/[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 1067 by chpe, Tue Oct 16 15:54:22 2012 UTC revision 1381 by zherczeg, Wed Oct 16 06:23:00 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    /* This table is used when converting repeating opcodes into possessified
781    versions as a result of an explicit possessive quantifier such as ++. A zero
782    value means there is no possessified version - in those cases the item in
783    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
784    because all relevant opcodes are less than that. */
785    
786    static const pcre_uint8 opcode_possessify[] = {
787      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
788      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
789    
790      0,                       /* NOTI */
791      OP_POSSTAR, 0,           /* STAR, MINSTAR */
792      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
793      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
794      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
795      0,                       /* EXACT */
796      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
797    
798      OP_POSSTARI, 0,          /* STARI, MINSTARI */
799      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
800      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
801      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
802      0,                       /* EXACTI */
803      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
804    
805      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
806      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
807      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
808      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
809      0,                       /* NOTEXACT */
810      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
811    
812      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
813      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
814      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
815      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
816      0,                       /* NOTEXACTI */
817      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
818    
819      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
820      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
821      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
822      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
823      0,                       /* TYPEEXACT */
824      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
825    
826      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
827      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
828      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
829      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
830      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
831    
832      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
833      0, 0,                    /* REF, REFI */
834      0, 0,                    /* DNREF, DNREFI */
835      0, 0                     /* RECURSE, CALLOUT */
836    };
837    
838    
839    
840  /*************************************************  /*************************************************
# Line 665  find_error_text(int n) Line 856  find_error_text(int n)
856  const char *s = error_texts;  const char *s = error_texts;
857  for (; n > 0; n--)  for (; n > 0; n--)
858    {    {
859    while (*s++ != 0) {};    while (*s++ != CHAR_NULL) {};
860    if (*s == 0) return "Error text not found (please report)";    if (*s == CHAR_NULL) return "Error text not found (please report)";
861    }    }
862  return s;  return s;
863  }  }
864    
865    
866    
867  /*************************************************  /*************************************************
868  *           Expand the workspace                 *  *           Expand the workspace                 *
869  *************************************************/  *************************************************/
# Line 749  return (*p == CHAR_RIGHT_CURLY_BRACKET); Line 941  return (*p == CHAR_RIGHT_CURLY_BRACKET);
941  *************************************************/  *************************************************/
942    
943  /* 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
944  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
945  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.
946  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
947  be returned in chptr.  chptr. On entry, ptr is pointing at the \. On exit, it is on the final
948  On entry,ptr is pointing at the \. On exit, it is on the final character of the  character of the escape sequence.
 escape sequence.  
949    
950  Arguments:  Arguments:
951    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
952    chptr          points to the data character    chptr          points to a returned data character
953    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
954    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
955    options        the options bits    options        the options bits
# Line 771  Returns:         zero => a data characte Line 962  Returns:         zero => a data characte
962  */  */
963    
964  static int  static int
965  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
966    int bracount, int options, BOOL isclass)    int bracount, int options, BOOL isclass)
967  {  {
968  /* PCRE_UTF16 has the same value as PCRE_UTF8. */  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
# Line 786  ptr--;                            /* Set Line 977  ptr--;                            /* Set
977    
978  /* 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. */
979    
980  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
981    
982  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
983  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 986  Otherwise further processing may be requ
986  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
987  /* Not alphanumeric */  /* Not alphanumeric */
988  else if (c < CHAR_0 || c > CHAR_z) {}  else if (c < CHAR_0 || c > CHAR_z) {}
989  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)
990      { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
991    
992  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
993  /* Not alphanumeric */  /* Not alphanumeric */
# Line 845  else Line 1037  else
1037            }            }
1038    
1039  #if defined COMPILE_PCRE8  #if defined COMPILE_PCRE8
1040          if (c > (utf ? 0x10ffff : 0xff))          if (c > (utf ? 0x10ffffU : 0xffU))
1041  #elif defined COMPILE_PCRE16  #elif defined COMPILE_PCRE16
1042          if (c > (utf ? 0x10ffff : 0xffff))          if (c > (utf ? 0x10ffffU : 0xffffU))
1043  #elif defined COMPILE_PCRE32  #elif defined COMPILE_PCRE32
1044          if (utf && c > 0x10ffff)          if (utf && c > 0x10ffffU)
1045  #endif  #endif
1046            {            {
1047            *errorcodeptr = ERR76;            *errorcodeptr = ERR76;
# Line 896  else Line 1088  else
1088      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1089        {        {
1090        const pcre_uchar *p;        const pcre_uchar *p;
1091        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++)
1092          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1093        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)        if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1094          {          {
1095          escape = ESC_k;          escape = ESC_k;
1096          break;          break;
# Line 961  else Line 1153  else
1153      break;      break;
1154    
1155      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1156      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
1157      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1158        recommended to avoid the ambiguities in the old syntax.
1159    
1160      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
1161      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
1162      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
1163      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
1164      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
1165      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
1166      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1167    
1168        Inside a character class, \ followed by a digit is always either a literal
1169        8 or 9 or an octal number. */
1170    
1171      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:
1172      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 1193  else
1193          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1194          break;          break;
1195          }          }
1196        if (s < 10 || s <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1197          {          {
1198          escape = -s;          escape = -s;
1199          break;          break;
# Line 1005  else Line 1201  else
1201        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1202        }        }
1203    
1204      /* 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
1205      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
1206      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
1207        changed so as not to insert the binary zero. */
1208    
1209      if ((c = *ptr) >= CHAR_8)      if ((c = *ptr) >= CHAR_8) break;
1210        {  
1211        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1212    
1213      /* \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
1214      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 1225  else
1225  #endif  #endif
1226      break;      break;
1227    
1228      /* \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
1229      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}. */
1230      If not, { is treated as a data character. */  
1231        case CHAR_o:
1232        if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1233          {
1234          ptr += 2;
1235          c = 0;
1236          overflow = FALSE;
1237          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1238            {
1239            register pcre_uint32 cc = *ptr++;
1240            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1241    #ifdef COMPILE_PCRE32
1242            if (c >= 0x20000000l) { overflow = TRUE; break; }
1243    #endif
1244            c = (c << 3) + cc - CHAR_0 ;
1245    #if defined COMPILE_PCRE8
1246            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1247    #elif defined COMPILE_PCRE16
1248            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1249    #elif defined COMPILE_PCRE32
1250            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1251    #endif
1252            }
1253          if (overflow)
1254            {
1255            while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1256            *errorcodeptr = ERR34;
1257            }
1258          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1259            {
1260            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1261            }
1262          else *errorcodeptr = ERR80;
1263          }
1264        break;
1265    
1266        /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1267        numbers. Otherwise it is a lowercase x letter. */
1268    
1269      case CHAR_x:      case CHAR_x:
1270      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1271        {        {
       /* In JavaScript, \x must be followed by two hexadecimal numbers.  
       Otherwise it is a lowercase x letter. */  
1272        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1273          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1274          {          {
# Line 1056  else Line 1285  else
1285  #endif  #endif
1286            }            }
1287          }          }
1288        break;        }    /* End JavaScript handling */
       }  
1289    
1290      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1291        {      greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1292        const pcre_uchar *pt = ptr + 2;      digits. If not, { used to be treated as a data character. However, Perl
1293        seems to read hex digits up to the first non-such, and ignore the rest, so
1294        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1295        now gives an error. */
1296    
1297        c = 0;      else
1298        overflow = FALSE;        {
1299        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1300          {          {
1301          register pcre_uint32 cc = *pt++;          ptr += 2;
1302          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          c = 0;
1303            overflow = FALSE;
1304            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1305              {
1306              register pcre_uint32 cc = *ptr++;
1307              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1308    
1309  #ifdef COMPILE_PCRE32  #ifdef COMPILE_PCRE32
1310          if (c >= 0x10000000l) { overflow = TRUE; break; }            if (c >= 0x10000000l) { overflow = TRUE; break; }
1311  #endif  #endif
1312    
1313  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1314          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1315          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1316  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1317          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 */
1318          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1319  #endif  #endif
1320    
1321  #if defined COMPILE_PCRE8  #if defined COMPILE_PCRE8
1322          if (c > (utf ? 0x10ffff : 0xff)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1323  #elif defined COMPILE_PCRE16  #elif defined COMPILE_PCRE16
1324          if (c > (utf ? 0x10ffff : 0xffff)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1325  #elif defined COMPILE_PCRE32  #elif defined COMPILE_PCRE32
1326          if (utf && c > 0x10ffff) { overflow = TRUE; break; }            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1327  #endif  #endif
1328          }            }
1329    
1330        if (overflow)          if (overflow)
1331          {            {
1332          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1333          *errorcodeptr = ERR34;            *errorcodeptr = ERR34;
1334          }            }
1335    
1336        if (*pt == CHAR_RIGHT_CURLY_BRACKET)          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1337          {            {
1338          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1339          ptr = pt;            }
         break;  
         }  
1340    
1341        /* 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.
1342        recognize this construct; fall through to the normal \x handling. */          We used just to recognize this construct and fall through to the normal
1343        }          \x handling, but nowadays Perl gives an error, which seems much more
1344            sensible, so we do too. */
1345    
1346      /* Read just a single-byte hex-defined char */          else *errorcodeptr = ERR79;
1347            }   /* End of \x{} processing */
1348    
1349      c = 0;        /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1350      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)  
1351        {        else
1352        pcre_uint32 cc;                          /* Some compilers don't like */          {
1353        cc = *(++ptr);                           /* ++ in initializers */          c = 0;
1354            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1355              {
1356              pcre_uint32 cc;                          /* Some compilers don't like */
1357              cc = *(++ptr);                           /* ++ in initializers */
1358  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1359        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1360        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1361  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1362        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */            if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1363        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1364  #endif  #endif
1365        }            }
1366            }     /* End of \xdd handling */
1367          }       /* End of Perl-style \x handling */
1368      break;      break;
1369    
1370      /* 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 1374  else
1374    
1375      case CHAR_c:      case CHAR_c:
1376      c = *(++ptr);      c = *(++ptr);
1377      if (c == 0)      if (c == CHAR_NULL)
1378        {        {
1379        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1380        break;        break;
# Line 1188  if ((options & PCRE_UCP) != 0 && escape Line 1430  if ((options & PCRE_UCP) != 0 && escape
1430  return escape;  return escape;
1431  }  }
1432    
1433    
1434    
1435  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
1436  /*************************************************  /*************************************************
1437  *               Handle \P and \p                 *  *               Handle \P and \p                 *
# Line 1201  escape sequence. Line 1445  escape sequence.
1445  Argument:  Argument:
1446    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1447    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
1448    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
1449      pdataptr       points to an unsigned int that is set to the detailed property value
1450    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1451    
1452  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
1453  */  */
1454    
1455  static int  static BOOL
1456  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1457      unsigned int *pdataptr, int *errorcodeptr)
1458  {  {
1459  pcre_uchar c;  pcre_uchar c;
1460  int i, bot, top;  int i, bot, top;
# Line 1216  const pcre_uchar *ptr = *ptrptr; Line 1462  const pcre_uchar *ptr = *ptrptr;
1462  pcre_uchar name[32];  pcre_uchar name[32];
1463    
1464  c = *(++ptr);  c = *(++ptr);
1465  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1466    
1467  *negptr = FALSE;  *negptr = FALSE;
1468    
# Line 1233  if (c == CHAR_LEFT_CURLY_BRACKET) Line 1479  if (c == CHAR_LEFT_CURLY_BRACKET)
1479    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1480      {      {
1481      c = *(++ptr);      c = *(++ptr);
1482      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1483      if (c == CHAR_RIGHT_CURLY_BRACKET) break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1484      name[i] = c;      name[i] = c;
1485      }      }
# Line 1263  while (bot < top) Line 1509  while (bot < top)
1509    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);
1510    if (r == 0)    if (r == 0)
1511      {      {
1512      *dptr = PRIV(utt)[i].value;      *ptypeptr = PRIV(utt)[i].type;
1513      return PRIV(utt)[i].type;      *pdataptr = PRIV(utt)[i].value;
1514        return TRUE;
1515      }      }
1516    if (r > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1517    }    }
1518    
1519  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1520  *ptrptr = ptr;  *ptrptr = ptr;
1521  return -1;  return FALSE;
1522    
1523  ERROR_RETURN:  ERROR_RETURN:
1524  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1525  *ptrptr = ptr;  *ptrptr = ptr;
1526  return -1;  return FALSE;
1527  }  }
1528  #endif  #endif
1529    
1530    
1531    
   
