if (*(++ptr) == CHAR_E) break;  
           }  
         continue;  
         }  
       }  
     continue;  
     }  
   
   /* Skip comments in /x mode */  
   
   if (xmode && *ptr == CHAR_NUMBER_SIGN)  
     {  
     ptr++;  
     while (*ptr != 0)  
       {  
       if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }  
       ptr++;  
 #ifdef SUPPORT_UTF  
       if (utf) FORWARDCHAR(ptr);  
 #endif  
       }  
     if (*ptr == 0) 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 == 0) 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++ == 0) break;  
   }  
   
 return rc;  
 }  
   
   
   
   
 /*************************************************  
1542  *      Find first significant op code            *  *      Find first significant op code            *
1543  *************************************************/  *************************************************/
1544    
# Line 1684  for (;;) Line 1577  for (;;)
1577    
1578      case OP_CALLOUT:      case OP_CALLOUT:
1579      case OP_CREF:      case OP_CREF:
1580      case OP_NCREF:      case OP_DNCREF:
1581      case OP_RREF:      case OP_RREF:
1582      case OP_NRREF:      case OP_DNRREF:
1583      case OP_DEF:      case OP_DEF:
1584      code += PRIV(OP_lengths)[*code];      code += PRIV(OP_lengths)[*code];
1585      break;      break;
# Line 1700  for (;;) Line 1593  for (;;)
1593    
1594    
1595    
   
1596  /*************************************************  /*************************************************
1597  *        Find the fixed length of a branch       *  *        Find the fixed length of a branch       *
1598  *************************************************/  *************************************************/
# Line 1824  for (;;) Line 1716  for (;;)
1716      case OP_COMMIT:      case OP_COMMIT:
1717      case OP_CREF:      case OP_CREF:
1718      case OP_DEF:      case OP_DEF:
1719        case OP_DNCREF:
1720        case OP_DNRREF:
1721      case OP_DOLL:      case OP_DOLL:
1722      case OP_DOLLM:      case OP_DOLLM:
1723      case OP_EOD:      case OP_EOD:
1724      case OP_EODN:      case OP_EODN:
1725      case OP_FAIL:      case OP_FAIL:
     case OP_NCREF:  
     case OP_NRREF:  
1726      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1727      case OP_PRUNE:      case OP_PRUNE:
1728      case OP_REVERSE:      case OP_REVERSE:
# Line 1852  for (;;) Line 1744  for (;;)
1744      case OP_NOTI:      case OP_NOTI:
1745      branchlength++;      branchlength++;
1746      cc += 2;      cc += 2;
1747  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #ifdef SUPPORT_UTF
1748      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1749  #endif  #endif
1750      break;      break;
# Line 1866  for (;;) Line 1758  for (;;)
1758      case OP_NOTEXACTI:      case OP_NOTEXACTI:
1759      branchlength += (int)GET2(cc,1);      branchlength += (int)GET2(cc,1);
1760      cc += 2 + IMM2_SIZE;      cc += 2 + IMM2_SIZE;
1761  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #ifdef SUPPORT_UTF
1762      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1763  #endif  #endif
1764      break;      break;
1765    
1766      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1767      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1768      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1769        cc += 2;        cc += 2;
1770      cc += 1 + IMM2_SIZE + 1;      cc += 1 + IMM2_SIZE + 1;
1771      break;      break;
# Line 1909  for (;;) Line 1801  for (;;)
1801    
1802      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1803    
 #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32  
     case OP_XCLASS:  
     cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];  
     /* Fall through */  
 #endif  
   
1804      case OP_CLASS:      case OP_CLASS:
1805      case OP_NCLASS:      case OP_NCLASS:
1806    #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1807        case OP_XCLASS:
1808        /* The original code caused an unsigned overflow in 64 bit systems,
1809        so now we use a conditional statement. */
1810        if (op == OP_XCLASS)
1811          cc += GET(cc, 1);
1812        else
1813          cc += PRIV(OP_lengths)[OP_CLASS];
1814    #else
1815      cc += PRIV(OP_lengths)[OP_CLASS];      cc += PRIV(OP_lengths)[OP_CLASS];
1816    #endif
1817    
1818      switch (*cc)      switch (*cc)
1819        {        {
       case OP_CRPLUS:  
       case OP_CRMINPLUS:  
1820        case OP_CRSTAR:        case OP_CRSTAR:
1821        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1822          case OP_CRPLUS:
1823          case OP_CRMINPLUS:
1824        case OP_CRQUERY:        case OP_CRQUERY:
1825        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1826          case OP_CRPOSSTAR:
1827          case OP_CRPOSPLUS:
1828          case OP_CRPOSQUERY:
1829        return -1;        return -1;
1830    
1831        case OP_CRRANGE:        case OP_CRRANGE:
1832        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1833          case OP_CRPOSRANGE:
1834        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1835        branchlength += (int)GET2(cc,1);        branchlength += (int)GET2(cc,1);
1836        cc += 1 + 2 * IMM2_SIZE;        cc += 1 + 2 * IMM2_SIZE;
# Line 1999  for (;;) Line 1899  for (;;)
1899      case OP_QUERYI:      case OP_QUERYI:
1900      case OP_REF:      case OP_REF:
1901      case OP_REFI:      case OP_REFI:
1902        case OP_DNREF:
1903        case OP_DNREFI:
1904      case OP_SBRA:      case OP_SBRA:
1905      case OP_SBRAPOS:      case OP_SBRAPOS:
1906      case OP_SCBRA:      case OP_SCBRA:
# Line 2035  for (;;) Line 1937  for (;;)
1937    
1938    
1939    
   
1940  /*************************************************  /*************************************************
1941  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
1942  *************************************************/  *************************************************/
# Line 2112  for (;;) Line 2013  for (;;)
2013        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2014        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2015        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2016        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2017          code += 2;          code += 2;
2018        break;        break;
2019    
2020        case OP_MARK:        case OP_MARK:
2021        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2022        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2023        case OP_THEN_ARG:        case OP_THEN_ARG:
2024        code += code[1];        code += code[1];
2025        break;        break;
# Line 2232  for (;;) Line 2130  for (;;)
2130        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2131        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2132        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2133        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2134          code += 2;          code += 2;
2135        break;        break;
2136    
2137        case OP_MARK:        case OP_MARK:
2138        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2139        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2140        case OP_THEN_ARG:        case OP_THEN_ARG:
2141        code += code[1];        code += code[1];
2142        break;        break;
# Line 2343  Arguments: Line 2238  Arguments:
2238    endcode     points to where to stop    endcode     points to where to stop
2239    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2240    cd          contains pointers to tables etc.    cd          contains pointers to tables etc.
2241      recurses    chain of recurse_check to catch mutual recursion
2242    
2243  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2244  */  */
2245    
2246    typedef struct recurse_check {
2247      struct recurse_check *prev;
2248      const pcre_uchar *group;
2249    } recurse_check;
2250    
2251  static BOOL  static BOOL
2252  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2253    BOOL utf, compile_data *cd)    BOOL utf, compile_data *cd, recurse_check *recurses)
2254  {  {
2255  register pcre_uchar c;  register pcre_uchar c;
2256    recurse_check this_recurse;
2257    
2258  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2259       code < endcode;       code < endcode;
2260       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
# Line 2379  for (code = first_significant_code(code Line 2282  for (code = first_significant_code(code
2282    
2283    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2284      {      {
2285      const pcre_uchar *scode;      const pcre_uchar *scode = cd->start_code + GET(code, 1);
2286      BOOL empty_branch;      BOOL empty_branch;
2287    
2288      /* Test for forward reference */      /* Test for forward reference or uncompleted reference. This is disabled
2289        when called to scan a completed pattern by setting cd->start_workspace to
2290        NULL. */
2291    
2292        if (cd->start_workspace != NULL)
2293          {
2294          const pcre_uchar *tcode;
2295          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2296            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2297          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2298          }
2299    
2300        /* If we are scanning a completed pattern, there are no forward references
2301        and all groups are complete. We need to detect whether this is a recursive
2302        call, as otherwise there will be an infinite loop. If it is a recursion,
2303        just skip over it. Simple recursions are easily detected. For mutual
2304        recursions we keep a chain on the stack. */
2305    
2306      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)      else
2307        if ((int)GET(scode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;        {
2308          recurse_check *r = recurses;
2309          const pcre_uchar *endgroup = scode;
2310    
2311      /* Not a forward reference, test for completed backward reference */        do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2312          if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2313    
2314      empty_branch = FALSE;        for (r = recurses; r != NULL; r = r->prev)
2315      scode = cd->start_code + GET(code, 1);          if (r->group == scode) break;
2316      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */        if (r != NULL) continue;   /* Mutual recursion */
2317          }
2318    
2319      /* Completed backwards reference */      /* Completed reference; scan the referenced group, remembering it on the
2320        stack chain to detect mutual recursions. */
2321    
2322      do      empty_branch = FALSE;
2323        this_recurse.prev = recurses;
2324        this_recurse.group = scode;
2325    
2326        do
2327        {        {
2328        if (could_be_empty_branch(scode, endcode, utf, cd))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2329          {          {
2330          empty_branch = TRUE;          empty_branch = TRUE;
2331          break;          break;
# Line 2453  for (code = first_significant_code(code Line 2381  for (code = first_significant_code(code
2381        empty_branch = FALSE;        empty_branch = FALSE;
2382        do        do
2383          {          {
2384          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2385            empty_branch = TRUE;            empty_branch = TRUE;
2386          code += GET(code, 1);          code += GET(code, 1);
2387          }          }
# Line 2495  for (code = first_significant_code(code Line 2423  for (code = first_significant_code(code
2423        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2424        case OP_CRQUERY:        case OP_CRQUERY:
2425        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2426          case OP_CRPOSSTAR:
2427          case OP_CRPOSQUERY:
2428        break;        break;
2429    
2430        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2431        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2432        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2433          case OP_CRPOSPLUS:
2434        return FALSE;        return FALSE;
2435    
2436        case OP_CRRANGE:        case OP_CRRANGE:
2437        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2438          case OP_CRPOSRANGE:
2439        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2440        break;        break;
2441        }        }
# Line 2511  for (code = first_significant_code(code Line 2443  for (code = first_significant_code(code
2443    
2444      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2445    
2446        case OP_ANY:
2447        case OP_ALLANY:
2448        case OP_ANYBYTE:
2449    
2450      case OP_PROP:      case OP_PROP:
2451      case OP_NOTPROP:      case OP_NOTPROP:
2452        case OP_ANYNL:
2453    
2454        case OP_NOT_HSPACE:
2455        case OP_HSPACE:
2456        case OP_NOT_VSPACE:
2457        case OP_VSPACE:
2458      case OP_EXTUNI:      case OP_EXTUNI:
2459    
2460      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2461      case OP_DIGIT:      case OP_DIGIT:
2462      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2463      case OP_WHITESPACE:      case OP_WHITESPACE:
2464      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2465      case OP_WORDCHAR:      case OP_WORDCHAR:
2466      case OP_ANY:  
     case OP_ALLANY:  
     case OP_ANYBYTE:  
2467      case OP_CHAR:      case OP_CHAR:
2468      case OP_CHARI:      case OP_CHARI:
2469      case OP_NOT:      case OP_NOT:
2470      case OP_NOTI:      case OP_NOTI:
2471    
2472      case OP_PLUS:      case OP_PLUS:
2473        case OP_PLUSI:
2474      case OP_MINPLUS:      case OP_MINPLUS:
2475      case OP_POSPLUS:      case OP_MINPLUSI:
2476      case OP_EXACT:  
2477      case OP_NOTPLUS:      case OP_NOTPLUS:
2478        case OP_NOTPLUSI:
2479      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2480        case OP_NOTMINPLUSI:
2481    
2482        case OP_POSPLUS:
2483        case OP_POSPLUSI:
2484      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2485        case OP_NOTPOSPLUSI:
2486    
2487        case OP_EXACT:
2488        case OP_EXACTI:
2489      case OP_NOTEXACT:      case OP_NOTEXACT:
2490        case OP_NOTEXACTI:
2491    
2492      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2493      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2494      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2495      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2496    
2497      return FALSE;      return FALSE;
2498    
2499      /* 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 2558  for (code = first_significant_code(code Line 2513  for (code = first_significant_code(code
2513      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2514      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2515      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2516      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2517        code += 2;        code += 2;
2518      break;      break;
2519    
# Line 2572  for (code = first_significant_code(code Line 2527  for (code = first_significant_code(code
2527      return TRUE;      return TRUE;
2528    
2529      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2530      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2531        followed by a multibyte character. */
2532    
2533  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2534      case OP_STAR:      case OP_STAR:
2535      case OP_STARI:      case OP_STARI:
2536        case OP_NOTSTAR:
2537        case OP_NOTSTARI:
2538    
2539      case OP_MINSTAR:      case OP_MINSTAR:
2540      case OP_MINSTARI:      case OP_MINSTARI:
2541        case OP_NOTMINSTAR:
2542        case OP_NOTMINSTARI:
2543    
2544      case OP_POSSTAR:      case OP_POSSTAR:
2545      case OP_POSSTARI:      case OP_POSSTARI:
2546        case OP_NOTPOSSTAR:
2547        case OP_NOTPOSSTARI:
2548    
2549      case OP_QUERY:      case OP_QUERY:
2550      case OP_QUERYI:      case OP_QUERYI:
2551        case OP_NOTQUERY:
2552        case OP_NOTQUERYI:
2553    
2554      case OP_MINQUERY:      case OP_MINQUERY:
2555      case OP_MINQUERYI:      case OP_MINQUERYI:
2556        case OP_NOTMINQUERY:
2557        case OP_NOTMINQUERYI:
2558    
2559      case OP_POSQUERY:      case OP_POSQUERY:
2560      case OP_POSQUERYI:      case OP_POSQUERYI:
2561        case OP_NOTPOSQUERY:
2562        case OP_NOTPOSQUERYI:
2563    
2564      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2565      break;      break;
2566    
2567      case OP_UPTO:      case OP_UPTO:
2568      case OP_UPTOI:      case OP_UPTOI:
2569        case OP_NOTUPTO:
2570        case OP_NOTUPTOI:
2571    
2572      case OP_MINUPTO:      case OP_MINUPTO:
2573      case OP_MINUPTOI:      case OP_MINUPTOI:
2574        case OP_NOTMINUPTO:
2575        case OP_NOTMINUPTOI:
2576    
2577      case OP_POSUPTO:      case OP_POSUPTO:
2578      case OP_POSUPTOI:      case OP_POSUPTOI:
2579        case OP_NOTPOSUPTO:
2580        case OP_NOTPOSUPTOI:
2581    
2582      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]);
2583      break;      break;
2584  #endif  #endif
# Line 2606  for (code = first_significant_code(code Line 2589  for (code = first_significant_code(code
2589      case OP_MARK:      case OP_MARK:
2590      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
2591      case OP_SKIP_ARG:      case OP_SKIP_ARG:
     code += code[1];  
     break;  
   