1532  /*************************************************  /*************************************************
1533  *         Read repeat counts                     *  *         Read repeat counts                     *
1534  *************************************************/  *************************************************/
# Line 1351  return p; Line 1597  return p;
1597    
1598    
1599  /*************************************************  /*************************************************
 *  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;  
 }  
   
   
   
   
 /*************************************************  
1600  *      Find first significant op code            *  *      Find first significant op code            *
1601  *************************************************/  *************************************************/
1602    
# Line 1684  for (;;) Line 1635  for (;;)
1635    
1636      case OP_CALLOUT:      case OP_CALLOUT:
1637      case OP_CREF:      case OP_CREF:
1638      case OP_NCREF:      case OP_DNCREF:
1639      case OP_RREF:      case OP_RREF:
1640      case OP_NRREF:      case OP_DNRREF:
1641      case OP_DEF:      case OP_DEF:
1642      code += PRIV(OP_lengths)[*code];      code += PRIV(OP_lengths)[*code];
1643      break;      break;
# Line 1700  for (;;) Line 1651  for (;;)
1651    
1652    
1653    
   
1654  /*************************************************  /*************************************************
1655  *        Find the fixed length of a branch       *  *        Find the fixed length of a branch       *
1656  *************************************************/  *************************************************/
# Line 1744  for (;;) Line 1694  for (;;)
1694    {    {
1695    int d;    int d;
1696    pcre_uchar *ce, *cs;    pcre_uchar *ce, *cs;
1697    register int op = *cc;    register pcre_uchar op = *cc;
1698    
1699    switch (op)    switch (op)
1700      {      {
# Line 1824  for (;;) Line 1774  for (;;)
1774      case OP_COMMIT:      case OP_COMMIT:
1775      case OP_CREF:      case OP_CREF:
1776      case OP_DEF:      case OP_DEF:
1777        case OP_DNCREF:
1778        case OP_DNRREF:
1779      case OP_DOLL:      case OP_DOLL:
1780      case OP_DOLLM:      case OP_DOLLM:
1781      case OP_EOD:      case OP_EOD:
1782      case OP_EODN:      case OP_EODN:
1783      case OP_FAIL:      case OP_FAIL:
     case OP_NCREF:  
     case OP_NRREF:  
1784      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1785      case OP_PRUNE:      case OP_PRUNE:
1786      case OP_REVERSE:      case OP_REVERSE:
# Line 1852  for (;;) Line 1802  for (;;)
1802      case OP_NOTI:      case OP_NOTI:
1803      branchlength++;      branchlength++;
1804      cc += 2;      cc += 2;
1805  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #ifdef SUPPORT_UTF
1806      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1807  #endif  #endif
1808      break;      break;
# Line 1864  for (;;) Line 1814  for (;;)
1814      case OP_EXACTI:      case OP_EXACTI:
1815      case OP_NOTEXACT:      case OP_NOTEXACT:
1816      case OP_NOTEXACTI:      case OP_NOTEXACTI:
1817      branchlength += GET2(cc,1);      branchlength += (int)GET2(cc,1);
1818      cc += 2 + IMM2_SIZE;      cc += 2 + IMM2_SIZE;
1819  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #ifdef SUPPORT_UTF
1820      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1821  #endif  #endif
1822      break;      break;
1823    
1824      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1825      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1826      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)
1827        cc += 2;        cc += 2;
1828      cc += 1 + IMM2_SIZE + 1;      cc += 1 + IMM2_SIZE + 1;
1829      break;      break;
# Line 1909  for (;;) Line 1859  for (;;)
1859    
1860      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1861    
 #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  
   
1862      case OP_CLASS:      case OP_CLASS:
1863      case OP_NCLASS:      case OP_NCLASS:
1864    #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1865        case OP_XCLASS:
1866        /* The original code caused an unsigned overflow in 64 bit systems,
1867        so now we use a conditional statement. */
1868        if (op == OP_XCLASS)
1869          cc += GET(cc, 1);
1870        else
1871          cc += PRIV(OP_lengths)[OP_CLASS];
1872    #else
1873      cc += PRIV(OP_lengths)[OP_CLASS];      cc += PRIV(OP_lengths)[OP_CLASS];
1874    #endif
1875    
1876      switch (*cc)      switch (*cc)
1877        {        {
       case OP_CRPLUS:  
       case OP_CRMINPLUS:  
1878        case OP_CRSTAR:        case OP_CRSTAR:
1879        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1880          case OP_CRPLUS:
1881          case OP_CRMINPLUS:
1882        case OP_CRQUERY:        case OP_CRQUERY:
1883        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1884          case OP_CRPOSSTAR:
1885          case OP_CRPOSPLUS:
1886          case OP_CRPOSQUERY:
1887        return -1;        return -1;
1888    
1889        case OP_CRRANGE:        case OP_CRRANGE:
1890        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1891          case OP_CRPOSRANGE:
1892        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1893        branchlength += GET2(cc,1);        branchlength += (int)GET2(cc,1);
1894        cc += 1 + 2 * IMM2_SIZE;        cc += 1 + 2 * IMM2_SIZE;
1895        break;        break;
1896    
# Line 1999  for (;;) Line 1957  for (;;)
1957      case OP_QUERYI:      case OP_QUERYI:
1958      case OP_REF:      case OP_REF:
1959      case OP_REFI:      case OP_REFI:
1960        case OP_DNREF:
1961        case OP_DNREFI:
1962      case OP_SBRA:      case OP_SBRA:
1963      case OP_SBRAPOS:      case OP_SBRAPOS:
1964      case OP_SCBRA:      case OP_SCBRA:
# Line 2035  for (;;) Line 1995  for (;;)
1995    
1996    
1997    
   
1998  /*************************************************  /*************************************************
1999  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
2000  *************************************************/  *************************************************/
# Line 2059  PRIV(find_bracket)(const pcre_uchar *cod Line 2018  PRIV(find_bracket)(const pcre_uchar *cod
2018  {  {
2019  for (;;)  for (;;)
2020    {    {
2021    register int c = *code;    register pcre_uchar c = *code;
2022    
2023    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2024    
# Line 2082  for (;;) Line 2041  for (;;)
2041    else if (c == OP_CBRA || c == OP_SCBRA ||    else if (c == OP_CBRA || c == OP_SCBRA ||
2042             c == OP_CBRAPOS || c == OP_SCBRAPOS)             c == OP_CBRAPOS || c == OP_SCBRAPOS)
2043      {      {
2044      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
2045      if (n == number) return (pcre_uchar *)code;      if (n == number) return (pcre_uchar *)code;
2046      code += PRIV(OP_lengths)[c];      code += PRIV(OP_lengths)[c];
2047      }      }
# Line 2112  for (;;) Line 2071  for (;;)
2071        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2072        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2073        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2074        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)
2075          code += 2;          code += 2;
2076        break;        break;
2077    
2078        case OP_MARK:        case OP_MARK:
2079        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2080        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2081        case OP_THEN_ARG:        case OP_THEN_ARG:
2082        code += code[1];        code += code[1];
2083        break;        break;
# Line 2197  find_recurse(const pcre_uchar *code, BOO Line 2153  find_recurse(const pcre_uchar *code, BOO
2153  {  {
2154  for (;;)  for (;;)
2155    {    {
2156    register int c = *code;    register pcre_uchar c = *code;
2157    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2158    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2159    
# Line 2232  for (;;) Line 2188  for (;;)
2188        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2189        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2190        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2191        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)
2192          code += 2;          code += 2;
2193        break;        break;
2194    
2195        case OP_MARK:        case OP_MARK:
2196        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2197        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2198        case OP_THEN_ARG:        case OP_THEN_ARG:
2199        code += code[1];        code += code[1];
2200        break;        break;
# Line 2343  Arguments: Line 2296  Arguments:
2296    endcode     points to where to stop    endcode     points to where to stop
2297    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2298    cd          contains pointers to tables etc.    cd          contains pointers to tables etc.
2299      recurses    chain of recurse_check to catch mutual recursion
2300    
2301  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2302  */  */
2303    
2304    typedef struct recurse_check {
2305      struct recurse_check *prev;
2306      const pcre_uchar *group;
2307    } recurse_check;
2308    
2309  static BOOL  static BOOL
2310  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2311    BOOL utf, compile_data *cd)    BOOL utf, compile_data *cd, recurse_check *recurses)
2312  {  {
2313  register int c;  register pcre_uchar c;
2314    recurse_check this_recurse;
2315    
2316  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2317       code < endcode;       code < endcode;
2318       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 2340  for (code = first_significant_code(code
2340    
2341    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2342      {      {
2343      const pcre_uchar *scode;      const pcre_uchar *scode = cd->start_code + GET(code, 1);
2344      BOOL empty_branch;      BOOL empty_branch;
2345    
2346      /* Test for forward reference */      /* Test for forward reference or uncompleted reference. This is disabled
2347        when called to scan a completed pattern by setting cd->start_workspace to
2348        NULL. */
2349    
2350        if (cd->start_workspace != NULL)
2351          {
2352          const pcre_uchar *tcode;
2353          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2354            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2355          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2356          }
2357    
2358        /* If we are scanning a completed pattern, there are no forward references
2359        and all groups are complete. We need to detect whether this is a recursive
2360        call, as otherwise there will be an infinite loop. If it is a recursion,
2361        just skip over it. Simple recursions are easily detected. For mutual
2362        recursions we keep a chain on the stack. */
2363    
2364      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)      else
2365        if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;        {
2366          recurse_check *r = recurses;
2367          const pcre_uchar *endgroup = scode;
2368    
2369      /* Not a forward reference, test for completed backward reference */        do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2370          if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2371    
2372      empty_branch = FALSE;        for (r = recurses; r != NULL; r = r->prev)
2373      scode = cd->start_code + GET(code, 1);          if (r->group == scode) break;
2374      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */        if (r != NULL) continue;   /* Mutual recursion */
2375          }
2376    
2377      /* Completed backwards reference */      /* Completed reference; scan the referenced group, remembering it on the
2378        stack chain to detect mutual recursions. */
2379    
2380        empty_branch = FALSE;
2381        this_recurse.prev = recurses;
2382        this_recurse.group = scode;
2383    
2384      do      do
2385        {        {
2386        if (could_be_empty_branch(scode, endcode, utf, cd))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2387          {          {
2388          empty_branch = TRUE;          empty_branch = TRUE;
2389          break;          break;
# Line 2453  for (code = first_significant_code(code Line 2439  for (code = first_significant_code(code
2439        empty_branch = FALSE;        empty_branch = FALSE;
2440        do        do
2441          {          {
2442          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2443            empty_branch = TRUE;            empty_branch = TRUE;
2444          code += GET(code, 1);          code += GET(code, 1);
2445          }          }
# Line 2495  for (code = first_significant_code(code Line 2481  for (code = first_significant_code(code
2481        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2482        case OP_CRQUERY:        case OP_CRQUERY:
2483        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2484          case OP_CRPOSSTAR:
2485          case OP_CRPOSQUERY:
2486        break;        break;
2487    
2488        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2489        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2490        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2491          case OP_CRPOSPLUS:
2492        return FALSE;        return FALSE;
2493    
2494        case OP_CRRANGE:        case OP_CRRANGE:
2495        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2496          case OP_CRPOSRANGE:
2497        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2498        break;        break;
2499        }        }
# Line 2511  for (code = first_significant_code(code Line 2501  for (code = first_significant_code(code
2501    
2502      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2503    
2504        case OP_ANY:
2505        case OP_ALLANY:
2506        case OP_ANYBYTE:
2507    
2508      case OP_PROP:      case OP_PROP:
2509      case OP_NOTPROP:      case OP_NOTPROP:
2510        case OP_ANYNL:
2511    
2512        case OP_NOT_HSPACE:
2513        case OP_HSPACE:
2514        case OP_NOT_VSPACE:
2515        case OP_VSPACE:
2516      case OP_EXTUNI:      case OP_EXTUNI:
2517    
2518      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2519      case OP_DIGIT:      case OP_DIGIT:
2520      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2521      case OP_WHITESPACE:      case OP_WHITESPACE:
2522      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2523      case OP_WORDCHAR:      case OP_WORDCHAR:
2524      case OP_ANY:  
     case OP_ALLANY:  
     case OP_ANYBYTE:  
2525      case OP_CHAR:      case OP_CHAR:
2526      case OP_CHARI:      case OP_CHARI:
2527      case OP_NOT:      case OP_NOT:
2528      case OP_NOTI:      case OP_NOTI:
2529    
2530      case OP_PLUS:      case OP_PLUS:
2531        case OP_PLUSI:
2532      case OP_MINPLUS:      case OP_MINPLUS:
2533      case OP_POSPLUS:      case OP_MINPLUSI:
2534      case OP_EXACT:  
2535      case OP_NOTPLUS:      case OP_NOTPLUS:
2536        case OP_NOTPLUSI:
2537      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2538        case OP_NOTMINPLUSI:
2539    
2540        case OP_POSPLUS:
2541        case OP_POSPLUSI:
2542      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2543        case OP_NOTPOSPLUSI:
2544    
2545        case OP_EXACT:
2546        case OP_EXACTI:
2547      case OP_NOTEXACT:      case OP_NOTEXACT:
2548        case OP_NOTEXACTI:
2549    
2550      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2551      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2552      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2553      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2554    
2555      return FALSE;      return FALSE;
2556    
2557      /* 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 2571  for (code = first_significant_code(code
2571      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2572      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2573      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2574      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)
2575        code += 2;        code += 2;
2576      break;      break;
2577    
# Line 2572  for (code = first_significant_code(code Line 2585  for (code = first_significant_code(code
2585      return TRUE;      return TRUE;
2586    
2587      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2588      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2589        followed by a multibyte character. */
2590    
2591  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2592      case OP_STAR:      case OP_STAR:
2593      case OP_STARI:      case OP_STARI:
2594        case OP_NOTSTAR:
2595        case OP_NOTSTARI:
2596    
2597      case OP_MINSTAR:      case OP_MINSTAR:
2598      case OP_MINSTARI:      case OP_MINSTARI:
2599        case OP_NOTMINSTAR:
2600        case OP_NOTMINSTARI:
2601    
2602      case OP_POSSTAR:      case OP_POSSTAR:
2603      case OP_POSSTARI:      case OP_POSSTARI:
2604        case OP_NOTPOSSTAR:
2605        case OP_NOTPOSSTARI:
2606    
2607      case OP_QUERY:      case OP_QUERY:
2608      case OP_QUERYI:      case OP_QUERYI:
2609        case OP_NOTQUERY:
2610        case OP_NOTQUERYI:
2611    
2612      case OP_MINQUERY:      case OP_MINQUERY:
2613      case OP_MINQUERYI:      case OP_MINQUERYI:
2614        case OP_NOTMINQUERY:
2615        case OP_NOTMINQUERYI:
2616    
2617      case OP_POSQUERY:      case OP_POSQUERY:
2618      case OP_POSQUERYI:      case OP_POSQUERYI:
2619        case OP_NOTPOSQUERY:
2620        case OP_NOTPOSQUERYI:
2621    
2622      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2623      break;      break;
2624    
2625      case OP_UPTO:      case OP_UPTO:
2626      case OP_UPTOI:      case OP_UPTOI:
2627        case OP_NOTUPTO:
2628        case OP_NOTUPTOI:
2629    
2630      case OP_MINUPTO:      case OP_MINUPTO:
2631      case OP_MINUPTOI:      case OP_MINUPTOI:
2632        case OP_NOTMINUPTO:
2633        case OP_NOTMINUPTOI:
2634    
2635      case OP_POSUPTO:      case OP_POSUPTO:
2636      case OP_POSUPTOI:      case OP_POSUPTOI:
2637        case OP_NOTPOSUPTO:
2638        case OP_NOTPOSUPTOI:
2639    
2640      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]);
2641      break;      break;
2642  #endif  #endif
# Line 2606  for (code = first_significant_code(code Line 2647  for (code = first_significant_code(code
2647      case OP_MARK:      case OP_MARK:
2648      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
2649      case OP_SKIP_ARG:      case OP_SKIP_ARG:
     code += code[1];  
     break;  
   