2592      case OP_THEN_ARG:      case OP_THEN_ARG:
2593      code += code[1];      code += code[1];
2594      break;      break;
# Line 2652  could_be_empty(const pcre_uchar *code, c Line 2632  could_be_empty(const pcre_uchar *code, c
2632  {  {
2633  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2634    {    {
2635    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2636      return FALSE;      return FALSE;
2637    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2638    }    }
# Line 2662  return TRUE; Line 2642  return TRUE;
2642    
2643    
2644  /*************************************************  /*************************************************
2645  *           Check for POSIX class syntax         *  *        Base opcode of repeated opcodes         *
2646  *************************************************/  *************************************************/
2647    
2648  /* This function is called when the sequence "[:" or "[." or "[=" is  /* Returns the base opcode for repeated single character type opcodes. If the
2649  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.  
   
 Originally, this function only recognized a sequence of letters between the  
 terminators, but it seems that Perl recognizes any sequence of characters,  
 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:].  
   
 The problem in trying to be exactly like Perl is in the handling of escapes. We  
 have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX  
 class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code  
 below handles the special case of \], but does not try to do any other escape  
 processing. This makes it different from Perl for cases such as [:l\ower:]  
 where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize  
 "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  
 I think.  
   
 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.  
   
 In Perl, unescaped square brackets may also appear as part of class names. For  
 example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for  
 [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not  
 seem right at all. PCRE does not allow closing square brackets in POSIX class  
 names.  
   
 Arguments:  
   ptr      pointer to the initial [  
   endptr   where to return the end pointer  
2650    
2651  Returns:   TRUE or FALSE  Arguments:  c opcode
2652    Returns:    base opcode for the type
2653  */  */
2654    
2655  static BOOL  static pcre_uchar
2656  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)  get_repeat_base(pcre_uchar c)
2657  {  {
2658  pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */  return (c > OP_TYPEPOSUPTO)? c :
2659  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */         (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2660  for (++ptr; *ptr != 0; ptr++)         (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2661    {         (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2662    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)         (c >= OP_STARI)?      OP_STARI :
2663      ptr++;                               OP_STAR;
   else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
   else  
     {  
     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;  
2664  }  }
2665    
2666    
2667    
2668    #ifdef SUPPORT_UCP
2669  /*************************************************  /*************************************************
2670  *          Check POSIX class name                *  *        Check a character and a property        *
2671  *************************************************/  *************************************************/
2672    
2673  /* This function is called to check the name given in a POSIX-style class entry  /* This function is called by check_auto_possessive() when a property item
2674  such as [:alnum:].  is adjacent to a fixed character.
2675    
2676  Arguments:  Arguments:
2677    ptr        points to the first letter    c            the character
2678    len        the length of the name    ptype        the property type
2679      pdata        the data for the type
2680      negated      TRUE if it's a negated property (\P or \p{^)
2681    
2682  Returns:     a value representing the name, or -1 if unknown  Returns:       TRUE if auto-possessifying is OK
2683  */  */
2684    
2685  static int  static BOOL
2686  check_posix_name(const pcre_uchar *ptr, int len)  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2687      BOOL negated)
2688  {  {
2689  const char *pn = posix_names;  const pcre_uint32 *p;
2690  register int yield = 0;  const ucd_record *prop = GET_UCD(c);
 while (posix_name_lengths[yield] != 0)  
   {  
   if (len == posix_name_lengths[yield] &&  
     STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;  
   pn += posix_name_lengths[yield] + 1;  
   yield++;  
   }  
 return -1;  
 }  
   
   
 /*************************************************  
 *    Adjust OP_RECURSE items in repeated group   *  
 *************************************************/  
2691    
2692  /* OP_RECURSE items contain an offset from the start of the regex to the group  switch(ptype)
2693  that is referenced. This means that groups can be replicated for fixed    {
2694  repetition simply by copying (because the recursion is allowed to refer to    case PT_LAMP:
2695  earlier groups that are outside the current group). However, when a group is    return (prop->chartype == ucp_Lu ||
2696  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is            prop->chartype == ucp_Ll ||
2697  inserted before it, after it has been compiled. This means that any OP_RECURSE            prop->chartype == ucp_Lt) == negated;
 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.  
2698    
2699  This function has been extended with the possibility of forward references for    case PT_GC:
2700  recursions and subroutine calls. It must also check the list of such references    return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
 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).  
2701    
2702  Arguments:    case PT_PC:
2703    group      points to the start of the group    return (pdata == prop->chartype) == negated;
   adjust     the amount by which the group is to be moved  
   utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode  
   cd         contains pointers to tables etc.  
   save_hwm   the hwm forward reference pointer at the start of the group  
2704    
2705  Returns:     nothing    case PT_SC:
2706  */    return (pdata == prop->script) == negated;
2707    
2708  static void    /* These are specials */
 adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,  
   pcre_uchar *save_hwm)  
 {  
 pcre_uchar *ptr = group;  
2709    
2710  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)    case PT_ALNUM:
2711    {    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2712    int offset;            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
   pcre_uchar *hc;  
2713    
2714    /* See if this recursion is on the forward reference list. If so, adjust the    /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2715    reference. */    means that Perl space and POSIX space are now identical. PCRE was changed
2716      at release 8.34. */
2717    
2718    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    case PT_SPACE:    /* Perl space */
2719      case PT_PXSPACE:  /* POSIX space */
2720      switch(c)
2721      {      {
2722      offset = GET(hc, 0);      HSPACE_CASES:
2723      if (cd->start_code + offset == ptr + 1)      VSPACE_CASES:
2724        {      return negated;
2725        PUT(hc, 0, offset + adjust);  
2726        break;      default:
2727        }      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2728      }      }
2729      break;  /* Control never reaches here */
2730    
2731    /* Otherwise, adjust the recursion offset if it's after the start of this    case PT_WORD:
2732    group. */    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2733              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2734              c == CHAR_UNDERSCORE) == negated;
2735    
2736    if (hc >= cd->hwm)    case PT_CLIST:
2737      p = PRIV(ucd_caseless_sets) + prop->caseset;
2738      for (;;)
2739      {      {
2740      offset = GET(ptr, 1);      if (c < *p) return !negated;
2741      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      if (c == *p++) return negated;
2742      }      }
2743      break;  /* Control never reaches here */
   ptr += 1 + LINK_SIZE;  
2744    }    }
2745    
2746    return FALSE;
2747  }  }
2748    #endif  /* SUPPORT_UCP */
2749    
2750    
2751    
2752  /*************************************************  /*************************************************
2753  *        Insert an automatic callout point       *  *        Fill the character property list        *
2754  *************************************************/  *************************************************/
2755    
2756  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert  /* Checks whether the code points to an opcode that can take part in auto-
2757  callout points before each pattern item.  possessification, and if so, fills a list with its properties.
2758    
2759  Arguments:  Arguments:
2760    code           current code pointer    code        points to start of expression
2761    ptr            current pattern pointer    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2762    cd             pointers to tables etc    fcc         points to case-flipping table
2763      list        points to output list
2764                  list[0] will be filled with the opcode
2765                  list[1] will be non-zero if this opcode
2766                    can match an empty character string
2767                  list[2..7] depends on the opcode
2768    
2769  Returns:         new code pointer  Returns:      points to the start of the next opcode if *code is accepted
2770                  NULL if *code is not accepted
2771  */  */
2772    
2773  static pcre_uchar *  static const pcre_uchar *
2774  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)  get_chr_property_list(const pcre_uchar *code, BOOL utf,
2775      const pcre_uint8 *fcc, pcre_uint32 *list)
2776  {  {
2777  *code++ = OP_CALLOUT;  pcre_uchar c = *code;
2778  *code++ = 255;  const pcre_uchar *end;
2779  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */  const pcre_uint32 *clist_src;
2780  PUT(code, LINK_SIZE, 0);                       /* Default length */  pcre_uint32 *clist_dest;
2781  return code + 2 * LINK_SIZE;  pcre_uint32 chr;
2782  }  pcre_uchar base;
2783    
2784    list[0] = c;
2785    list[1] = FALSE;
2786    code++;
2787    
2788    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2789      {
2790      base = get_repeat_base(c);
2791      c -= (base - OP_STAR);
2792    
2793  /*************************************************    if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2794  *         Complete a callout item                *      code += IMM2_SIZE;
 *************************************************/  
2795    
2796  /* A callout item contains the length of the next item in the pattern, which    list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
 we can't fill in till after we have reached the relevant point. This is used  
 for both automatic and manual callouts.  
2797    
2798  Arguments:    switch(base)
2799    previous_callout   points to previous callout item      {
2800    ptr                current pattern pointer      case OP_STAR:
2801    cd                 pointers to tables etc      list[0] = OP_CHAR;
2802        break;
2803    
2804  Returns:             nothing      case OP_STARI:
2805  */      list[0] = OP_CHARI;
2806        break;
2807    
2808  static void      case OP_NOTSTAR:
2809  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)      list[0] = OP_NOT;
2810  {      break;
 int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));  
 PUT(previous_callout, 2 + LINK_SIZE, length);  
 }  
2811    
2812        case OP_NOTSTARI:
2813        list[0] = OP_NOTI;
2814        break;
2815    
2816        case OP_TYPESTAR:
2817        list[0] = *code;
2818        code++;
2819        break;
2820        }
2821      c = list[0];
2822      }
2823    
2824  #ifdef SUPPORT_UCP  switch(c)
2825  /*************************************************    {
2826  *           Get othercase range                  *    case OP_NOT_DIGIT:
2827  *************************************************/    case OP_DIGIT:
2828      case OP_NOT_WHITESPACE:
2829      case OP_WHITESPACE:
2830      case OP_NOT_WORDCHAR:
2831      case OP_WORDCHAR:
2832      case OP_ANY:
2833      case OP_ALLANY:
2834      case OP_ANYNL:
2835      case OP_NOT_HSPACE:
2836      case OP_HSPACE:
2837      case OP_NOT_VSPACE:
2838      case OP_VSPACE:
2839      case OP_EXTUNI:
2840      case OP_EODN:
2841      case OP_EOD:
2842      case OP_DOLL:
2843      case OP_DOLLM:
2844      return code;
2845    
2846  /* This function is passed the start and end of a class range, in UTF-8 mode    case OP_CHAR:
2847  with UCP support. It searches up the characters, looking for ranges of    case OP_NOT:
2848  characters in the "other" case. Each call returns the next one, updating the    GETCHARINCTEST(chr, code);
2849  start address. A character with multiple other cases is returned on its own    list[2] = chr;
2850  with a special return value.    list[3] = NOTACHAR;
2851      return code;
2852    
2853  Arguments:    case OP_CHARI:
2854    cptr        points to starting character value; updated    case OP_NOTI:
2855    d           end value    list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2856    ocptr       where to put start of othercase range    GETCHARINCTEST(chr, code);
2857    odptr       where to put end of othercase range    list[2] = chr;
2858    
2859  Yield:        -1 when no more  #ifdef SUPPORT_UCP
2860                 0 when a range is returned    if (chr < 128 || (chr < 256 && !utf))
2861                >0 the CASESET offset for char with multiple other cases      list[3] = fcc[chr];
2862                  in this case, ocptr contains the original    else
2863  */      list[3] = UCD_OTHERCASE(chr);
2864    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2865      list[3] = (chr < 256) ? fcc[chr] : chr;
2866    #else
2867      list[3] = fcc[chr];
2868    #endif
2869    
2870  static int    /* The othercase might be the same value. */
 get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,  
   pcre_uint32 *odptr)  
 {  
 pcre_uint32 c, othercase, next;  
 int co;  
2871    
2872  /* Find the first character that has an other case. If it has multiple other    if (chr == list[3])
2873  cases, return its case offset value. */      list[3] = NOTACHAR;
2874      else
2875        list[4] = NOTACHAR;
2876      return code;
2877    
2878  for (c = *cptr; c <= d; c++)  #ifdef SUPPORT_UCP
2879    {    case OP_PROP:
2880    if ((co = UCD_CASESET(c)) != 0)    case OP_NOTPROP:
2881      if (code[0] != PT_CLIST)
2882      {      {
2883      *ocptr = c++;   /* Character that has the set */      list[2] = code[0];
2884      *cptr = c;      /* Rest of input range */      list[3] = code[1];
2885      return co;      return code + 2;
2886      }      }
   if ((othercase = UCD_OTHERCASE(c)) != c) break;  
   }  
2887    
2888  if (c > d) return -1;  /* Reached end of range */    /* Convert only if we have enough space. */
2889    
2890  *ocptr = othercase;    clist_src = PRIV(ucd_caseless_sets) + code[1];
2891  next = othercase + 1;    clist_dest = list + 2;
2892      code += 2;
2893    
2894  for (++c; c <= d; c++)    do {
2895    {       if (clist_dest >= list + 8)
2896    if (UCD_OTHERCASE(c) != next) break;         {
2897    next++;         /* Early return if there is not enough space. This should never
2898    }         happen, since all clists are shorter than 5 character now. */
2899           list[2] = code[0];
2900           list[3] = code[1];
2901           return code;
2902           }
2903         *clist_dest++ = *clist_src;
2904         }
2905      while(*clist_src++ != NOTACHAR);
2906    
2907  *odptr = next - 1;     /* End of othercase range */    /* All characters are stored. The terminating NOTACHAR
2908  *cptr = c;             /* Rest of input range */    is copied form the clist itself. */
2909  return 0;  
2910      list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
2911      return code;
2912    #endif
2913    
2914      case OP_NCLASS:
2915      case OP_CLASS:
2916    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2917      case OP_XCLASS:
2918    
2919      if (c == OP_XCLASS)
2920        end = code + GET(code, 0);
2921      else
2922    #endif
2923        end = code + 32 / sizeof(pcre_uchar);
2924    
2925      switch(*end)
2926        {
2927        case OP_CRSTAR:
2928        case OP_CRMINSTAR:
2929        case OP_CRQUERY:
2930        case OP_CRMINQUERY:
2931        case OP_CRPOSSTAR:
2932        case OP_CRPOSQUERY:
2933        list[1] = TRUE;
2934        end++;
2935        break;
2936    
2937        case OP_CRPLUS:
2938        case OP_CRMINPLUS:
2939        case OP_CRPOSPLUS:
2940        end++;
2941        break;
2942    
2943        case OP_CRRANGE:
2944        case OP_CRMINRANGE:
2945        case OP_CRPOSRANGE:
2946        list[1] = (GET2(end, 1) == 0);
2947        end += 1 + 2 * IMM2_SIZE;
2948        break;
2949        }
2950      list[2] = end - code;
2951      return end;
2952      }
2953    return NULL;    /* Opcode not accepted */
2954  }  }
2955    
2956    
2957    
2958  /*************************************************  /*************************************************
2959  *        Check a character and a property        *  *    Scan further character sets for match       *
2960  *************************************************/  *************************************************/
2961    
2962  /* This function is called by check_auto_possessive() when a property item  /* Checks whether the base and the current opcode have a common character, in
2963  is adjacent to a fixed character.  which case the base cannot be possessified.
2964    
2965  Arguments:  Arguments:
2966    c            the character    code        points to the byte code
2967    ptype        the property type    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2968    pdata        the data for the type    cd          static compile data
2969    negated      TRUE if it's a negated property (\P or \p{^)    base_list   the data list of the base opcode
2970    
2971  Returns:       TRUE if auto-possessifying is OK  Returns:      TRUE if the auto-possessification is possible