2650      case OP_THEN_ARG:      case OP_THEN_ARG:
2651      code += code[1];      code += code[1];
2652      break;      break;
# Line 2652  could_be_empty(const pcre_uchar *code, c Line 2690  could_be_empty(const pcre_uchar *code, c
2690  {  {
2691  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2692    {    {
2693    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2694      return FALSE;      return FALSE;
2695    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2696    }    }
# Line 2662  return TRUE; Line 2700  return TRUE;
2700    
2701    
2702  /*************************************************  /*************************************************
2703  *           Check for POSIX class syntax         *  *        Base opcode of repeated opcodes         *
2704  *************************************************/  *************************************************/
2705    
2706  /* This function is called when the sequence "[:" or "[." or "[=" is  /* Returns the base opcode for repeated single character type opcodes. If the
2707  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  
2708    
2709  Returns:   TRUE or FALSE  Arguments:  c opcode
2710    Returns:    base opcode for the type
2711  */  */
2712    
2713  static BOOL  static pcre_uchar
2714  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)  get_repeat_base(pcre_uchar c)
2715  {  {
2716  int terminator;          /* Don't combine these lines; the Solaris cc */  return (c > OP_TYPEPOSUPTO)? c :
2717  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */         (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2718  for (++ptr; *ptr != 0; ptr++)         (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2719    {         (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2720    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)         (c >= OP_STARI)?      OP_STARI :
2721      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;  
2722  }  }
2723    
2724    
2725    
2726    #ifdef SUPPORT_UCP
2727  /*************************************************  /*************************************************
2728  *          Check POSIX class name                *  *        Check a character and a property        *
2729  *************************************************/  *************************************************/
2730    
2731  /* 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
2732  such as [:alnum:].  is adjacent to a fixed character.
2733    
2734  Arguments:  Arguments:
2735    ptr        points to the first letter    c            the character
2736    len        the length of the name    ptype        the property type
2737      pdata        the data for the type
2738      negated      TRUE if it's a negated property (\P or \p{^)
2739    
2740  Returns:     a value representing the name, or -1 if unknown  Returns:       TRUE if auto-possessifying is OK
2741  */  */
2742    
2743  static int  static BOOL
2744  check_posix_name(const pcre_uchar *ptr, int len)  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2745      BOOL negated)
2746  {  {
2747  const char *pn = posix_names;  const pcre_uint32 *p;
2748  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   *  
 *************************************************/  
2749    
2750  /* OP_RECURSE items contain an offset from the start of the regex to the group  switch(ptype)
2751  that is referenced. This means that groups can be replicated for fixed    {
2752  repetition simply by copying (because the recursion is allowed to refer to    case PT_LAMP:
2753  earlier groups that are outside the current group). However, when a group is    return (prop->chartype == ucp_Lu ||
2754  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is            prop->chartype == ucp_Ll ||
2755  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.  
2756    
2757  This function has been extended with the possibility of forward references for    case PT_GC:
2758  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).  
2759    
2760  Arguments:    case PT_PC:
2761    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  
2762    
2763  Returns:     nothing    case PT_SC:
2764  */    return (pdata == prop->script) == negated;
2765    
2766  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;  
2767    
2768  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)    case PT_ALNUM:
2769    {    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2770    int offset;            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
   pcre_uchar *hc;  
2771    
2772    /* 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
2773    reference. */    means that Perl space and POSIX space are now identical. PCRE was changed
2774      at release 8.34. */
2775    
2776    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    case PT_SPACE:    /* Perl space */
2777      case PT_PXSPACE:  /* POSIX space */
2778      switch(c)
2779      {      {
2780      offset = GET(hc, 0);      HSPACE_CASES:
2781      if (cd->start_code + offset == ptr + 1)      VSPACE_CASES:
2782        {      return negated;
2783        PUT(hc, 0, offset + adjust);  
2784        break;      default:
2785        }      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2786      }      }
2787      break;  /* Control never reaches here */
2788    
2789    /* Otherwise, adjust the recursion offset if it's after the start of this    case PT_WORD:
2790    group. */    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2791              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2792              c == CHAR_UNDERSCORE) == negated;
2793    
2794    if (hc >= cd->hwm)    case PT_CLIST:
2795      p = PRIV(ucd_caseless_sets) + prop->caseset;
2796      for (;;)
2797      {      {
2798      offset = GET(ptr, 1);      if (c < *p) return !negated;
2799      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      if (c == *p++) return negated;
2800      }      }
2801      break;  /* Control never reaches here */
   ptr += 1 + LINK_SIZE;  
2802    }    }
2803    
2804    return FALSE;
2805  }  }
2806    #endif  /* SUPPORT_UCP */
2807    
2808    
2809    
2810  /*************************************************  /*************************************************
2811  *        Insert an automatic callout point       *  *        Fill the character property list        *
2812  *************************************************/  *************************************************/
2813    
2814  /* 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-
2815  callout points before each pattern item.  possessification, and if so, fills a list with its properties.
2816    
2817  Arguments:  Arguments:
2818    code           current code pointer    code        points to start of expression
2819    ptr            current pattern pointer    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2820    cd             pointers to tables etc    fcc         points to case-flipping table
2821      list        points to output list
2822                  list[0] will be filled with the opcode
2823                  list[1] will be non-zero if this opcode
2824                    can match an empty character string
2825                  list[2..7] depends on the opcode
2826    
2827  Returns:         new code pointer  Returns:      points to the start of the next opcode if *code is accepted
2828                  NULL if *code is not accepted
2829  */  */
2830    
2831  static pcre_uchar *  static const pcre_uchar *
2832  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)  get_chr_property_list(const pcre_uchar *code, BOOL utf,
2833      const pcre_uint8 *fcc, pcre_uint32 *list)
2834  {  {
2835  *code++ = OP_CALLOUT;  pcre_uchar c = *code;
2836  *code++ = 255;  const pcre_uchar *end;
2837  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */  const pcre_uint32 *clist_src;
2838  PUT(code, LINK_SIZE, 0);                       /* Default length */  pcre_uint32 *clist_dest;
2839  return code + 2 * LINK_SIZE;  pcre_uint32 chr;
2840  }  pcre_uchar base;
2841    
2842    list[0] = c;
2843    list[1] = FALSE;
2844    code++;
2845    
2846    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2847      {
2848      base = get_repeat_base(c);
2849      c -= (base - OP_STAR);
2850    
2851  /*************************************************    if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2852  *         Complete a callout item                *      code += IMM2_SIZE;
 *************************************************/  
2853    
2854  /* 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.  
2855    
2856  Arguments:    switch(base)
2857    previous_callout   points to previous callout item      {
2858    ptr                current pattern pointer      case OP_STAR:
2859    cd                 pointers to tables etc      list[0] = OP_CHAR;
2860        break;
2861    
2862  Returns:             nothing      case OP_STARI:
2863  */      list[0] = OP_CHARI;
2864        break;
2865    
2866  static void      case OP_NOTSTAR:
2867  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)      list[0] = OP_NOT;
2868  {      break;
 int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));  
 PUT(previous_callout, 2 + LINK_SIZE, length);  
 }  
2869    
2870        case OP_NOTSTARI:
2871        list[0] = OP_NOTI;
2872        break;
2873    
2874        case OP_TYPESTAR:
2875        list[0] = *code;
2876        code++;
2877        break;
2878        }
2879      c = list[0];
2880      }
2881    
2882  #ifdef SUPPORT_UCP  switch(c)
2883  /*************************************************    {
2884  *           Get othercase range                  *    case OP_NOT_DIGIT:
2885  *************************************************/    case OP_DIGIT:
2886      case OP_NOT_WHITESPACE:
2887      case OP_WHITESPACE:
2888      case OP_NOT_WORDCHAR:
2889      case OP_WORDCHAR:
2890      case OP_ANY:
2891      case OP_ALLANY:
2892      case OP_ANYNL:
2893      case OP_NOT_HSPACE:
2894      case OP_HSPACE:
2895      case OP_NOT_VSPACE:
2896      case OP_VSPACE:
2897      case OP_EXTUNI:
2898      case OP_EODN:
2899      case OP_EOD:
2900      case OP_DOLL:
2901      case OP_DOLLM:
2902      return code;
2903    
2904  /* This function is passed the start and end of a class range, in UTF-8 mode    case OP_CHAR:
2905  with UCP support. It searches up the characters, looking for ranges of    case OP_NOT:
2906  characters in the "other" case. Each call returns the next one, updating the    GETCHARINCTEST(chr, code);
2907  start address. A character with multiple other cases is returned on its own    list[2] = chr;
2908  with a special return value.    list[3] = NOTACHAR;
2909      return code;
2910    
2911  Arguments:    case OP_CHARI:
2912    cptr        points to starting character value; updated    case OP_NOTI:
2913    d           end value    list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2914    ocptr       where to put start of othercase range    GETCHARINCTEST(chr, code);
2915    odptr       where to put end of othercase range    list[2] = chr;
2916    
2917  Yield:        -1 when no more  #ifdef SUPPORT_UCP
2918                 0 when a range is returned    if (chr < 128 || (chr < 256 && !utf))
2919                >0 the CASESET offset for char with multiple other cases      list[3] = fcc[chr];
2920                  in this case, ocptr contains the original    else
2921  */      list[3] = UCD_OTHERCASE(chr);
2922    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2923      list[3] = (chr < 256) ? fcc[chr] : chr;
2924    #else
2925      list[3] = fcc[chr];
2926    #endif
2927    
2928  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;  
2929    
2930  /* Find the first character that has an other case. If it has multiple other    if (chr == list[3])
2931  cases, return its case offset value. */      list[3] = NOTACHAR;
2932      else
2933        list[4] = NOTACHAR;
2934      return code;
2935    
2936  for (c = *cptr; c <= d; c++)  #ifdef SUPPORT_UCP
2937    {    case OP_PROP:
2938    if ((co = UCD_CASESET(c)) != 0)    case OP_NOTPROP:
2939      if (code[0] != PT_CLIST)
2940      {      {
2941      *ocptr = c++;   /* Character that has the set */      list[2] = code[0];
2942      *cptr = c;      /* Rest of input range */      list[3] = code[1];
2943      return co;      return code + 2;
2944      }      }
   if ((othercase = UCD_OTHERCASE(c)) != c) break;  
   }  
2945    
2946  if (c > d) return -1;  /* Reached end of range */    /* Convert only if we have enough space. */
2947    
2948  *ocptr = othercase;    clist_src = PRIV(ucd_caseless_sets) + code[1];
2949  next = othercase + 1;    clist_dest = list + 2;
2950      code += 2;
2951    
2952  for (++c; c <= d; c++)    do {
2953    {       if (clist_dest >= list + 8)
2954    if (UCD_OTHERCASE(c) != next) break;         {
2955    next++;         /* Early return if there is not enough space. This should never
2956    }         happen, since all clists are shorter than 5 character now. */
2957           list[2] = code[0];
2958           list[3] = code[1];
2959           return code;
2960           }
2961         *clist_dest++ = *clist_src;
2962         }
2963      while(*clist_src++ != NOTACHAR);
2964    
2965  *odptr = next - 1;     /* End of othercase range */    /* All characters are stored. The terminating NOTACHAR
2966  *cptr = c;             /* Rest of input range */    is copied form the clist itself. */
2967  return 0;  
2968      list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
2969      return code;
2970    #endif
2971    
2972      case OP_NCLASS:
2973      case OP_CLASS:
2974    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2975      case OP_XCLASS:
2976    
2977      if (c == OP_XCLASS)
2978        end = code + GET(code, 0);
2979      else
2980    #endif
2981        end = code + 32 / sizeof(pcre_uchar);
2982    
2983      switch(*end)
2984        {
2985        case OP_CRSTAR:
2986        case OP_CRMINSTAR:
2987        case OP_CRQUERY:
2988        case OP_CRMINQUERY:
2989        case OP_CRPOSSTAR:
2990        case OP_CRPOSQUERY:
2991        list[1] = TRUE;
2992        end++;
2993        break;
2994    
2995        case OP_CRPLUS:
2996        case OP_CRMINPLUS:
2997        case OP_CRPOSPLUS:
2998        end++;
2999        break;
3000    
3001        case OP_CRRANGE:
3002        case OP_CRMINRANGE:
3003        case OP_CRPOSRANGE:
3004        list[1] = (GET2(end, 1) == 0);
3005        end += 1 + 2 * IMM2_SIZE;
3006        break;
3007        }
3008      list[2] = end - code;
3009      return end;
3010      }
3011    return NULL;    /* Opcode not accepted */
3012  }  }
3013    
3014    
3015    
3016  /*************************************************  /*************************************************
3017  *        Check a character and a property        *  *    Scan further character sets for match       *
3018  *************************************************/  *************************************************/
3019    
3020  /* 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
3021  is adjacent to a fixed character.  which case the base cannot be possessified.
3022    
3023  Arguments:  Arguments:
3024    c            the character    code        points to the byte code
3025    ptype        the property type    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3026    pdata        the data for the type    cd          static compile data
3027    negated      TRUE if it's a negated property (\P or \p{^)    base_list   the data list of the base opcode
3028    
3029  Returns:       TRUE if auto-possessifying is OK  Returns:      TRUE if the auto-possessification is possible
3030  */  */
3031    
3032  static BOOL  static BOOL
3033  check_char_prop(pcre_uint32 c, int ptype, int pdata, BOOL negated)  compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3034      const pcre_uint32* base_list, const pcre_uchar *base_end)
3035  {  {
3036  #ifdef SUPPORT_UCP  pcre_uchar c;
3037  const pcre_uint32 *p;  pcre_uint32 list[8];
3038  #endif  const pcre_uint32* chr_ptr;
3039    const pcre_uint32* ochr_ptr;
3040  const ucd_record *prop = GET_UCD(c);  const pcre_uint32* list_ptr;
3041    const pcre_uchar *next_code;
3042    const pcre_uint8 *class_bits;
3043    pcre_uint32 chr;
3044    BOOL accepted;
3045    
3046    /* Note: the base_list[1] contains whether the current opcode has greedy
3047    (represented by a non-zero value) quantifier. This is a different from
3048    other character type lists, which stores here that the character iterator
3049    matches to an empty string (also represented by a non-zero value). */
3050    
3051  switch(ptype)  for(;;)
3052    {    {
3053    case PT_LAMP:    c = *code;
   return (prop->chartype == ucp_Lu ||  
           prop->chartype == ucp_Ll ||  
           prop->chartype == ucp_Lt) == negated;  
   