2972  */  */
2973    
2974  static BOOL  static BOOL
2975  check_char_prop(pcre_uint32 c, int ptype, int pdata, BOOL negated)  compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
2976      const pcre_uint32* base_list, const pcre_uchar *base_end)
2977  {  {
2978  #ifdef SUPPORT_UCP  pcre_uchar c;
2979  const pcre_uint32 *p;  pcre_uint32 list[8];
2980  #endif  const pcre_uint32* chr_ptr;
2981    const pcre_uint32* ochr_ptr;
2982  const ucd_record *prop = GET_UCD(c);  const pcre_uint32* list_ptr;
2983    const pcre_uchar *next_code;
2984    const pcre_uint8 *class_bits;
2985    pcre_uint32 chr;
2986    
2987    /* Note: the base_list[1] contains whether the current opcode has greedy
2988    (represented by a non-zero value) quantifier. This is a different from
2989    other character type lists, which stores here that the character iterator
2990    matches to an empty string (also represented by a non-zero value). */
2991    
2992  switch(ptype)  for(;;)
2993    {    {
2994    case PT_LAMP:    c = *code;
   return (prop->chartype == ucp_Lu ||  
           prop->chartype == ucp_Ll ||  
           prop->chartype == ucp_Lt) == negated;  
2995    
2996    case PT_GC:    /* Skip over callouts */
   return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;  
2997    
2998    case PT_PC:    if (c == OP_CALLOUT)
2999    return (pdata == prop->chartype) == negated;      {
3000        code += PRIV(OP_lengths)[c];
3001        continue;
3002        }
3003    
3004    case PT_SC:    if (c == OP_ALT)
3005    return (pdata == prop->script) == negated;      {
3006        do code += GET(code, 1); while (*code == OP_ALT);
3007        c = *code;
3008        }
3009    
3010    /* These are specials */    switch(c)
3011        {
3012        case OP_END:
3013        case OP_KETRPOS:
3014        /* TRUE only in greedy case. The non-greedy case could be replaced by
3015        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3016        uses more memory, which we cannot get at this stage.) */
3017    
3018    case PT_ALNUM:      return base_list[1] != 0;
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||  
           PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;  
3019    
3020    case PT_SPACE:    /* Perl space */      case OP_KET:
3021    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||      /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3022            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)      it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3023            == negated;      cannot be converted to a possessive form. */
3024    
3025        if (base_list[1] == 0) return FALSE;
3026    
3027        switch(*(code - GET(code, 1)))
3028          {
3029          case OP_ASSERT:
3030          case OP_ASSERT_NOT:
3031          case OP_ASSERTBACK:
3032          case OP_ASSERTBACK_NOT:
3033          case OP_ONCE:
3034          case OP_ONCE_NC:
3035          /* Atomic sub-patterns and assertions can always auto-possessify their
3036          last iterator. */
3037          return TRUE;
3038          }
3039    
3040    case PT_PXSPACE:  /* POSIX space */      code += PRIV(OP_lengths)[c];
3041    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||      continue;
           c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||  
           c == CHAR_FF || c == CHAR_CR)  
           == negated;  
3042    
3043    case PT_WORD:      case OP_ONCE:
3044    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||      case OP_ONCE_NC:
3045            PRIV(ucp_gentype)[prop->chartype] == ucp_N ||      case OP_BRA:
3046            c == CHAR_UNDERSCORE) == negated;      case OP_CBRA:
3047        next_code = code;
3048  #ifdef SUPPORT_UCP      do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
   case PT_CLIST:  
   p = PRIV(ucd_caseless_sets) + prop->caseset;  
   for (;;)  
     {  
     if ((unsigned int)c < *p) return !negated;  
     if ((unsigned int)c == *p++) return negated;  
     }  
   break;  /* Control never reaches here */  
 #endif  
   }  
3049    
3050  return FALSE;      /* We do not support repeated brackets, because they can lead to
3051  }      infinite recursion. */
 #endif  /* SUPPORT_UCP */  
3052    
3053        if (*next_code != OP_KET) return FALSE;
3054    
3055        next_code = code + GET(code, 1);
3056        code += PRIV(OP_lengths)[c];
3057    
3058  /*************************************************      while (*next_code == OP_ALT)
3059  *     Check if auto-possessifying is possible    *        {
3060  *************************************************/        if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3061          code = next_code + 1 + LINK_SIZE;
3062          next_code += GET(next_code, 1);
3063          }
3064        continue;
3065    
3066        case OP_BRAZERO:
3067        case OP_BRAMINZERO:
3068    
3069  /* This function is called for unlimited repeats of certain items, to see      next_code = code + 1;
3070  whether the next thing could possibly match the repeated item. If not, it makes      if (*next_code != OP_BRA && *next_code != OP_CBRA)
3071  sense to automatically possessify the repeated item.        return FALSE;
3072    
3073  Arguments:      do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
   previous      pointer to the repeated opcode  
   utf           TRUE in UTF-8 / UTF-16 / UTF-32 mode  
   ptr           next character in pattern  
   options       options bits  
   cd            contains pointers to tables etc.  
3074    
3075  Returns:        TRUE if possessifying is wanted      /* We do not support repeated brackets, because they can lead to
3076  */      infinite recursion. */
3077        if (*next_code != OP_KET) return FALSE;
3078    
3079        /* The bracket content will be checked by the
3080        OP_BRA/OP_CBRA case above. */
3081        next_code += 1 + LINK_SIZE;
3082        if (!compare_opcodes(next_code, utf, cd, base_list, base_end)) return FALSE;
3083    
3084  static BOOL      code += PRIV(OP_lengths)[c];
3085  check_auto_possessive(const pcre_uchar *previous, BOOL utf,      continue;
3086    const pcre_uchar *ptr, int options, compile_data *cd)      }
 {  
 pcre_uint32 c = NOTACHAR;  
 pcre_uint32 next;  
 int escape;  
 int op_code = *previous++;  
3087    
3088  /* Skip whitespace and comments in extended mode */    /* Check for a supported opcode, and load its properties. */
3089    
3090  if ((options & PCRE_EXTENDED) != 0)    code = get_chr_property_list(code, utf, cd->fcc, list);
3091    {    if (code == NULL) return FALSE;    /* Unsupported */
3092    for (;;)  
3093      /* If either opcode is a small character list, set pointers for comparing
3094      characters from that list with another list, or with a property. */
3095    
3096      if (base_list[0] == OP_CHAR)
3097      {      {
3098      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      chr_ptr = base_list + 2;
3099      if (*ptr == CHAR_NUMBER_SIGN)      list_ptr = list;
3100        {      }
3101        ptr++;    else if (list[0] == OP_CHAR)
3102        while (*ptr != 0)      {
3103          {      chr_ptr = list + 2;
3104          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }      list_ptr = base_list;
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
       }  
     else break;  
3105      }      }
   }  
3106    
3107  /* If the next item is one that we can handle, get its value. A non-negative    /* Some property combinations also acceptable. Unicode property opcodes are
3108  value is a character, a negative value is an escape value. */    processed specially; the rest can be handled with a lookup table. */
3109    
3110  if (*ptr == CHAR_BACKSLASH)    else
3111    {      {
3112    int temperrorcode = 0;      pcre_uint32 leftop, rightop;
   escape = check_escape(&ptr, &next, &temperrorcode, cd->bracount, options, FALSE);  
   if (temperrorcode != 0) return FALSE;  
   ptr++;    /* Point after the escape sequence */  
   }  
 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;  
3113    
3114  /* Skip whitespace and comments in extended mode */      if (list[1] != 0) return FALSE;   /* Must match at least one character */
3115        leftop = base_list[0];
3116        rightop = list[0];
3117    
3118  if ((options & PCRE_EXTENDED) != 0)  #ifdef SUPPORT_UCP
3119    {      if (leftop == OP_PROP || leftop == OP_NOTPROP)
   for (;;)  
     {  
     while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;  
     if (*ptr == CHAR_NUMBER_SIGN)  
3120        {        {
3121        ptr++;        if (rightop == OP_EOD) return TRUE;
3122        while (*ptr != 0)        if (rightop == OP_PROP || rightop == OP_NOTPROP)
3123          {          {
3124          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          int n;
3125          ptr++;          const pcre_uint8 *p;
3126  #ifdef SUPPORT_UTF          BOOL same = leftop == rightop;
3127          if (utf) FORWARDCHAR(ptr);          BOOL lisprop = leftop == OP_PROP;
3128  #endif          BOOL risprop = rightop == OP_PROP;
3129            BOOL bothprop = lisprop && risprop;
3130    
3131            /* There's a table that specifies how each combination is to be
3132            processed:
3133              0   Always return FALSE (never auto-possessify)
3134              1   Character groups are distinct (possessify if both are OP_PROP)
3135              2   Check character categories in the same group (general or particular)
3136              3   Return TRUE if the two opcodes are not the same
3137              ... see comments below
3138            */
3139    
3140            n = propposstab[base_list[2]][list[2]];
3141            switch(n)
3142              {
3143              case 0: return FALSE;
3144              case 1: return bothprop;
3145              case 2: return (base_list[3] == list[3]) != same;
3146              case 3: return !same;
3147    
3148              case 4:  /* Left general category, right particular category */
3149              return risprop && catposstab[base_list[3]][list[3]] == same;
3150    
3151              case 5:  /* Right general category, left particular category */
3152              return lisprop && catposstab[list[3]][base_list[3]] == same;
3153    
3154              /* This code is logically tricky. Think hard before fiddling with it.
3155              The posspropstab table has four entries per row. Each row relates to
3156              one of PCRE's special properties such as ALNUM or SPACE or WORD.
3157              Only WORD actually needs all four entries, but using repeats for the
3158              others means they can all use the same code below.
3159    
3160              The first two entries in each row are Unicode general categories, and
3161              apply always, because all the characters they include are part of the
3162              PCRE character set. The third and fourth entries are a general and a
3163              particular category, respectively, that include one or more relevant
3164              characters. One or the other is used, depending on whether the check
3165              is for a general or a particular category. However, in both cases the
3166              category contains more characters than the specials that are defined
3167              for the property being tested against. Therefore, it cannot be used
3168              in a NOTPROP case.
3169    
3170              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3171              Underscore is covered by ucp_P or ucp_Po. */
3172    
3173              case 6:  /* Left alphanum vs right general category */
3174              case 7:  /* Left space vs right general category */
3175              case 8:  /* Left word vs right general category */
3176              p = posspropstab[n-6];
3177              return risprop && lisprop ==
3178                (list[3] != p[0] &&
3179                 list[3] != p[1] &&
3180                (list[3] != p[2] || !lisprop));
3181    
3182              case 9:   /* Right alphanum vs left general category */
3183              case 10:  /* Right space vs left general category */
3184              case 11:  /* Right word vs left general category */
3185              p = posspropstab[n-9];
3186              return lisprop && risprop ==
3187                (base_list[3] != p[0] &&
3188                 base_list[3] != p[1] &&
3189                (base_list[3] != p[2] || !risprop));
3190    
3191              case 12:  /* Left alphanum vs right particular category */
3192              case 13:  /* Left space vs right particular category */
3193              case 14:  /* Left word vs right particular category */
3194              p = posspropstab[n-12];
3195              return risprop && lisprop ==
3196                (catposstab[p[0]][list[3]] &&
3197                 catposstab[p[1]][list[3]] &&
3198                (list[3] != p[3] || !lisprop));
3199    
3200              case 15:  /* Right alphanum vs left particular category */
3201              case 16:  /* Right space vs left particular category */
3202              case 17:  /* Right word vs left particular category */
3203              p = posspropstab[n-15];
3204              return lisprop && risprop ==
3205                (catposstab[p[0]][base_list[3]] &&
3206                 catposstab[p[1]][base_list[3]] &&
3207                (base_list[3] != p[3] || !risprop));
3208              }
3209          }          }
3210          return FALSE;
3211        }        }
     else break;  
     }  
   }  
3212    
3213  /* If the next thing is itself optional, we have to give up. */      else
3214    #endif  /* SUPPORT_UCP */
 if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  
   STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)  
     return FALSE;  
   
 /* If the previous item is a character, get its value. */  
3215    
3216  if (op_code == OP_CHAR || op_code == OP_CHARI ||      return leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3217      op_code == OP_NOT || op_code == OP_NOTI)             rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3218    //if (escape == 0) switch(op_code)             autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3219    {      }
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   }  
3220    
3221  /* Now compare the next item with the previous opcode. First, handle cases when    /* Control reaches here only if one of the items is a small character list.
3222  the next item is a character. */    All characters are checked against the other side. */
3223    
3224  if (escape == 0)    do
   {  
   /* 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. */  
   
 #ifdef SUPPORT_UCP  
   if (utf && c != NOTACHAR && (options & PCRE_CASELESS) != 0)  
3225      {      {
3226      int ocs = UCD_CASESET(next);      chr = *chr_ptr;
     if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);  
     }  
 #endif  
3227    
3228    switch(op_code)      switch(list_ptr[0])
3229      {        {
3230      case OP_CHAR:        case OP_CHAR:
3231      return c != next;        ochr_ptr = list_ptr + 2;
3232          do
3233            {
3234            if (chr == *ochr_ptr) return FALSE;
3235            ochr_ptr++;
3236            }
3237          while(*ochr_ptr != NOTACHAR);
3238          break;
3239    
3240          case OP_NOT:
3241          ochr_ptr = list_ptr + 2;
3242          do
3243            {
3244            if (chr == *ochr_ptr)
3245              break;
3246            ochr_ptr++;
3247            }
3248          while(*ochr_ptr != NOTACHAR);
3249          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3250          break;
3251    
3252      /* For CHARI (caseless character) we must check the other case. If we have        /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3253      Unicode property support, we can use it to test the other case of        set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
     high-valued characters. We know that next can have only one other case,  
     because multi-other-case characters are dealt with above. */  
3254    
3255      case OP_CHARI:        case OP_DIGIT:
3256      if (c == next) return FALSE;        if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3257  #ifdef SUPPORT_UTF        break;
3258      if (utf)  
3259        {        case OP_NOT_DIGIT:
3260        pcre_uint32 othercase;        if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3261        if (next < 128) othercase = cd->fcc[next]; else        break;
3262  #ifdef SUPPORT_UCP  
3263        othercase = UCD_OTHERCASE(next);        case OP_WHITESPACE:
3264  #else        if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3265        othercase = NOTACHAR;        break;
3266    
3267          case OP_NOT_WHITESPACE:
3268          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3269          break;
3270    
3271          case OP_WORDCHAR:
3272          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3273          break;
3274    
3275          case OP_NOT_WORDCHAR:
3276          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3277          break;
3278    
3279          case OP_HSPACE:
3280          switch(chr)
3281            {
3282            HSPACE_CASES: return FALSE;
3283            default: break;
3284            }
3285          break;
3286    
3287          case OP_NOT_HSPACE:
3288          switch(chr)
3289            {
3290            HSPACE_CASES: break;
3291            default: return FALSE;
3292            }
3293          break;
3294    
3295          case OP_ANYNL:
3296          case OP_VSPACE:
3297          switch(chr)
3298            {
3299            VSPACE_CASES: return FALSE;
3300            default: break;
3301            }
3302          break;
3303    
3304          case OP_NOT_VSPACE:
3305          switch(chr)
3306            {
3307            VSPACE_CASES: break;
3308            default: return FALSE;
3309            }
3310          break;
3311    
3312          case OP_DOLL:
3313          case OP_EODN:
3314          switch (chr)
3315            {
3316            case CHAR_CR:
3317            case CHAR_LF:
3318            case CHAR_VT:
3319            case CHAR_FF:
3320            case CHAR_NEL:
3321    #ifndef EBCDIC
3322            case 0x2028:
3323            case 0x2029:
3324    #endif  /* Not EBCDIC */
3325            return FALSE;
3326            }
3327          break;
3328    
3329          case OP_EOD:    /* Can always possessify before \z */
3330          break;
3331    
3332          case OP_PROP:
3333          case OP_NOTPROP:
3334          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3335                list_ptr[0] == OP_NOTPROP))
3336            return FALSE;
3337          break;
3338    
3339          case OP_NCLASS:
3340          if (chr > 255) return FALSE;
3341          /* Fall through */
3342    
3343          case OP_CLASS:
3344          if (chr > 255) break;
3345          class_bits = (pcre_uint8 *)((list_ptr == list ? code : base_end) - list_ptr[2]);
3346          if ((class_bits[chr >> 3] & (1 << (chr & 7))) != 0)
3347            return FALSE;
3348          break;
3349    
3350    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3351          case OP_XCLASS:
3352          if (list_ptr != list) return FALSE;   /* Class is first opcode */
3353          if (PRIV(xclass)(chr, code - list_ptr[2] + LINK_SIZE, utf))
3354            return FALSE;
3355          break;
3356  #endif  #endif
3357        return c != othercase;  
3358          default:
3359          return FALSE;
3360        }        }
3361      else  
3362  #endif  /* SUPPORT_UTF */      chr_ptr++;
3363      return (c != TABLE_GET(next, cd->fcc, next));  /* Not UTF */      }
3364      while(*chr_ptr != NOTACHAR);
3365    
3366      /* At least one character must be matched from this opcode. */
3367    
3368      if (list[1] == 0) return TRUE;
3369      }
3370    
3371    return FALSE;
3372    }
3373    
3374    
3375    
3376    /*************************************************
3377    *    Scan compiled regex for auto-possession     *
3378    *************************************************/
3379    
3380    /* Replaces single character iterations with their possessive alternatives
3381    if appropriate. This function modifies the compiled opcode!
3382    
3383    Arguments:
3384      code        points to start of the byte code
3385      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3386      cd          static compile data
3387    
3388    Returns:      nothing
3389    */
3390    
3391    static void
3392    auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3393    {
3394    register pcre_uchar c, d;
3395    const pcre_uchar *end;
3396    pcre_uchar *repeat_code;
3397    pcre_uint32 list[8];
3398    
3399    for (;;)
3400      {
3401      c = *code;
3402    
3403      if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3404        {
3405        c -= get_repeat_base(c) - OP_STAR;
3406        end = (c <= OP_MINUPTO) ?
3407          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3408        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3409    
3410        if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3411          {
3412          switch(c)
3413            {
3414            case OP_STAR:
3415            *code += OP_POSSTAR - OP_STAR;
3416            break;
3417    
3418            case OP_MINSTAR:
3419            *code += OP_POSSTAR - OP_MINSTAR;
3420            break;
3421    
3422            case OP_PLUS:
3423            *code += OP_POSPLUS - OP_PLUS;
3424            break;
3425    
3426            case OP_MINPLUS:
3427            *code += OP_POSPLUS - OP_MINPLUS;
3428            break;
3429    
3430            case OP_QUERY:
3431            *code += OP_POSQUERY - OP_QUERY;
3432            break;
3433    
3434            case OP_MINQUERY:
3435            *code += OP_POSQUERY - OP_MINQUERY;
3436            break;
3437    
3438            case OP_UPTO:
3439            *code += OP_POSUPTO - OP_UPTO;
3440            break;
3441    
3442            case OP_MINUPTO:
3443            *code += OP_MINUPTO - OP_UPTO;
3444            break;
3445            }
3446          }
3447        c = *code;
3448        }
3449      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3450        {
3451    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3452        if (c == OP_XCLASS)
3453          repeat_code = code + 1 + GET(code, 1);
3454        else
3455    #endif
3456          repeat_code = code + 1 + (32 / sizeof(pcre_uchar));
3457    
3458        d = *repeat_code;
3459        if (d >= OP_CRSTAR && d <= OP_CRMINRANGE)
3460          {
3461          /* end must not be NULL. */
3462          end = get_chr_property_list(code, utf, cd->fcc, list);
3463    
3464          list[1] = d == OP_CRSTAR || d == OP_CRPLUS || d == OP_CRQUERY ||
3465            d == OP_CRRANGE;
3466    
3467          if (compare_opcodes(end, utf, cd, list, end))
3468            {
3469            switch (d)
3470              {
3471              case OP_CRSTAR:
3472              *repeat_code = OP_CRPOSSTAR;
3473              break;
3474    
3475              case OP_CRPLUS:
3476              *repeat_code = OP_CRPOSPLUS;
3477              break;
3478    
3479              case OP_CRQUERY:
3480              *repeat_code = OP_CRPOSQUERY;
3481              break;
3482    
3483              case OP_CRRANGE:
3484              *repeat_code = OP_CRPOSRANGE;
3485              break;
3486              }
3487            }
3488          }
3489        }
3490    
3491      switch(c)
3492        {
3493        case OP_END:
3494        return;
3495    
3496        case OP_TYPESTAR:
3497        case OP_TYPEMINSTAR:
3498        case OP_TYPEPLUS:
3499        case OP_TYPEMINPLUS:
3500        case OP_TYPEQUERY:
3501        case OP_TYPEMINQUERY:
3502        case OP_TYPEPOSSTAR:
3503        case OP_TYPEPOSPLUS:
3504        case OP_TYPEPOSQUERY:
3505        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3506        break;
3507    
3508        case OP_TYPEUPTO:
3509        case OP_TYPEMINUPTO:
3510        case OP_TYPEEXACT:
3511        case OP_TYPEPOSUPTO:
3512        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3513          code += 2;
3514        break;
3515    
3516    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3517        case OP_XCLASS:
3518        code += GET(code, 1);
3519        break;
3520    #endif
3521    
3522        case OP_MARK:
3523        case OP_PRUNE_ARG:
3524        case OP_SKIP_ARG:
3525        case OP_THEN_ARG:
3526        code += code[1];
3527        break;
3528        }
3529    
3530      /* Add in the fixed length from the table */
3531    
3532      code += PRIV(OP_lengths)[c];
3533    
3534      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3535      a multi-byte character. The length in the table is a minimum, so we have to
3536      arrange to skip the extra bytes. */
3537    
3538    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3539      if (utf) switch(c)
3540        {
3541        case OP_CHAR:
3542        case OP_CHARI:
3543      case OP_NOT:      case OP_NOT:
     return c == next;  
   