   case PT_GC:  
   return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;  
3054    
3055    case PT_PC:    /* Skip over callouts */
   return (pdata == prop->chartype) == negated;  
3056    
3057    case PT_SC:    if (c == OP_CALLOUT)
3058    return (pdata == prop->script) == negated;      {
3059        code += PRIV(OP_lengths)[c];
3060        continue;
3061        }
3062    
3063    /* These are specials */    if (c == OP_ALT)
3064        {
3065        do code += GET(code, 1); while (*code == OP_ALT);
3066        c = *code;
3067        }
3068    
3069    case PT_ALNUM:    switch(c)
3070    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||      {
3071            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;      case OP_END:
3072        case OP_KETRPOS:
3073        /* TRUE only in greedy case. The non-greedy case could be replaced by
3074        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3075        uses more memory, which we cannot get at this stage.) */
3076    
3077    case PT_SPACE:    /* Perl space */      return base_list[1] != 0;
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||  
           c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)  
           == negated;  
3078    
3079    case PT_PXSPACE:  /* POSIX space */      case OP_KET:
3080    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||      /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3081            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||      it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3082            c == CHAR_FF || c == CHAR_CR)      cannot be converted to a possessive form. */
3083            == negated;  
3084        if (base_list[1] == 0) return FALSE;
3085    
3086        switch(*(code - GET(code, 1)))
3087          {
3088          case OP_ASSERT:
3089          case OP_ASSERT_NOT:
3090          case OP_ASSERTBACK:
3091          case OP_ASSERTBACK_NOT:
3092          case OP_ONCE:
3093          case OP_ONCE_NC:
3094          /* Atomic sub-patterns and assertions can always auto-possessify their
3095          last iterator. */
3096          return TRUE;
3097          }
3098    
3099    case PT_WORD:      code += PRIV(OP_lengths)[c];
3100    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||      continue;
           PRIV(ucp_gentype)[prop->chartype] == ucp_N ||  
           c == CHAR_UNDERSCORE) == negated;  
   
 #ifdef SUPPORT_UCP  
   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  
   }  
3101    
3102  return FALSE;      case OP_ONCE:
3103  }      case OP_ONCE_NC:
3104  #endif  /* SUPPORT_UCP */      case OP_BRA:
3105        case OP_CBRA:
3106        next_code = code;
3107        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3108    
3109        /* We do not support repeated brackets, because they can lead to
3110        infinite recursion. */
3111    
3112        if (*next_code != OP_KET) return FALSE;
3113    
3114  /*************************************************      next_code = code + GET(code, 1);
3115  *     Check if auto-possessifying is possible    *      code += PRIV(OP_lengths)[c];
 *************************************************/  
3116    
3117  /* This function is called for unlimited repeats of certain items, to see      while (*next_code == OP_ALT)
3118  whether the next thing could possibly match the repeated item. If not, it makes        {
3119  sense to automatically possessify the repeated item.        if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3120          code = next_code + 1 + LINK_SIZE;
3121          next_code += GET(next_code, 1);
3122          }
3123        continue;
3124    
3125  Arguments:      case OP_BRAZERO:
3126    previous      pointer to the repeated opcode      case OP_BRAMINZERO:
   utf           TRUE in UTF-8 / UTF-16 / UTF-32 mode  
   ptr           next character in pattern  
   options       options bits  
   cd            contains pointers to tables etc.  
3127    
3128  Returns:        TRUE if possessifying is wanted      next_code = code + 1;
3129  */      if (*next_code != OP_BRA && *next_code != OP_CBRA)
3130          return FALSE;
3131    
3132  static BOOL      do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
 check_auto_possessive(const pcre_uchar *previous, BOOL utf,  
   const pcre_uchar *ptr, int options, compile_data *cd)  
 {  
 pcre_uint32 c = NOTACHAR;  
 pcre_uint32 next;  
 int escape;  
 int op_code = *previous++;  
3133    
3134  /* Skip whitespace and comments in extended mode */      /* We do not support repeated brackets, because they can lead to
3135        infinite recursion. */
3136        if (*next_code != OP_KET) return FALSE;
3137    
3138        /* The bracket content will be checked by the
3139        OP_BRA/OP_CBRA case above. */
3140        next_code += 1 + LINK_SIZE;
3141        if (!compare_opcodes(next_code, utf, cd, base_list, base_end)) return FALSE;
3142    
3143  if ((options & PCRE_EXTENDED) != 0)      code += PRIV(OP_lengths)[c];
3144    {      continue;
   for (;;)  
     {  
     while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;  
     if (*ptr == CHAR_NUMBER_SIGN)  
       {  
       ptr++;  
       while (*ptr != 0)  
         {  
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
       }  
     else break;  
3145      }      }
   }  
3146    
3147  /* If the next item is one that we can handle, get its value. A non-negative    /* Check for a supported opcode, and load its properties. */
 value is a character, a negative value is an escape value. */  
3148    
3149  if (*ptr == CHAR_BACKSLASH)    code = get_chr_property_list(code, utf, cd->fcc, list);
3150    {    if (code == NULL) return FALSE;    /* Unsupported */
   int temperrorcode = 0;  
   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;  
3151    
3152  /* Skip whitespace and comments in extended mode */    /* If either opcode is a small character list, set pointers for comparing
3153      characters from that list with another list, or with a property. */
3154    
3155  if ((options & PCRE_EXTENDED) != 0)    if (base_list[0] == OP_CHAR)
   {  
   for (;;)  
3156      {      {
3157      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      chr_ptr = base_list + 2;
3158      if (*ptr == CHAR_NUMBER_SIGN)      list_ptr = list;
3159        {      }
3160        ptr++;    else if (list[0] == OP_CHAR)
3161        while (*ptr != 0)      {
3162          {      chr_ptr = list + 2;
3163          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }      list_ptr = base_list;
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
       }  
     else break;  
3164      }      }
   }  
3165    
3166  /* If the next thing is itself optional, we have to give up. */    /* Some property combinations also acceptable. Unicode property opcodes are
3167      processed specially; the rest can be handled with a lookup table. */
3168    
3169  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||    else
3170    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)      {
3171      return FALSE;      pcre_uint32 leftop, rightop;
3172    
3173  /* If the previous item is a character, get its value. */      leftop = base_list[0];
3174        rightop = list[0];
3175    
3176  if (op_code == OP_CHAR || op_code == OP_CHARI ||  #ifdef SUPPORT_UCP
3177      op_code == OP_NOT || op_code == OP_NOTI)      accepted = FALSE; /* Always set in non-unicode case. */
3178    //if (escape == 0) switch(op_code)      if (leftop == OP_PROP || leftop == OP_NOTPROP)
3179    {        {
3180  #ifdef SUPPORT_UTF        if (rightop == OP_EOD)
3181    GETCHARTEST(c, previous);          accepted = TRUE;
3182  #else        else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3183    c = *previous;          {
3184  #endif          int n;
3185    }          const pcre_uint8 *p;
3186            BOOL same = leftop == rightop;
3187            BOOL lisprop = leftop == OP_PROP;
3188            BOOL risprop = rightop == OP_PROP;
3189            BOOL bothprop = lisprop && risprop;
3190    
3191            /* There's a table that specifies how each combination is to be
3192            processed:
3193              0   Always return FALSE (never auto-possessify)
3194              1   Character groups are distinct (possessify if both are OP_PROP)
3195              2   Check character categories in the same group (general or particular)
3196              3   Return TRUE if the two opcodes are not the same
3197              ... see comments below
3198            */
3199    
3200            n = propposstab[base_list[2]][list[2]];
3201            switch(n)
3202              {
3203              case 0: break;
3204              case 1: accepted = bothprop; break;
3205              case 2: accepted = (base_list[3] == list[3]) != same; break;
3206              case 3: accepted = !same; break;
3207    
3208  /* Now compare the next item with the previous opcode. First, handle cases when            case 4:  /* Left general category, right particular category */
3209  the next item is a character. */            accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3210              break;
3211    
3212  if (escape == 0)            case 5:  /* Right general category, left particular category */
3213    {            accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3214    /* For a caseless UTF match, the next character may have more than one other            break;
   case, which maps to the special PT_CLIST property. Check this first. */  
   
 #ifdef SUPPORT_UCP  
   if (utf && c != NOTACHAR && (options & PCRE_CASELESS) != 0)  
     {  
     int ocs = UCD_CASESET(next);  
     if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);  
     }  
 #endif  
3215    
3216    switch(op_code)            /* This code is logically tricky. Think hard before fiddling with it.
3217      {            The posspropstab table has four entries per row. Each row relates to
3218      case OP_CHAR:            one of PCRE's special properties such as ALNUM or SPACE or WORD.
3219      return c != next;            Only WORD actually needs all four entries, but using repeats for the
3220              others means they can all use the same code below.
3221    
3222              The first two entries in each row are Unicode general categories, and
3223              apply always, because all the characters they include are part of the
3224              PCRE character set. The third and fourth entries are a general and a
3225              particular category, respectively, that include one or more relevant
3226              characters. One or the other is used, depending on whether the check
3227              is for a general or a particular category. However, in both cases the
3228              category contains more characters than the specials that are defined
3229              for the property being tested against. Therefore, it cannot be used
3230              in a NOTPROP case.
3231    
3232              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3233              Underscore is covered by ucp_P or ucp_Po. */
3234    
3235              case 6:  /* Left alphanum vs right general category */
3236              case 7:  /* Left space vs right general category */
3237              case 8:  /* Left word vs right general category */
3238              p = posspropstab[n-6];
3239              accepted = risprop && lisprop ==
3240                (list[3] != p[0] &&
3241                 list[3] != p[1] &&
3242                (list[3] != p[2] || !lisprop));
3243              break;
3244    
3245      /* For CHARI (caseless character) we must check the other case. If we have            case 9:   /* Right alphanum vs left general category */
3246      Unicode property support, we can use it to test the other case of            case 10:  /* Right space vs left general category */
3247      high-valued characters. We know that next can have only one other case,            case 11:  /* Right word vs left general category */
3248      because multi-other-case characters are dealt with above. */            p = posspropstab[n-9];
3249              accepted = lisprop && risprop ==
3250                (base_list[3] != p[0] &&
3251                 base_list[3] != p[1] &&
3252                (base_list[3] != p[2] || !risprop));
3253              break;
3254    
3255      case OP_CHARI:            case 12:  /* Left alphanum vs right particular category */
3256      if (c == next) return FALSE;            case 13:  /* Left space vs right particular category */
3257  #ifdef SUPPORT_UTF            case 14:  /* Left word vs right particular category */
3258      if (utf)            p = posspropstab[n-12];
3259              accepted = risprop && lisprop ==
3260                (catposstab[p[0]][list[3]] &&
3261                 catposstab[p[1]][list[3]] &&
3262                (list[3] != p[3] || !lisprop));
3263              break;
3264    
3265              case 15:  /* Right alphanum vs left particular category */
3266              case 16:  /* Right space vs left particular category */
3267              case 17:  /* Right word vs left particular category */
3268              p = posspropstab[n-15];
3269              accepted = lisprop && risprop ==
3270                (catposstab[p[0]][base_list[3]] &&
3271                 catposstab[p[1]][base_list[3]] &&
3272                (base_list[3] != p[3] || !risprop));
3273              break;
3274              }
3275            }
3276          }
3277    
3278        else
3279    #endif  /* SUPPORT_UCP */
3280    
3281        accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3282               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3283               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3284    
3285        if (!accepted)
3286          return FALSE;
3287    
3288        if (list[1] == 0) return TRUE;
3289        /* Might be an empty repeat. */
3290        continue;
3291        }
3292    
3293      /* Control reaches here only if one of the items is a small character list.
3294      All characters are checked against the other side. */
3295    
3296      do
3297        {
3298        chr = *chr_ptr;
3299    
3300        switch(list_ptr[0])
3301        {        {
3302        pcre_uint32 othercase;        case OP_CHAR:
3303        if (next < 128) othercase = cd->fcc[next]; else        ochr_ptr = list_ptr + 2;
3304  #ifdef SUPPORT_UCP        do
3305        othercase = UCD_OTHERCASE(next);          {
3306  #else          if (chr == *ochr_ptr) return FALSE;
3307        othercase = NOTACHAR;          ochr_ptr++;
3308            }
3309          while(*ochr_ptr != NOTACHAR);
3310          break;
3311    
3312          case OP_NOT:
3313          ochr_ptr = list_ptr + 2;
3314          do
3315            {
3316            if (chr == *ochr_ptr)
3317              break;
3318            ochr_ptr++;
3319            }
3320          while(*ochr_ptr != NOTACHAR);
3321          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3322          break;
3323    
3324          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3325          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3326    
3327          case OP_DIGIT:
3328          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3329          break;
3330    
3331          case OP_NOT_DIGIT:
3332          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3333          break;
3334    
3335          case OP_WHITESPACE:
3336          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3337          break;
3338    
3339          case OP_NOT_WHITESPACE:
3340          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3341          break;
3342    
3343          case OP_WORDCHAR:
3344          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3345          break;
3346    
3347          case OP_NOT_WORDCHAR:
3348          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3349          break;
3350    
3351          case OP_HSPACE:
3352          switch(chr)
3353            {
3354            HSPACE_CASES: return FALSE;
3355            default: break;
3356            }
3357          break;
3358    
3359          case OP_NOT_HSPACE:
3360          switch(chr)
3361            {
3362            HSPACE_CASES: break;
3363            default: return FALSE;
3364            }
3365          break;
3366    
3367          case OP_ANYNL:
3368          case OP_VSPACE:
3369          switch(chr)
3370            {
3371            VSPACE_CASES: return FALSE;
3372            default: break;
3373            }
3374          break;
3375    
3376          case OP_NOT_VSPACE:
3377          switch(chr)
3378            {
3379            VSPACE_CASES: break;
3380            default: return FALSE;
3381            }
3382          break;
3383    
3384          case OP_DOLL:
3385          case OP_EODN:
3386          switch (chr)
3387            {
3388            case CHAR_CR:
3389            case CHAR_LF:
3390            case CHAR_VT:
3391            case CHAR_FF:
3392            case CHAR_NEL:
3393    #ifndef EBCDIC
3394            case 0x2028:
3395            case 0x2029:
3396    #endif  /* Not EBCDIC */
3397            return FALSE;
3398            }
3399          break;
3400    
3401          case OP_EOD:    /* Can always possessify before \z */
3402          break;
3403    
3404          case OP_PROP:
3405          case OP_NOTPROP:
3406          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3407                list_ptr[0] == OP_NOTPROP))
3408            return FALSE;
3409          break;
3410    
3411          case OP_NCLASS:
3412          if (chr > 255) return FALSE;
3413          /* Fall through */
3414    
3415          case OP_CLASS:
3416          if (chr > 255) break;
3417          class_bits = (pcre_uint8 *)((list_ptr == list ? code : base_end) - list_ptr[2]);
3418          if ((class_bits[chr >> 3] & (1 << (chr & 7))) != 0)
3419            return FALSE;
3420          break;
3421    
3422    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3423          case OP_XCLASS:
3424          if (list_ptr != list) return FALSE;   /* Class is first opcode */
3425          if (PRIV(xclass)(chr, code - list_ptr[2] + LINK_SIZE, utf))
3426            return FALSE;
3427          break;
3428  #endif  #endif
3429        return c != othercase;  
3430          default:
3431          return FALSE;
3432          }
3433    
3434        chr_ptr++;
3435        }
3436      while(*chr_ptr != NOTACHAR);
3437    
3438      /* At least one character must be matched from this opcode. */
3439    
3440      if (list[1] == 0) return TRUE;
3441      }
3442    
3443    return FALSE;
3444    }
3445    
3446    
3447    
3448    /*************************************************
3449    *    Scan compiled regex for auto-possession     *
3450    *************************************************/
3451    
3452    /* Replaces single character iterations with their possessive alternatives
3453    if appropriate. This function modifies the compiled opcode!
3454    
3455    Arguments:
3456      code        points to start of the byte code
3457      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3458      cd          static compile data
3459    
3460    Returns:      nothing
3461    */
3462    
3463    static void
3464    auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3465    {
3466    register pcre_uchar c;
3467    const pcre_uchar *end;
3468    pcre_uchar *repeat_code;
3469    pcre_uint32 list[8];
3470    
3471    for (;;)
3472      {
3473      c = *code;
3474    
3475      if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3476        {
3477        c -= get_repeat_base(c) - OP_STAR;
3478        end = (c <= OP_MINUPTO) ?
3479          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3480        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3481    
3482        if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3483          {
3484          switch(c)
3485            {
3486            case OP_STAR:
3487            *code += OP_POSSTAR - OP_STAR;
3488            break;
3489    
3490            case OP_MINSTAR:
3491            *code += OP_POSSTAR - OP_MINSTAR;
3492            break;
3493    
3494            case OP_PLUS:
3495            *code += OP_POSPLUS - OP_PLUS;
3496            break;
3497    
3498            case OP_MINPLUS:
3499            *code += OP_POSPLUS - OP_MINPLUS;
3500            break;
3501    
3502            case OP_QUERY:
3503            *code += OP_POSQUERY - OP_QUERY;
3504            break;
3505    
3506            case OP_MINQUERY:
3507            *code += OP_POSQUERY - OP_MINQUERY;
3508            break;
3509    
3510            case OP_UPTO:
3511            *code += OP_POSUPTO - OP_UPTO;
3512            break;
3513    
3514            case OP_MINUPTO:
3515            *code += OP_MINUPTO - OP_UPTO;
3516            break;
3517            }
3518        }        }
3519        c = *code;
3520        }
3521      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3522        {
3523    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3524        if (c == OP_XCLASS)
3525          repeat_code = code + 1 + GET(code, 1);
3526      else      else
3527  #endif  /* SUPPORT_UTF */  #endif
3528      return (c != TABLE_GET(next, cd->fcc, next));  /* Not UTF */        repeat_code = code + 1 + (32 / sizeof(pcre_uchar));
3529    
3530        c = *repeat_code;
3531        if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3532          {
3533          /* end must not be NULL. */
3534          end = get_chr_property_list(code, utf, cd->fcc, list);
3535    
3536          list[1] = (c & 1) == 0;
3537    
3538          if (compare_opcodes(end, utf, cd, list, end))
3539            {
3540            switch (c)
3541              {
3542              case OP_CRSTAR:
3543              *repeat_code = OP_CRPOSSTAR;
3544              break;
3545    
3546              case OP_CRPLUS:
3547              *repeat_code = OP_CRPOSPLUS;
3548              break;
3549    
3550              case OP_CRQUERY:
3551              *repeat_code = OP_CRPOSQUERY;
3552              break;
3553    
3554              case OP_CRRANGE:
3555              *repeat_code = OP_CRPOSRANGE;
3556              break;
3557              }
3558            }
3559          }
3560        c = *code;
3561        }
3562    
3563      switch(c)
3564        {
3565        case OP_END:
3566        return;
3567    
3568        case OP_TYPESTAR:
3569        case OP_TYPEMINSTAR:
3570        case OP_TYPEPLUS:
3571        case OP_TYPEMINPLUS:
3572        case OP_TYPEQUERY:
3573        case OP_TYPEMINQUERY:
3574        case OP_TYPEPOSSTAR:
3575        case OP_TYPEPOSPLUS:
3576        case OP_TYPEPOSQUERY:
3577        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3578        break;
3579    
3580        case OP_TYPEUPTO:
3581        case OP_TYPEMINUPTO:
3582        case OP_TYPEEXACT:
3583        case OP_TYPEPOSUPTO:
3584        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3585          code += 2;
3586        break;
3587    
3588    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3589        case OP_XCLASS:
3590        code += GET(code, 1);
3591        break;
3592    #endif
3593    
3594        case OP_MARK:
3595        case OP_PRUNE_ARG:
3596        case OP_SKIP_ARG:
3597        case OP_THEN_ARG:
3598        code += code[1];
3599        break;
3600        }
3601    
3602      /* Add in the fixed length from the table */
3603    
3604      code += PRIV(OP_lengths)[c];
3605    
3606      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3607      a multi-byte character. The length in the table is a minimum, so we have to
3608      arrange to skip the extra bytes. */
3609    
3610    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3611      if (utf) switch(c)
3612        {
3613        case OP_CHAR:
3614        case OP_CHARI:
3615      case OP_NOT:      case OP_NOT:
     return c == next;  
   