3544      case OP_NOTI:      case OP_NOTI:
3545      if (c == next) return TRUE;      case OP_STAR:
3546  #ifdef SUPPORT_UTF      case OP_MINSTAR:
3547      if (utf)      case OP_PLUS:
3548        {      case OP_MINPLUS:
3549        pcre_uint32 othercase;      case OP_QUERY:
3550        if (next < 128) othercase = cd->fcc[next]; else      case OP_MINQUERY:
3551  #ifdef SUPPORT_UCP      case OP_UPTO:
3552        othercase = UCD_OTHERCASE(next);      case OP_MINUPTO:
3553        case OP_EXACT:
3554        case OP_POSSTAR:
3555        case OP_POSPLUS:
3556        case OP_POSQUERY:
3557        case OP_POSUPTO:
3558        case OP_STARI:
3559        case OP_MINSTARI:
3560        case OP_PLUSI:
3561        case OP_MINPLUSI:
3562        case OP_QUERYI:
3563        case OP_MINQUERYI:
3564        case OP_UPTOI:
3565        case OP_MINUPTOI:
3566        case OP_EXACTI:
3567        case OP_POSSTARI:
3568        case OP_POSPLUSI:
3569        case OP_POSQUERYI:
3570        case OP_POSUPTOI:
3571        case OP_NOTSTAR:
3572        case OP_NOTMINSTAR:
3573        case OP_NOTPLUS:
3574        case OP_NOTMINPLUS:
3575        case OP_NOTQUERY:
3576        case OP_NOTMINQUERY:
3577        case OP_NOTUPTO:
3578        case OP_NOTMINUPTO:
3579        case OP_NOTEXACT:
3580        case OP_NOTPOSSTAR:
3581        case OP_NOTPOSPLUS:
3582        case OP_NOTPOSQUERY:
3583        case OP_NOTPOSUPTO:
3584        case OP_NOTSTARI:
3585        case OP_NOTMINSTARI:
3586        case OP_NOTPLUSI:
3587        case OP_NOTMINPLUSI:
3588        case OP_NOTQUERYI:
3589        case OP_NOTMINQUERYI:
3590        case OP_NOTUPTOI:
3591        case OP_NOTMINUPTOI:
3592        case OP_NOTEXACTI:
3593        case OP_NOTPOSSTARI:
3594        case OP_NOTPOSPLUSI:
3595        case OP_NOTPOSQUERYI:
3596        case OP_NOTPOSUPTOI:
3597        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3598        break;
3599        }
3600  #else  #else
3601        othercase = NOTACHAR;    (void)(utf);  /* Keep compiler happy by referencing function argument */
3602  #endif  #endif
3603        return c == othercase;    }
3604    }
3605    
3606    
3607    
3608    /*************************************************
3609    *           Check for POSIX class syntax         *
3610    *************************************************/
3611    
3612    /* This function is called when the sequence "[:" or "[." or "[=" is
3613    encountered in a character class. It checks whether this is followed by a
3614    sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3615    reach an unescaped ']' without the special preceding character, return FALSE.
3616    
3617    Originally, this function only recognized a sequence of letters between the
3618    terminators, but it seems that Perl recognizes any sequence of characters,
3619    though of course unknown POSIX names are subsequently rejected. Perl gives an
3620    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3621    didn't consider this to be a POSIX class. Likewise for [:1234:].
3622    
3623    The problem in trying to be exactly like Perl is in the handling of escapes. We
3624    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3625    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3626    below handles the special case of \], but does not try to do any other escape
3627    processing. This makes it different from Perl for cases such as [:l\ower:]
3628    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3629    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
3630    I think.
3631    
3632    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3633    It seems that the appearance of a nested POSIX class supersedes an apparent
3634    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3635    a digit.
3636    
3637    In Perl, unescaped square brackets may also appear as part of class names. For
3638    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3639    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3640    seem right at all. PCRE does not allow closing square brackets in POSIX class
3641    names.
3642    
3643    Arguments:
3644      ptr      pointer to the initial [
3645      endptr   where to return the end pointer
3646    
3647    Returns:   TRUE or FALSE
3648    */
3649    
3650    static BOOL
3651    check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3652    {
3653    pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
3654    terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
3655    for (++ptr; *ptr != CHAR_NULL; ptr++)
3656      {
3657      if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3658        ptr++;
3659      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3660      else
3661        {
3662        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3663          {
3664          *endptr = ptr;
3665          return TRUE;
3666        }        }
3667      else      if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3668  #endif  /* SUPPORT_UTF */           (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3669      return (c == TABLE_GET(next, cd->fcc, next));  /* Not UTF */            ptr[1] == CHAR_EQUALS_SIGN) &&
3670            check_posix_syntax(ptr, endptr))
3671          return FALSE;
3672        }
3673      }
3674    return FALSE;
3675    }
3676    
     /* 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. */  
3677    
     case OP_DIGIT:  
     return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;  
3678    
     case OP_NOT_DIGIT:  
     return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;  
3679    
3680      case OP_WHITESPACE:  /*************************************************
3681      return next > 255 || (cd->ctypes[next] & ctype_space) == 0;  *          Check POSIX class name                *
3682    *************************************************/
3683    
3684      case OP_NOT_WHITESPACE:  /* This function is called to check the name given in a POSIX-style class entry
3685      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;  such as [:alnum:].
3686    
3687      case OP_WORDCHAR:  Arguments:
3688      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;    ptr        points to the first letter
3689      len        the length of the name
3690    
3691      case OP_NOT_WORDCHAR:  Returns:     a value representing the name, or -1 if unknown
3692      return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;  */
3693    
3694      case OP_HSPACE:  static int
3695      case OP_NOT_HSPACE:  check_posix_name(const pcre_uchar *ptr, int len)
3696      switch(next)  {
3697        {  const char *pn = posix_names;
3698        HSPACE_CASES:  register int yield = 0;
3699        return op_code == OP_NOT_HSPACE;  while (posix_name_lengths[yield] != 0)
3700      {
3701      if (len == posix_name_lengths[yield] &&
3702        STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3703      pn += posix_name_lengths[yield] + 1;
3704      yield++;
3705      }
3706    return -1;
3707    }
3708    
       default:  
       return op_code != OP_NOT_HSPACE;  
       }  
3709    
3710      case OP_ANYNL:  /*************************************************
3711      case OP_VSPACE:  *    Adjust OP_RECURSE items in repeated group   *
3712      case OP_NOT_VSPACE:  *************************************************/
     switch(next)  
       {  
       VSPACE_CASES:  
       return op_code == OP_NOT_VSPACE;  
3713    
3714        default:  /* OP_RECURSE items contain an offset from the start of the regex to the group
3715        return op_code != OP_NOT_VSPACE;  that is referenced. This means that groups can be replicated for fixed
3716        }  repetition simply by copying (because the recursion is allowed to refer to
3717    earlier groups that are outside the current group). However, when a group is
3718    optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3719    inserted before it, after it has been compiled. This means that any OP_RECURSE
3720    items within it that refer to the group itself or any contained groups have to
3721    have their offsets adjusted. That one of the jobs of this function. Before it
3722    is called, the partially compiled regex must be temporarily terminated with
3723    OP_END.
3724    
3725  #ifdef SUPPORT_UCP  This function has been extended with the possibility of forward references for
3726      case OP_PROP:  recursions and subroutine calls. It must also check the list of such references
3727      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
3728    the current group is on this list, it adjusts the offset in the list, not the
3729    value in the reference (which is a group number).
3730    
3731      case OP_NOTPROP:  Arguments:
3732      return check_char_prop(next, previous[0], previous[1], TRUE);    group      points to the start of the group
3733  #endif    adjust     the amount by which the group is to be moved
3734      utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
3735      cd         contains pointers to tables etc.
3736      save_hwm   the hwm forward reference pointer at the start of the group
3737    
3738      default:  Returns:     nothing
3739      return FALSE;  */
     }  
   }  
3740    
3741  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP  static void
3742  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,
3743  generated only when PCRE_UCP is *not* set, that is, when only ASCII    pcre_uchar *save_hwm)
3744  characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are  {
3745  replaced by OP_PROP codes when PCRE_UCP is set. */  pcre_uchar *ptr = group;
3746    
3747  switch(op_code)  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3748    {    {
3749    case OP_CHAR:    int offset;
3750    case OP_CHARI:    pcre_uchar *hc;
3751    switch(escape)  
3752      /* See if this recursion is on the forward reference list. If so, adjust the
3753      reference. */
3754    
3755      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3756      {      {
3757      case ESC_d:      offset = (int)GET(hc, 0);
3758      return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;      if (cd->start_code + offset == ptr + 1)
3759          {
3760          PUT(hc, 0, offset + adjust);
3761          break;
3762          }
3763        }
3764    
3765      case ESC_D:    /* Otherwise, adjust the recursion offset if it's after the start of this
3766      return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;    group. */
3767    
3768      case ESC_s:    if (hc >= cd->hwm)
3769      return c > 255 || (cd->ctypes[c] & ctype_space) == 0;      {
3770        offset = (int)GET(ptr, 1);
3771        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3772        }
3773    
3774      case ESC_S:    ptr += 1 + LINK_SIZE;
3775      return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;    }
3776    }
3777    
     case ESC_w:  
     return c > 255 || (cd->ctypes[c] & ctype_word) == 0;  
3778    
     case ESC_W:  
     return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;  
3779    
3780      case ESC_h:  /*************************************************
3781      case ESC_H:  *        Insert an automatic callout point       *
3782      switch(c)  *************************************************/
       {  
       HSPACE_CASES:  
       return escape != ESC_h;  
   