3616      case OP_NOTI:      case OP_NOTI:
3617      if (c == next) return TRUE;      case OP_STAR:
3618  #ifdef SUPPORT_UTF      case OP_MINSTAR:
3619      if (utf)      case OP_PLUS:
3620        {      case OP_MINPLUS:
3621        pcre_uint32 othercase;      case OP_QUERY:
3622        if (next < 128) othercase = cd->fcc[next]; else      case OP_MINQUERY:
3623  #ifdef SUPPORT_UCP      case OP_UPTO:
3624        othercase = UCD_OTHERCASE(next);      case OP_MINUPTO:
3625        case OP_EXACT:
3626        case OP_POSSTAR:
3627        case OP_POSPLUS:
3628        case OP_POSQUERY:
3629        case OP_POSUPTO:
3630        case OP_STARI:
3631        case OP_MINSTARI:
3632        case OP_PLUSI:
3633        case OP_MINPLUSI:
3634        case OP_QUERYI:
3635        case OP_MINQUERYI:
3636        case OP_UPTOI:
3637        case OP_MINUPTOI:
3638        case OP_EXACTI:
3639        case OP_POSSTARI:
3640        case OP_POSPLUSI:
3641        case OP_POSQUERYI:
3642        case OP_POSUPTOI:
3643        case OP_NOTSTAR:
3644        case OP_NOTMINSTAR:
3645        case OP_NOTPLUS:
3646        case OP_NOTMINPLUS:
3647        case OP_NOTQUERY:
3648        case OP_NOTMINQUERY:
3649        case OP_NOTUPTO:
3650        case OP_NOTMINUPTO:
3651        case OP_NOTEXACT:
3652        case OP_NOTPOSSTAR:
3653        case OP_NOTPOSPLUS:
3654        case OP_NOTPOSQUERY:
3655        case OP_NOTPOSUPTO:
3656        case OP_NOTSTARI:
3657        case OP_NOTMINSTARI:
3658        case OP_NOTPLUSI:
3659        case OP_NOTMINPLUSI:
3660        case OP_NOTQUERYI:
3661        case OP_NOTMINQUERYI:
3662        case OP_NOTUPTOI:
3663        case OP_NOTMINUPTOI:
3664        case OP_NOTEXACTI:
3665        case OP_NOTPOSSTARI:
3666        case OP_NOTPOSPLUSI:
3667        case OP_NOTPOSQUERYI:
3668        case OP_NOTPOSUPTOI:
3669        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3670        break;
3671        }
3672  #else  #else
3673        othercase = NOTACHAR;    (void)(utf);  /* Keep compiler happy by referencing function argument */
3674  #endif  #endif
3675        return c == othercase;    }
3676    }
3677    
3678    
3679    
3680    /*************************************************
3681    *           Check for POSIX class syntax         *
3682    *************************************************/
3683    
3684    /* This function is called when the sequence "[:" or "[." or "[=" is
3685    encountered in a character class. It checks whether this is followed by a
3686    sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3687    reach an unescaped ']' without the special preceding character, return FALSE.
3688    
3689    Originally, this function only recognized a sequence of letters between the
3690    terminators, but it seems that Perl recognizes any sequence of characters,
3691    though of course unknown POSIX names are subsequently rejected. Perl gives an
3692    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3693    didn't consider this to be a POSIX class. Likewise for [:1234:].
3694    
3695    The problem in trying to be exactly like Perl is in the handling of escapes. We
3696    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3697    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3698    below handles the special case of \], but does not try to do any other escape
3699    processing. This makes it different from Perl for cases such as [:l\ower:]
3700    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3701    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
3702    I think.
3703    
3704    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3705    It seems that the appearance of a nested POSIX class supersedes an apparent
3706    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3707    a digit.
3708    
3709    In Perl, unescaped square brackets may also appear as part of class names. For
3710    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3711    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3712    seem right at all. PCRE does not allow closing square brackets in POSIX class
3713    names.
3714    
3715    Arguments:
3716      ptr      pointer to the initial [
3717      endptr   where to return the end pointer
3718    
3719    Returns:   TRUE or FALSE
3720    */
3721    
3722    static BOOL
3723    check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3724    {
3725    pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
3726    terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
3727    for (++ptr; *ptr != CHAR_NULL; ptr++)
3728      {
3729      if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3730        ptr++;
3731      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3732      else
3733        {
3734        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3735          {
3736          *endptr = ptr;
3737          return TRUE;
3738        }        }
3739      else      if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3740  #endif  /* SUPPORT_UTF */           (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3741      return (c == TABLE_GET(next, cd->fcc, next));  /* Not UTF */            ptr[1] == CHAR_EQUALS_SIGN) &&
3742            check_posix_syntax(ptr, endptr))
3743          return FALSE;
3744        }
3745      }
3746    return FALSE;
3747    }
3748    
3749    
3750    
3751    
3752    /*************************************************
3753    *          Check POSIX class name                *
3754    *************************************************/
3755    
3756    /* This function is called to check the name given in a POSIX-style class entry
3757    such as [:alnum:].
3758    
3759    Arguments:
3760      ptr        points to the first letter
3761      len        the length of the name
3762    
3763    Returns:     a value representing the name, or -1 if unknown
3764    */
3765    
3766    static int
3767    check_posix_name(const pcre_uchar *ptr, int len)
3768    {
3769    const char *pn = posix_names;
3770    register int yield = 0;
3771    while (posix_name_lengths[yield] != 0)
3772      {
3773      if (len == posix_name_lengths[yield] &&
3774        STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3775      pn += posix_name_lengths[yield] + 1;
3776      yield++;
3777      }
3778    return -1;
3779    }
3780    
     /* 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. */  
3781    
3782      case OP_DIGIT:  /*************************************************
3783      return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;  *    Adjust OP_RECURSE items in repeated group   *
3784    *************************************************/
3785    
3786      case OP_NOT_DIGIT:  /* OP_RECURSE items contain an offset from the start of the regex to the group
3787      return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;  that is referenced. This means that groups can be replicated for fixed
3788    repetition simply by copying (because the recursion is allowed to refer to
3789    earlier groups that are outside the current group). However, when a group is
3790    optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3791    inserted before it, after it has been compiled. This means that any OP_RECURSE
3792    items within it that refer to the group itself or any contained groups have to
3793    have their offsets adjusted. That one of the jobs of this function. Before it
3794    is called, the partially compiled regex must be temporarily terminated with
3795    OP_END.
3796    
3797      case OP_WHITESPACE:  This function has been extended with the possibility of forward references for
3798      return next > 255 || (cd->ctypes[next] & ctype_space) == 0;  recursions and subroutine calls. It must also check the list of such references
3799    for the group we are dealing with. If it finds that one of the recursions in
3800    the current group is on this list, it adjusts the offset in the list, not the
3801    value in the reference (which is a group number).
3802    
3803      case OP_NOT_WHITESPACE:  Arguments:
3804      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;    group      points to the start of the group
3805      adjust     the amount by which the group is to be moved
3806      utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
3807      cd         contains pointers to tables etc.
3808      save_hwm   the hwm forward reference pointer at the start of the group
3809    
3810      case OP_WORDCHAR:  Returns:     nothing
3811      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;  */
3812    
3813      case OP_NOT_WORDCHAR:  static void
3814      return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3815      pcre_uchar *save_hwm)
3816    {
3817    pcre_uchar *ptr = group;
3818    
3819      case OP_HSPACE:  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3820      case OP_NOT_HSPACE:    {
3821      switch(next)    int offset;
3822        {    pcre_uchar *hc;
       HSPACE_CASES:  
       return op_code == OP_NOT_HSPACE;  
3823    
3824        default:    /* See if this recursion is on the forward reference list. If so, adjust the
3825        return op_code != OP_NOT_HSPACE;    reference. */
       }  
3826    
3827      case OP_ANYNL:    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3828      case OP_VSPACE:      {
3829      case OP_NOT_VSPACE:      offset = (int)GET(hc, 0);
3830      switch(next)      if (cd->start_code + offset == ptr + 1)
3831        {        {
3832        VSPACE_CASES:        PUT(hc, 0, offset + adjust);
3833        return op_code == OP_NOT_VSPACE;        break;
   
       default:  
       return op_code != OP_NOT_VSPACE;  
3834        }        }
   
 #ifdef SUPPORT_UCP  
     case OP_PROP:  
     return check_char_prop(next, previous[0], previous[1], FALSE);  
   
     case OP_NOTPROP:  
     return check_char_prop(next, previous[0], previous[1], TRUE);  
 #endif  
   
     default:  
     return FALSE;  
3835      }      }
   }  
3836    
3837  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP    /* Otherwise, adjust the recursion offset if it's after the start of this
3838  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are    group. */
 generated only when PCRE_UCP is *not* set, that is, when only ASCII  
 characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are  
 replaced by OP_PROP codes when PCRE_UCP is set. */  
3839    
3840  switch(op_code)    if (hc >= cd->hwm)
   {  
   case OP_CHAR:  
   case OP_CHARI:  
   switch(escape)  
3841      {      {
3842      case ESC_d:      offset = (int)GET(ptr, 1);
3843      return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3844        }
     case ESC_D:  
     return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;  
   
     case ESC_s:  
     return c > 255 || (cd->ctypes[c] & ctype_space) == 0;  
3845    
3846      case ESC_S:    ptr += 1 + LINK_SIZE;
3847      return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;    }
3848    }
3849    
     case ESC_w:  
     return c > 255 || (cd->ctypes[c] & ctype_word) == 0;  
3850    
     case ESC_W:  
     return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;  
3851    
3852      case ESC_h:  /*************************************************
3853      case ESC_H:  *        Insert an automatic callout point       *
3854      switch(c)  *************************************************/
       {  
       HSPACE_CASES:  
       return escape != ESC_h;  
   