       default:  
       return escape == ESC_h;  
       }  
3783    
3784      case ESC_v:  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
3785      case ESC_V:  callout points before each pattern item.
     switch(c)  
       {  
       VSPACE_CASES:  
       return escape != ESC_v;  
3786    
3787        default:  Arguments:
3788        return escape == ESC_v;    code           current code pointer
3789        }    ptr            current pattern pointer
3790      cd             pointers to tables etc
3791    
3792      /* When PCRE_UCP is set, these values get generated for \d etc. Find  Returns:         new code pointer
3793      their substitutions and process them. The result will always be either  */
     ESC_p or ESC_P. Then fall through to process those values. */  
3794    
3795  #ifdef SUPPORT_UCP  static pcre_uchar *
3796      case ESC_du:  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
3797      case ESC_DU:  {
3798      case ESC_wu:  *code++ = OP_CALLOUT;
3799      case ESC_WU:  *code++ = 255;
3800      case ESC_su:  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
3801      case ESC_SU:  PUT(code, LINK_SIZE, 0);                       /* Default length */
3802        {  return code + 2 * LINK_SIZE;
3803        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 */  
3804    
     case ESC_p:  
     case ESC_P:  
       {  
       int ptype, pdata, errorcodeptr;  
       BOOL negated;  
3805    
       ptr--;      /* Make ptr point at the p or P */  
       ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);  
       if (ptype < 0) return FALSE;  
       ptr++;      /* Point past the final curly ket */  
3806    
3807        /* If the property item is optional, we have to give up. (When generated  /*************************************************
3808        from \d etc by PCRE_UCP, this test will have been applied much earlier,  *         Complete a callout item                *
3809        to the original \d etc. At this point, ptr will point to a zero byte. */  *************************************************/
3810    
3811        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  /* A callout item contains the length of the next item in the pattern, which
3812          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
3813            return FALSE;  for both automatic and manual callouts.
3814    
3815        /* Do the property check. */  Arguments:
3816      previous_callout   points to previous callout item
3817      ptr                current pattern pointer
3818      cd                 pointers to tables etc
3819    
3820        return check_char_prop(c, ptype, pdata, (escape == ESC_P) != negated);  Returns:             nothing
3821        }  */
 #endif  
3822    
3823      default:  static void
3824      return FALSE;  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
3825      }  {
3826    int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
3827    PUT(previous_callout, 2 + LINK_SIZE, length);
3828    }
3829    
   /* 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.) */  
3830    
   case OP_DIGIT:  
   return escape == ESC_D || escape == ESC_s || escape == ESC_W ||  
          escape == ESC_h || escape == ESC_v || escape == ESC_R;  
3831    
3832    case OP_NOT_DIGIT:  #ifdef SUPPORT_UCP
3833    return escape == ESC_d;  /*************************************************
3834    *           Get othercase range                  *
3835    *************************************************/
3836    
3837    case OP_WHITESPACE:  /* This function is passed the start and end of a class range, in UTF-8 mode
3838    return escape == ESC_S || escape == ESC_d || escape == ESC_w;  with UCP support. It searches up the characters, looking for ranges of
3839    characters in the "other" case. Each call returns the next one, updating the
3840    start address. A character with multiple other cases is returned on its own
3841    with a special return value.
3842    
3843    case OP_NOT_WHITESPACE:  Arguments:
3844    return escape == ESC_s || escape == ESC_h || escape == ESC_v || escape == ESC_R;    cptr        points to starting character value; updated
3845      d           end value
3846      ocptr       where to put start of othercase range
3847      odptr       where to put end of othercase range
3848    
3849    case OP_HSPACE:  Yield:        -1 when no more
3850    return escape == ESC_S || escape == ESC_H || escape == ESC_d ||                 0 when a range is returned
3851           escape == ESC_w || escape == ESC_v || escape == ESC_R;                >0 the CASESET offset for char with multiple other cases
3852                    in this case, ocptr contains the original
3853    */
3854    
3855    case OP_NOT_HSPACE:  static int
3856    return escape == ESC_h;  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
3857      pcre_uint32 *odptr)
3858    {
3859    pcre_uint32 c, othercase, next;
3860    unsigned int co;
3861    
3862    /* 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
3863    case OP_ANYNL:  cases, return its case offset value. */
   case OP_VSPACE:  
   return escape == ESC_V || escape == ESC_d || escape == ESC_w;  
3864    
3865    case OP_NOT_VSPACE:  for (c = *cptr; c <= d; c++)
3866    return escape == ESC_v || escape == ESC_R;    {
3867      if ((co = UCD_CASESET(c)) != 0)
3868        {
3869        *ocptr = c++;   /* Character that has the set */
3870        *cptr = c;      /* Rest of input range */
3871        return (int)co;
3872        }
3873      if ((othercase = UCD_OTHERCASE(c)) != c) break;
3874      }
3875    
3876    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;  
3877    
3878    case OP_NOT_WORDCHAR:  *ocptr = othercase;
3879    return escape == ESC_w || escape == ESC_d;  next = othercase + 1;
3880    
3881    default:  for (++c; c <= d; c++)
3882    return FALSE;    {
3883      if (UCD_OTHERCASE(c) != next) break;
3884      next++;
3885    }    }
3886    
3887  /* Control does not reach here */  *odptr = next - 1;     /* End of othercase range */
3888    *cptr = c;             /* Rest of input range */
3889    return 0;
3890  }  }
3891    #endif  /* SUPPORT_UCP */
3892    
3893    
3894    
# Line 3418  switch(op_code) Line 3897  switch(op_code)
3897  *************************************************/  *************************************************/
3898    
3899  /* This function packages up the logic of adding a character or range of  /* This function packages up the logic of adding a character or range of
3900  characters to a class. The character values in the arguments will be within the  characters to a class. The character values in the arguments will be within the
3901  valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is  valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
3902  mutually recursive with the function immediately below.  mutually recursive with the function immediately below.
3903    
3904  Arguments:  Arguments:
3905    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
3906    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
3907    options       the options word    options       the options word
3908    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
3909    start         start of range character    start         start of range character
3910    end           end of range character    end           end of range character
3911    
3912  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
3913                  the pointer to extra data is updated                  the pointer to extra data is updated
3914  */  */
# Line 3441  add_to_class(pcre_uint8 *classbits, pcre Line 3920  add_to_class(pcre_uint8 *classbits, pcre
3920  pcre_uint32 c;  pcre_uint32 c;
3921  int n8 = 0;  int n8 = 0;
3922    
3923  /* If caseless matching is required, scan the range and process alternate  /* If caseless matching is required, scan the range and process alternate
3924  cases. In Unicode, there are 8-bit characters that have alternate cases that  cases. In Unicode, there are 8-bit characters that have alternate cases that
3925  are greater than 255 and vice-versa. Sometimes we can just extend the original  are greater than 255 and vice-versa. Sometimes we can just extend the original
3926  range. */  range. */
3927    
3928  if ((options & PCRE_CASELESS) != 0)  if ((options & PCRE_CASELESS) != 0)
3929    {    {
3930  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3931    if ((options & PCRE_UTF8) != 0)    if ((options & PCRE_UTF8) != 0)
3932      {      {
3933      int rc;      int rc;
3934      pcre_uint32 oc, od;      pcre_uint32 oc, od;
3935    
3936      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
3937      c = start;      c = start;
3938    
3939      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
3940        {        {
3941        /* Handle a single character that has more than one other case. */        /* Handle a single character that has more than one other case. */
3942    
3943        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
3944          PRIV(ucd_caseless_sets) + rc, oc);          PRIV(ucd_caseless_sets) + rc, oc);
3945    
3946        /* Do nothing if the other case range is within the original range. */        /* Do nothing if the other case range is within the original range. */
3947    
3948        else if (oc >= start && od <= end) continue;        else if (oc >= start && od <= end) continue;
3949    
3950        /* Extend the original range if there is overlap, noting that if oc < c, we        /* Extend the original range if there is overlap, noting that if oc < c, we
3951        can't have od > end because a subrange is always shorter than the basic        can't have od > end because a subrange is always shorter than the basic
3952        range. Otherwise, use a recursive call to add the additional range. */        range. Otherwise, use a recursive call to add the additional range. */
3953    
3954        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
3955        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
3956        else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);        else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
# Line 3481  if ((options & PCRE_CASELESS) != 0) Line 3960  if ((options & PCRE_CASELESS) != 0)
3960  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3961    
3962    /* Not UTF-mode, or no UCP */    /* Not UTF-mode, or no UCP */
3963    
3964    for (c = start; c <= end && c < 256; c++)    for (c = start; c <= end && c < 256; c++)
3965      {      {
3966      SETBIT(classbits, cd->fcc[c]);      SETBIT(classbits, cd->fcc[c]);
3967      n8++;      n8++;
3968      }      }
3969    }    }
3970    
3971  /* Now handle the original range. Adjust the final value according to the bit  /* Now handle the original range. Adjust the final value according to the bit
3972  length - this means that the same lists of (e.g.) horizontal spaces can be used  length - this means that the same lists of (e.g.) horizontal spaces can be used
3973  in all cases. */  in all cases. */
# Line 3514  if (end < 0x100) Line 3993  if (end < 0x100)
3993    {    {
3994    for (c = start; c <= end; c++)    for (c = start; c <= end; c++)
3995      {      {
3996      n8++;      n8++;
3997      SETBIT(classbits, c);      SETBIT(classbits, c);
3998      }      }
3999    }    }
4000    
4001  else  else
4002    {    {
4003    pcre_uchar *uchardata = *uchardptr;    pcre_uchar *uchardata = *uchardptr;
4004    
4005  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4006    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
4007      {      {
4008      if (start < end)      if (start < end)
4009        {        {
4010        *uchardata++ = XCL_RANGE;        *uchardata++ = XCL_RANGE;
4011        uchardata += PRIV(ord2utf)(start, uchardata);        uchardata += PRIV(ord2utf)(start, uchardata);
4012        uchardata += PRIV(ord2utf)(end, uchardata);        uchardata += PRIV(ord2utf)(end, uchardata);
4013        }        }
4014      else if (start == end)      else if (start == end)
4015        {        {
4016        *uchardata++ = XCL_SINGLE;        *uchardata++ = XCL_SINGLE;
4017        uchardata += PRIV(ord2utf)(start, uchardata);        uchardata += PRIV(ord2utf)(start, uchardata);
4018        }        }
4019      }      }
4020    else    else
4021  #endif  /* SUPPORT_UTF */  #endif  /* SUPPORT_UTF */
4022    
4023    /* Without UTF support, character values are constrained by the bit length,    /* Without UTF support, character values are constrained by the bit length,
4024    and can only be > 256 for 16-bit and 32-bit libraries. */    and can only be > 256 for 16-bit and 32-bit libraries. */
4025    
4026  #ifdef COMPILE_PCRE8  #ifdef COMPILE_PCRE8
4027      {}      {}
4028  #else  #else
4029    if (start < end)    if (start < end)
4030      {      {
4031      *uchardata++ = XCL_RANGE;      *uchardata++ = XCL_RANGE;
# Line 3557  else Line 4036  else
4036      {      {
4037      *uchardata++ = XCL_SINGLE;      *uchardata++ = XCL_SINGLE;
4038      *uchardata++ = start;      *uchardata++ = start;
4039      }      }
4040  #endif  #endif
4041    
4042    *uchardptr = uchardata;   /* Updata extra data pointer */    *uchardptr = uchardata;   /* Updata extra data pointer */
4043    }    }
4044    
4045  return n8;    /* Number of 8-bit characters */  return n8;    /* Number of 8-bit characters */
4046  }  }
4047    
4048    
4049    
4050    
4051  /*************************************************  /*************************************************
4052  *        Add a list of characters to a class     *  *        Add a list of characters to a class     *
4053  *************************************************/  *************************************************/
4054    
4055  /* This function is used for adding a list of case-equivalent characters to a  /* This function is used for adding a list of case-equivalent characters to a
4056  class, and also for adding a list of horizontal or vertical whitespace. If the  class, and also for adding a list of horizontal or vertical whitespace. If the
4057  list is in order (which it should be), ranges of characters are detected and  list is in order (which it should be), ranges of characters are detected and
4058  handled appropriately. This function is mutually recursive with the function  handled appropriately. This function is mutually recursive with the function
# Line 3583  Arguments: Line 4062  Arguments:
4062    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4063    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4064    options       the options word    options       the options word
4065    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4066    p             points to row of 32-bit values, terminated by NOTACHAR    p             points to row of 32-bit values, terminated by NOTACHAR
4067    except        character to omit; this is used when adding lists of    except        character to omit; this is used when adding lists of
4068                    case-equivalent characters to avoid including the one we                    case-equivalent characters to avoid including the one we
4069                    already know about                    already know about
4070    
4071  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4072                  the pointer to extra data is updated                  the pointer to extra data is updated
4073  */  */
# Line 3602  while (p[0] < NOTACHAR) Line 4081  while (p[0] < NOTACHAR)
4081    {    {
4082    int n = 0;    int n = 0;
4083    if (p[0] != except)    if (p[0] != except)
4084      {      {
4085      while(p[n+1] == p[0] + n + 1) n++;      while(p[n+1] == p[0] + n + 1) n++;
4086      n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);      n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4087      }      }
4088    p += n + 1;    p += n + 1;
4089    }    }
4090  return n8;  return n8;
4091  }  }
4092    
4093    
4094    
# Line 3624  Arguments: Line 4103  Arguments:
4103    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4104    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4105    options       the options word    options       the options word
4106    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4107    p             points to row of 32-bit values, terminated by NOTACHAR    p             points to row of 32-bit values, terminated by NOTACHAR
4108    
4109  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4110                  the pointer to extra data is updated                  the pointer to extra data is updated
4111  */  */
4112    
4113  static int  static int
4114  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
4115    int options, compile_data *cd, const pcre_uint32 *p)    int options, compile_data *cd, const pcre_uint32 *p)
4116  {  {
4117  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
# Line 3644  while (p[0] < NOTACHAR) Line 4123  while (p[0] < NOTACHAR)
4123    while (p[1] == p[0] + 1) p++;    while (p[1] == p[0] + 1) p++;
4124    n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,    n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
4125      (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);      (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
4126    p++;    p++;
4127    }    }
4128  return n8;  return n8;
4129  }  }
4130    
4131    
4132    
# Line 3662  to find out the amount of memory needed, Line 4141  to find out the amount of memory needed,
4141  phase. The value of lengthptr distinguishes the two phases.  phase. The value of lengthptr distinguishes the two phases.
4142    
4143  Arguments:  Arguments:
4144    optionsptr     pointer to the option bits    optionsptr        pointer to the option bits
4145    codeptr        points to the pointer to the current code point    codeptr           points to the pointer to the current code point
4146    ptrptr         points to the current pattern pointer    ptrptr            points to the current pattern pointer
4147    errorcodeptr   points to error code variable    errorcodeptr      points to error code variable
4148    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr      place to put the first required character
4149    reqcharptr     set to the last literal character required, else < 0    firstcharflagsptr place to put the first character flags, or a negative number
4150    bcptr          points to current branch chain    reqcharptr        place to put the last required character
4151    cond_depth     conditional nesting depth    reqcharflagsptr   place to put the last required character flags, or a negative number
4152    cd             contains pointers to tables etc.    bcptr             points to current branch chain
4153    lengthptr      NULL during the real compile phase    cond_depth        conditional nesting depth
4154                   points to length accumulator during pre-compile phase    cd                contains pointers to tables etc.
4155      lengthptr         NULL during the real compile phase
4156                        points to length accumulator during pre-compile phase
4157    
4158  Returns:         TRUE on success  Returns:            TRUE on success
4159                   FALSE, with *errorcodeptr set non-zero on error                      FALSE, with *errorcodeptr set non-zero on error
4160  */  */
4161    
4162  static BOOL  static BOOL
4163  compile_branch(int *optionsptr, pcre_uchar **codeptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
4164    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,    const pcre_uchar **ptrptr, int *errorcodeptr,
4165    pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,    pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
4166      pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
4167      branch_chain *bcptr, int cond_depth,
4168    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
4169  {  {
4170  int repeat_type, op_type;  int repeat_type, op_type;
4171  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
4172  int bravalue = 0;  int bravalue = 0;
4173  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
4174  pcre_int32 firstchar, reqchar;  pcre_uint32 firstchar, reqchar;
4175  pcre_int32 zeroreqchar, zerofirstchar;  pcre_int32 firstcharflags, reqcharflags;
4176    pcre_uint32 zeroreqchar, zerofirstchar;
4177    pcre_int32 zeroreqcharflags, zerofirstcharflags;
4178  pcre_int32 req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
4179  int options = *optionsptr;               /* May change dynamically */  int options = *optionsptr;               /* May change dynamically */
4180  int after_manual_callout = 0;  int after_manual_callout = 0;
# Line 3717  dynamically as we process the pattern. * Line 4202  dynamically as we process the pattern. *
4202  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4203  /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */  /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
4204  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
4205    #ifndef COMPILE_PCRE32
4206  pcre_uchar utf_chars[6];  pcre_uchar utf_chars[6];
4207    #endif
4208  #else  #else
4209  BOOL utf = FALSE;  BOOL utf = FALSE;
4210  #endif  #endif
4211    
4212  /* Helper variables for OP_XCLASS opcode (for characters > 255). We define  /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
4213  class_uchardata always so that it can be passed to add_to_class() always,  class_uchardata always so that it can be passed to add_to_class() always,
4214  though it will not be used in non-UTF 8-bit cases. This avoids having to supply  though it will not be used in non-UTF 8-bit cases. This avoids having to supply
4215  alternative calls for the different cases. */  alternative calls for the different cases. */
4216    
4217  pcre_uchar *class_uchardata;  pcre_uchar *class_uchardata;
# Line 3752  to take the zero repeat into account. Th Line 4239  to take the zero repeat into account. Th
4239  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
4240  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
4241    
4242  firstchar = reqchar = zerofirstchar = zeroreqchar = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
4243    firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
4244    
4245  /* The variable req_caseopt contains either the REQ_CASELESS value  /* The variable req_caseopt contains either the REQ_CASELESS value
4246  or zero, according to the current setting of the caseless flag. The  or zero, according to the current setting of the caseless flag. The
# Line 3778  for (;; ptr++) Line 4266  for (;; ptr++)
4266    int recno;    int recno;
4267    int refsign;    int refsign;
4268    int skipbytes;    int skipbytes;
4269    int subreqchar;    pcre_uint32 subreqchar, subfirstchar;
4270    int subfirstchar;    pcre_int32 subreqcharflags, subfirstcharflags;
4271    int terminator;    int terminator;
4272    int mclength;    unsigned int mclength;
4273    int tempbracount;    unsigned int tempbracount;
4274    int ec; // FIXMEchpe pcre_uint32    pcre_uint32 ec;
4275    pcre_uchar mcbuffer[8];    pcre_uchar mcbuffer[8];
4276    
4277    /* Get next character in the pattern */    /* Get next character in the pattern */
# Line 3793  for (;; ptr++) Line 4281  for (;; ptr++)
4281    /* If we are at the end of a nested substitution, revert to the outer level    /* If we are at the end of a nested substitution, revert to the outer level
4282    string. Nesting only happens one level deep. */    string. Nesting only happens one level deep. */
4283    
4284    if (c == 0 && nestptr != NULL)    if (c == CHAR_NULL && nestptr != NULL)
4285      {      {
4286      ptr = nestptr;      ptr = nestptr;
4287      nestptr = NULL;      nestptr = NULL;
# Line 3868  for (;; ptr++) Line 4356  for (;; ptr++)
4356    
4357    /* If in \Q...\E, check for the end; if not, we have a literal */    /* If in \Q...\E, check for the end; if not, we have a literal */
4358    
4359    if (inescq && c != 0)    if (inescq && c != CHAR_NULL)
4360      {      {
4361      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4362        {        {
# Line 3893  for (;; ptr++) Line 4381  for (;; ptr++)
4381        }        }
4382      }      }
4383    
   /* Fill in length of a previous callout, except when the next thing is  
   a quantifier. */  
   
4384    is_quantifier =    is_quantifier =
4385      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4386      (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));      (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4387    
4388    if (!is_quantifier && previous_callout != NULL &&    /* Fill in length of a previous callout, except when the next thing is a
4389      quantifier or when processing a property substitution string in UCP mode. */
4390    
4391      if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4392         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
4393      {      {
4394      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
# Line 3916  for (;; ptr++) Line 4404  for (;; ptr++)
4404      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
4405        {        {
4406        ptr++;        ptr++;
4407        while (*ptr != 0)        while (*ptr != CHAR_NULL)
4408          {          {
4409          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
4410          ptr++;          ptr++;
# Line 3924  for (;; ptr++) Line 4412  for (;; ptr++)
4412          if (utf) FORWARDCHAR(ptr);          if (utf) FORWARDCHAR(ptr);
4413  #endif  #endif
4414          }          }
4415        if (*ptr != 0) continue;        if (*ptr != CHAR_NULL) continue;
4416    
4417        /* Else fall through to handle end of string */        /* Else fall through to handle end of string */
4418        c = 0;        c = 0;
4419        }        }
4420      }      }
4421    
4422    /* No auto callout for quantifiers. */    /* No auto callout for quantifiers, or while processing property strings that
4423      are substituted for \w etc in UCP mode. */
4424    
4425    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4426      {      {
4427      previous_callout = code;      previous_callout = code;
4428      code = auto_callout(code, ptr, cd);      code = auto_callout(code, ptr, cd);
# Line 3946  for (;; ptr++) Line 4435  for (;; ptr++)
4435      case CHAR_VERTICAL_LINE:       /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
4436      case CHAR_RIGHT_PARENTHESIS:      case CHAR_RIGHT_PARENTHESIS:
4437      *firstcharptr = firstchar;      *firstcharptr = firstchar;
4438        *firstcharflagsptr = firstcharflags;
4439      *reqcharptr = reqchar;      *reqcharptr = reqchar;
4440        *reqcharflagsptr = reqcharflags;
4441      *codeptr = code;      *codeptr = code;
4442      *ptrptr = ptr;      *ptrptr = ptr;
4443      if (lengthptr != NULL)      if (lengthptr != NULL)
# Line 3970  for (;; ptr++) Line 4461  for (;; ptr++)
4461      previous = NULL;      previous = NULL;
4462      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
4463        {        {
4464        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;        if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4465        *code++ = OP_CIRCM;        *code++ = OP_CIRCM;
4466        }        }
4467      else *code++ = OP_CIRC;      else *code++ = OP_CIRC;
# Line 3985  for (;; ptr++) Line 4476  for (;; ptr++)
4476      repeats. The value of reqchar doesn't change either. */      repeats. The value of reqchar doesn't change either. */
4477    
4478      case CHAR_DOT:      case CHAR_DOT:
4479      if (firstchar == REQ_UNSET) firstchar = REQ_NONE;      if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4480      zerofirstchar = firstchar;      zerofirstchar = firstchar;
4481        zerofirstcharflags = firstcharflags;
4482      zeroreqchar = reqchar;      zeroreqchar = reqchar;
4483        zeroreqcharflags = reqcharflags;
4484      previous = code;      previous = code;
4485      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
4486      break;      break;
# Line 4061  for (;; ptr++) Line 4554  for (;; ptr++)
4554          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4555        {        {
4556        *code++ = negate_class? OP_ALLANY : OP_FAIL;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
4557        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;        if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4558        zerofirstchar = firstchar;        zerofirstchar = firstchar;
4559          zerofirstcharflags = firstcharflags;
4560        break;        break;
4561        }        }
4562    
# Line 4097  for (;; ptr++) Line 4591  for (;; ptr++)
4591      means that an initial ] is taken as a data character. At the start of the      means that an initial ] is taken as a data character. At the start of the
4592      loop, c contains the first byte of the character. */      loop, c contains the first byte of the character. */
4593    
4594      if (c != 0) do      if (c != CHAR_NULL) do
4595        {        {
4596        const pcre_uchar *oldptr;        const pcre_uchar *oldptr;
4597    
4598  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #ifdef SUPPORT_UTF
4599        if (utf && HAS_EXTRALEN(c))        if (utf && HAS_EXTRALEN(c))
4600          {                           /* Braces are required because the */          {                           /* Braces are required because the */
4601          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
# Line 4112  for (;; ptr++) Line 4606  for (;; ptr++)
4606        /* In the pre-compile phase, accumulate the length of any extra        /* In the pre-compile phase, accumulate the length of any extra
4607        data and reset the pointer. This is so that very large classes that        data and reset the pointer. This is so that very large classes that
4608        contain a zillion > 255 characters no longer overwrite the work space        contain a zillion > 255 characters no longer overwrite the work space
4609        (which is on the stack). We have to remember that there was XCLASS data,        (which is on the stack). We have to remember that there was XCLASS data,
4610        however. */        however. */
4611    
4612        if (lengthptr != NULL && class_uchardata > class_uchardata_base)        if (lengthptr != NULL && class_uchardata > class_uchardata_base)
# Line 4176  for (;; ptr++) Line 4670  for (;; ptr++)
4670          alpha. This relies on the fact that the class table starts with          alpha. This relies on the fact that the class table starts with
4671          alpha, lower, upper as the first 3 entries. */          alpha, lower, upper as the first 3 entries. */
4672    
4673          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
4674            posix_class = 0;            posix_class = 0;
4675    
4676          /* When PCRE_UCP is set, some of the POSIX classes are converted to          /* When PCRE_UCP is set, some of the POSIX classes are converted to
# Line 4253  for (;; ptr++) Line 4747  for (;; ptr++)
4747    
4748        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
4749          {          {
4750          escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options, TRUE);          escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options,
4751              TRUE);
4752          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
4753            if (escape == 0) c = ec;
         if (escape == 0)  
           c = ec;  
4754          else if (escape == ESC_b) c = CHAR_BS; /* \b is backspace in a class */          else if (escape == ESC_b) c = CHAR_BS; /* \b is backspace in a class */
4755          else if (escape == ESC_N)            /* \N is not supported in a class */          else if (escape == ESC_N)          /* \N is not supported in a class */
4756            {            {
4757            *errorcodeptr = ERR71;            *errorcodeptr = ERR71;
4758            goto FAILED;            goto FAILED;
# Line 4316  for (;; ptr++) Line 4808  for (;; ptr++)
4808              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
4809              continue;              continue;
4810    
4811              /* 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
4812              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
4813              class. Luckily, the value of CHAR_VT is 0x0b in both ASCII and              previously set by something earlier in the character class.
4814              EBCDIC, so we lazily just adjust the appropriate bit. */              Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
4815                we could just adjust the appropriate bit. From PCRE 8.34 we no
4816                longer treat \s and \S specially. */
4817    
4818              case ESC_s:              case ESC_s:
4819              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];  
4820              continue;              continue;
4821    
4822              case ESC_S:              case ESC_S:
4823              should_flip_negation = TRUE;              should_flip_negation = TRUE;
4824              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 */  
4825              continue;              continue;
4826    
4827              /* The rest apply in both UCP and non-UCP cases. */              /* The rest apply in both UCP and non-UCP cases. */
4828    
4829              case ESC_h:              case ESC_h:
4830              (void)add_list_to_class(classbits, &class_uchardata, options, cd,              (void)add_list_to_class(classbits, &class_uchardata, options, cd,
4831                PRIV(hspace_list), NOTACHAR);                PRIV(hspace_list), NOTACHAR);
4832              continue;              continue;
4833    
4834              case ESC_H:              case ESC_H:
4835              (void)add_not_list_to_class(classbits, &class_uchardata, options,              (void)add_not_list_to_class(classbits, &class_uchardata, options,
4836                cd, PRIV(hspace_list));                cd, PRIV(hspace_list));
4837              continue;              continue;
4838    
4839              case ESC_v:              case ESC_v:
4840              (void)add_list_to_class(classbits, &class_uchardata, options, cd,              (void)add_list_to_class(classbits, &class_uchardata, options, cd,
4841                PRIV(vspace_list), NOTACHAR);                PRIV(vspace_list), NOTACHAR);
4842              continue;              continue;
4843    
4844              case ESC_V:              case ESC_V:
4845              (void)add_not_list_to_class(classbits, &class_uchardata, options,              (void)add_not_list_to_class(classbits, &class_uchardata, options,
4846                cd, PRIV(vspace_list));                cd, PRIV(vspace_list));
4847              continue;              continue;
4848    
4849  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
# Line 4360  for (;; ptr++) Line 4851  for (;; ptr++)
4851              case ESC_P:              case ESC_P:
4852                {                {
4853                BOOL negated;                BOOL negated;
4854                int pdata;                unsigned int ptype = 0, pdata = 0;
4855                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                if (!get_ucp(&ptr, &negated, &ptype, &pdata, errorcodeptr))
4856                if (ptype < 0) goto FAILED;                  goto FAILED;
4857                *class_uchardata++ = ((escape == ESC_p) != negated)?                *class_uchardata++ = ((escape == ESC_p) != negated)?
4858                  XCL_PROP : XCL_NOTPROP;                  XCL_PROP : XCL_NOTPROP;
4859                *class_uchardata++ = ptype;                *class_uchardata++ = ptype;
# Line 4390  for (;; ptr++) Line 4881  for (;; ptr++)
4881    
4882          /* Fall through if the escape just defined a single character (c >= 0).          /* Fall through if the escape just defined a single character (c >= 0).
4883          This may be greater than 256. */          This may be greater than 256. */
4884    
4885          escape = 0;          escape = 0;
4886    
4887          }   /* End of backslash handling */          }   /* End of backslash handling */
# Line 4416  for (;; ptr++) Line 4907  for (;; ptr++)
4907    
4908        if (!inescq && ptr[1] == CHAR_MINUS)        if (!inescq && ptr[1] == CHAR_MINUS)
4909          {          {
4910          int d;          pcre_uint32 d;
4911          ptr += 2;          ptr += 2;
4912          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
4913    
# Line 4431  for (;; ptr++) Line 4922  for (;; ptr++)
4922            inescq = TRUE;            inescq = TRUE;
4923            break;            break;
4924            }            }
4925    
4926          /* Minus (hyphen) at the end of a class is treated as a literal, so put          /* Minus (hyphen) at the end of a class is treated as a literal, so put
4927          back the pointer and jump to handle the character that preceded it. */          back the pointer and jump to handle the character that preceded it. */
4928    
4929          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))          if (*ptr == CHAR_NULL || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
4930            {            {
4931            ptr = oldptr;            ptr = oldptr;
4932            goto CLASS_SINGLE_CHARACTER;            goto CLASS_SINGLE_CHARACTER;
4933            }            }
4934    
4935          /* Otherwise, we have a potential range; pick up the next character */          /* Otherwise, we have a potential range; pick up the next character */
4936    
4937  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
# Line 4487  for (;; ptr++) Line 4978  for (;; ptr++)
4978          /* We have found a character range, so single character optimizations          /* We have found a character range, so single character optimizations
4979          cannot be done anymore. Any value greater than 1 indicates that there          cannot be done anymore. Any value greater than 1 indicates that there
4980          is more than one character. */          is more than one character. */
4981    
4982          class_one_char = 2;          class_one_char = 2;
4983    
4984          /* Remember an explicit \r or \n, and add the range to the class. */          /* Remember an explicit \r or \n, and add the range to the class. */
4985    
4986          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
4987    
4988          class_has_8bitchar +=          class_has_8bitchar +=
4989            add_to_class(classbits, &class_uchardata, options, cd, c, d);            add_to_class(classbits, &class_uchardata, options, cd, c, d);
4990    
4991          continue;   /* Go get the next char in the class */          continue;   /* Go get the next char in the class */
4992          }          }
4993    
4994        /* Handle a single character - we can get here for a normal non-escape        /* Handle a single character - we can get here for a normal non-escape
4995        char, or after \ that introduces a single character or for an apparent        char, or after \ that introduces a single character or for an apparent
4996        range that isn't. Only the value 1 matters for class_one_char, so don't        range that isn't. Only the value 1 matters for class_one_char, so don't
4997        increase it if it is already 2 or more ... just in case there's a class        increase it if it is already 2 or more ... just in case there's a class
4998        with a zillion characters in it. */        with a zillion characters in it. */
4999    
5000        CLASS_SINGLE_CHARACTER:        CLASS_SINGLE_CHARACTER:
# Line 4522  for (;; ptr++) Line 5013  for (;; ptr++)
5013          {          {
5014          ptr++;          ptr++;
5015          zeroreqchar = reqchar;          zeroreqchar = reqchar;
5016            zeroreqcharflags = reqcharflags;
5017    
5018          if (negate_class)          if (negate_class)
5019            {            {
5020  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
             // FIXMEchpe pcreuint32?  
5021            int d;            int d;
5022  #endif  #endif
5023            if (firstchar == REQ_UNSET) firstchar = REQ_NONE;            if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5024            zerofirstchar = firstchar;            zerofirstchar = firstchar;
5025              zerofirstcharflags = firstcharflags;
5026    
5027            /* For caseless UTF-8 mode when UCP support is available, check            /* For caseless UTF-8 mode when UCP support is available, check
5028            whether this character has more than one other case. If so, generate            whether this character has more than one other case. If so, generate
5029            a special OP_NOTPROP item instead of OP_NOTI. */            a special OP_NOTPROP item instead of OP_NOTI. */
5030    
5031  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
5032            if (utf && (options & PCRE_CASELESS) != 0 &&            if (utf && (options & PCRE_CASELESS) != 0 &&
5033                (d = UCD_CASESET(c)) != 0)                (d = UCD_CASESET(c)) != 0)
5034              {              {
5035              *code++ = OP_NOTPROP;              *code++ = OP_NOTPROP;
5036              *code++ = PT_CLIST;              *code++ = PT_CLIST;
5037              *code++ = d;              *code++ = d;
5038              }              }
5039            else            else
5040  #endif  #endif
5041            /* Char has only one other case, or UCP not available */            /* Char has only one other case, or UCP not available */
5042    
# Line 4557  for (;; ptr++) Line 5049  for (;; ptr++)
5049  #endif  #endif
5050                *code++ = c;                *code++ = c;
5051              }              }
5052    
5053            /* We are finished with this character class */            /* We are finished with this character class */
5054    
5055            goto END_CLASS;            goto END_CLASS;
5056            }            }
5057    
# Line 4577  for (;; ptr++) Line 5069  for (;; ptr++)
5069            }            }
5070          goto ONE_CHAR;          goto ONE_CHAR;
5071          }       /* End of 1-char optimization */          }       /* End of 1-char optimization */
5072    
5073        /* There is more than one character in the class, or an XCLASS item        /* There is more than one character in the class, or an XCLASS item
5074        has been generated. Add this character to the class. */        has been generated. Add this character to the class. */
5075    
5076        class_has_8bitchar +=        class_has_8bitchar +=
5077          add_to_class(classbits, &class_uchardata, options, cd, c, c);          add_to_class(classbits, &class_uchardata, options, cd, c, c);
5078        }        }
5079    
# Line 4589  for (;; ptr++) Line 5081  for (;; ptr++)
5081      If we are at the end of an internal nested string, revert to the outer      If we are at the end of an internal nested string, revert to the outer
5082      string. */      string. */
5083    
5084      while (((c = *(++ptr)) != 0 ||      while (((c = *(++ptr)) != CHAR_NULL ||
5085             (nestptr != NULL &&             (nestptr != NULL &&
5086               (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&               (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != CHAR_NULL)) &&
5087             (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));             (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
5088    
5089      /* Check for missing terminating ']' */      /* Check for missing terminating ']' */
5090    
5091      if (c == 0)      if (c == CHAR_NULL)
5092        {        {
5093        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
5094        goto FAILED;        goto FAILED;
5095        }        }
5096    
5097      /* We will need an XCLASS if data has been placed in class_uchardata. In      /* We will need an XCLASS if data has been placed in class_uchardata. In
5098      the second phase this is a sufficient test. However, in the pre-compile      the second phase this is a sufficient test. However, in the pre-compile
5099      phase, class_uchardata gets emptied to prevent workspace overflow, so it      phase, class_uchardata gets emptied to prevent workspace overflow, so it
5100      only if the very last character in the class needs XCLASS will it contain      only if the very last character in the class needs XCLASS will it contain
5101      anything at this point. For this reason, xclass gets set TRUE above when      anything at this point. For this reason, xclass gets set TRUE above when
5102      uchar_classdata is emptied, and that's why this code is the way it is here      uchar_classdata is emptied, and that's why this code is the way it is here
5103      instead of just doing a test on class_uchardata below. */      instead of just doing a test on class_uchardata below. */
5104    
5105  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5106      if (class_uchardata > class_uchardata_base) xclass = TRUE;      if (class_uchardata > class_uchardata_base) xclass = TRUE;
5107  #endif  #endif
# Line 4618  for (;; ptr++) Line 5110  for (;; ptr++)
5110      setting, whatever the repeat count. Any reqchar setting must remain      setting, whatever the repeat count. Any reqchar setting must remain
5111      unchanged after any kind of repeat. */      unchanged after any kind of repeat. */
5112    
5113      if (firstchar == REQ_UNSET) firstchar = REQ_NONE;      if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5114      zerofirstchar = firstchar;      zerofirstchar = firstchar;
5115        zerofirstcharflags = firstcharflags;
5116      zeroreqchar = reqchar;      zeroreqchar = reqchar;
5117        zeroreqcharflags = reqcharflags;
5118    
5119      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
5120      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
# Line 4676  for (;; ptr++) Line 5170  for (;; ptr++)
5170        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
5171        }        }
5172      code += 32 / sizeof(pcre_uchar);      code += 32 / sizeof(pcre_uchar);
5173    
5174      END_CLASS:      END_CLASS:
5175      break;      break;
5176    
# Line 4715  for (;; ptr++) Line 5209  for (;; ptr++)
5209      if (repeat_min == 0)      if (repeat_min == 0)
5210        {        {
5211        firstchar = zerofirstchar;    /* Adjust for zero repeat */        firstchar = zerofirstchar;    /* Adjust for zero repeat */
5212          firstcharflags = zerofirstcharflags;
5213        reqchar = zeroreqchar;        /* Ditto */        reqchar = zeroreqchar;        /* Ditto */
5214          reqcharflags = zeroreqcharflags;
5215        }        }
5216    
5217      /* Remember whether this is a variable length repeat */      /* Remember whether this is a variable length repeat */
# Line 4818  for (;; ptr++) Line 5314  for (;; ptr++)
5314          {          {
5315          c = code[-1];          c = code[-1];
5316          if (*previous <= OP_CHARI && repeat_min > 1)          if (*previous <= OP_CHARI && repeat_min > 1)
5317            reqchar = c | req_caseopt | cd->req_varyopt;            {
5318          }            reqchar = c;
5319              reqcharflags = req_caseopt | cd->req_varyopt;
5320        /* 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;  
5321          }          }
5322    
5323        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
# Line 4851  for (;; ptr++) Line 5337  for (;; ptr++)
5337        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
5338        c = *previous;        c = *previous;
5339    
       if (!possessive_quantifier &&  
           repeat_max < 0 &&  
           check_auto_possessive(previous, utf, ptr + 1, options, cd))  
         {  
         repeat_type = 0;    /* Force greedy */