       default:  
       return escape == ESC_h;  
       }  
3855    
3856      case ESC_v:  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
3857      case ESC_V:  callout points before each pattern item.
     switch(c)  
       {  
       VSPACE_CASES:  
       return escape != ESC_v;  
3858    
3859        default:  Arguments:
3860        return escape == ESC_v;    code           current code pointer
3861        }    ptr            current pattern pointer
3862      cd             pointers to tables etc
3863    
3864      /* When PCRE_UCP is set, these values get generated for \d etc. Find  Returns:         new code pointer
3865      their substitutions and process them. The result will always be either  */
     ESC_p or ESC_P. Then fall through to process those values. */  
3866    
3867  #ifdef SUPPORT_UCP  static pcre_uchar *
3868      case ESC_du:  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
3869      case ESC_DU:  {
3870      case ESC_wu:  *code++ = OP_CALLOUT;
3871      case ESC_WU:  *code++ = 255;
3872      case ESC_su:  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
3873      case ESC_SU:  PUT(code, LINK_SIZE, 0);                       /* Default length */
3874        {  return code + 2 * LINK_SIZE;
3875        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 */  
3876    
     case ESC_p:  
     case ESC_P:  
       {  
       int ptype, pdata, errorcodeptr;  
       BOOL negated;  
3877    
       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 */  
3878    
3879        /* If the property item is optional, we have to give up. (When generated  /*************************************************
3880        from \d etc by PCRE_UCP, this test will have been applied much earlier,  *         Complete a callout item                *
3881        to the original \d etc. At this point, ptr will point to a zero byte. */  *************************************************/
3882    
3883        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  /* A callout item contains the length of the next item in the pattern, which
3884          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
3885            return FALSE;  for both automatic and manual callouts.
3886    
3887        /* Do the property check. */  Arguments:
3888      previous_callout   points to previous callout item
3889      ptr                current pattern pointer
3890      cd                 pointers to tables etc
3891    
3892        return check_char_prop(c, ptype, pdata, (escape == ESC_P) != negated);  Returns:             nothing
3893        }  */
 #endif  
3894    
3895      default:  static void
3896      return FALSE;  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
3897      }  {
3898    int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
3899    PUT(previous_callout, 2 + LINK_SIZE, length);
3900    }
3901    
   /* 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.) */  
3902    
   case OP_DIGIT:  
   return escape == ESC_D || escape == ESC_s || escape == ESC_W ||  
          escape == ESC_h || escape == ESC_v || escape == ESC_R;  
3903    
3904    case OP_NOT_DIGIT:  #ifdef SUPPORT_UCP
3905    return escape == ESC_d;  /*************************************************
3906    *           Get othercase range                  *
3907    *************************************************/
3908    
3909    case OP_WHITESPACE:  /* This function is passed the start and end of a class range, in UTF-8 mode
3910    return escape == ESC_S || escape == ESC_d || escape == ESC_w;  with UCP support. It searches up the characters, looking for ranges of
3911    characters in the "other" case. Each call returns the next one, updating the
3912    start address. A character with multiple other cases is returned on its own
3913    with a special return value.
3914    
3915    case OP_NOT_WHITESPACE:  Arguments:
3916    return escape == ESC_s || escape == ESC_h || escape == ESC_v || escape == ESC_R;    cptr        points to starting character value; updated
3917      d           end value
3918      ocptr       where to put start of othercase range
3919      odptr       where to put end of othercase range
3920    
3921    case OP_HSPACE:  Yield:        -1 when no more
3922    return escape == ESC_S || escape == ESC_H || escape == ESC_d ||                 0 when a range is returned
3923           escape == ESC_w || escape == ESC_v || escape == ESC_R;                >0 the CASESET offset for char with multiple other cases
3924                    in this case, ocptr contains the original
3925    */
3926    
3927    case OP_NOT_HSPACE:  static int
3928    return escape == ESC_h;  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
3929      pcre_uint32 *odptr)
3930    {
3931    pcre_uint32 c, othercase, next;
3932    unsigned int co;
3933    
3934    /* 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
3935    case OP_ANYNL:  cases, return its case offset value. */
   case OP_VSPACE:  
   return escape == ESC_V || escape == ESC_d || escape == ESC_w;  
3936    
3937    case OP_NOT_VSPACE:  for (c = *cptr; c <= d; c++)
3938    return escape == ESC_v || escape == ESC_R;    {
3939      if ((co = UCD_CASESET(c)) != 0)
3940        {
3941        *ocptr = c++;   /* Character that has the set */
3942        *cptr = c;      /* Rest of input range */
3943        return (int)co;
3944        }
3945      if ((othercase = UCD_OTHERCASE(c)) != c) break;
3946      }
3947    
3948    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;  
3949    
3950    case OP_NOT_WORDCHAR:  *ocptr = othercase;
3951    return escape == ESC_w || escape == ESC_d;  next = othercase + 1;
3952    
3953    default:  for (++c; c <= d; c++)
3954    return FALSE;    {
3955      if (UCD_OTHERCASE(c) != next) break;
3956      next++;
3957    }    }
3958    
3959  /* Control does not reach here */  *odptr = next - 1;     /* End of othercase range */
3960    *cptr = c;             /* Rest of input range */
3961    return 0;
3962  }  }
3963    #endif  /* SUPPORT_UCP */
3964    
3965    
3966    
# Line 3418  switch(op_code) Line 3969  switch(op_code)
3969  *************************************************/  *************************************************/
3970    
3971  /* 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
3972  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
3973  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
3974  mutually recursive with the function immediately below.  mutually recursive with the function immediately below.
3975    
3976  Arguments:  Arguments:
3977    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
3978    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
3979    options       the options word    options       the options word
3980    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
3981    start         start of range character    start         start of range character
3982    end           end of range character    end           end of range character
3983    
3984  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
3985                  the pointer to extra data is updated                  the pointer to extra data is updated
3986  */  */
# Line 3441  add_to_class(pcre_uint8 *classbits, pcre Line 3992  add_to_class(pcre_uint8 *classbits, pcre
3992  pcre_uint32 c;  pcre_uint32 c;
3993  int n8 = 0;  int n8 = 0;
3994    
3995  /* If caseless matching is required, scan the range and process alternate  /* If caseless matching is required, scan the range and process alternate
3996  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
3997  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
3998  range. */  range. */
3999    
4000  if ((options & PCRE_CASELESS) != 0)  if ((options & PCRE_CASELESS) != 0)
4001    {    {
4002  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4003    if ((options & PCRE_UTF8) != 0)    if ((options & PCRE_UTF8) != 0)
4004      {      {
4005      int rc;      int rc;
4006      pcre_uint32 oc, od;      pcre_uint32 oc, od;
4007    
4008      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
4009      c = start;      c = start;
4010    
4011      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
4012        {        {
4013        /* Handle a single character that has more than one other case. */        /* Handle a single character that has more than one other case. */
4014    
4015        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4016          PRIV(ucd_caseless_sets) + rc, oc);          PRIV(ucd_caseless_sets) + rc, oc);
4017    
4018        /* Do nothing if the other case range is within the original range. */        /* Do nothing if the other case range is within the original range. */
4019    
4020        else if (oc >= start && od <= end) continue;        else if (oc >= start && od <= end) continue;
4021    
4022        /* 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
4023        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
4024        range. Otherwise, use a recursive call to add the additional range. */        range. Otherwise, use a recursive call to add the additional range. */
4025    
4026        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4027        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
4028        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 4032  if ((options & PCRE_CASELESS) != 0)
4032  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
4033    
4034    /* Not UTF-mode, or no UCP */    /* Not UTF-mode, or no UCP */
4035    
4036    for (c = start; c <= end && c < 256; c++)    for (c = start; c <= end && c < 256; c++)
4037      {      {
4038      SETBIT(classbits, cd->fcc[c]);      SETBIT(classbits, cd->fcc[c]);
4039      n8++;      n8++;
4040      }      }
4041    }    }
4042    
4043  /* 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
4044  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
4045  in all cases. */  in all cases. */
# Line 3514  if (end < 0x100) Line 4065  if (end < 0x100)
4065    {    {
4066    for (c = start; c <= end; c++)    for (c = start; c <= end; c++)
4067      {      {
4068      n8++;      n8++;
4069      SETBIT(classbits, c);      SETBIT(classbits, c);
4070      }      }
4071    }    }
4072    
4073  else  else
4074    {    {
4075    pcre_uchar *uchardata = *uchardptr;    pcre_uchar *uchardata = *uchardptr;
4076    
4077  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4078    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
4079      {      {
4080      if (start < end)      if (start < end)
4081        {        {
4082        *uchardata++ = XCL_RANGE;        *uchardata++ = XCL_RANGE;
4083        uchardata += PRIV(ord2utf)(start, uchardata);        uchardata += PRIV(ord2utf)(start, uchardata);
4084        uchardata += PRIV(ord2utf)(end, uchardata);        uchardata += PRIV(ord2utf)(end, uchardata);
4085        }        }
4086      else if (start == end)      else if (start == end)
4087        {        {
4088        *uchardata++ = XCL_SINGLE;        *uchardata++ = XCL_SINGLE;
4089        uchardata += PRIV(ord2utf)(start, uchardata);        uchardata += PRIV(ord2utf)(start, uchardata);
4090        }        }
4091      }      }
4092    else    else
4093  #endif  /* SUPPORT_UTF */  #endif  /* SUPPORT_UTF */
4094    
4095    /* Without UTF support, character values are constrained by the bit length,    /* Without UTF support, character values are constrained by the bit length,
4096    and can only be > 256 for 16-bit and 32-bit libraries. */    and can only be > 256 for 16-bit and 32-bit libraries. */
4097    
4098  #ifdef COMPILE_PCRE8  #ifdef COMPILE_PCRE8
4099      {}      {}
4100  #else  #else
4101    if (start < end)    if (start < end)
4102      {      {
4103      *uchardata++ = XCL_RANGE;      *uchardata++ = XCL_RANGE;
# Line 3557  else Line 4108  else
4108      {      {
4109      *uchardata++ = XCL_SINGLE;      *uchardata++ = XCL_SINGLE;
4110      *uchardata++ = start;      *uchardata++ = start;
4111      }      }
4112  #endif  #endif
4113    
4114    *uchardptr = uchardata;   /* Updata extra data pointer */    *uchardptr = uchardata;   /* Updata extra data pointer */
4115    }    }
4116    
4117  return n8;    /* Number of 8-bit characters */  return n8;    /* Number of 8-bit characters */
4118  }  }
4119    
4120    
4121    
4122    
4123  /*************************************************  /*************************************************
4124  *        Add a list of characters to a class     *  *        Add a list of characters to a class     *
4125  *************************************************/  *************************************************/
4126    
4127  /* 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
4128  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
4129  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
4130  handled appropriately. This function is mutually recursive with the function  handled appropriately. This function is mutually recursive with the function
# Line 3583  Arguments: Line 4134  Arguments:
4134    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4135    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4136    options       the options word    options       the options word
4137    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4138    p             points to row of 32-bit values, terminated by NOTACHAR    p             points to row of 32-bit values, terminated by NOTACHAR
4139    except        character to omit; this is used when adding lists of    except        character to omit; this is used when adding lists of
4140                    case-equivalent characters to avoid including the one we                    case-equivalent characters to avoid including the one we
4141                    already know about                    already know about
4142    
4143  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4144                  the pointer to extra data is updated                  the pointer to extra data is updated
4145  */  */
# Line 3602  while (p[0] < NOTACHAR) Line 4153  while (p[0] < NOTACHAR)
4153    {    {
4154    int n = 0;    int n = 0;
4155    if (p[0] != except)    if (p[0] != except)
4156      {      {
4157      while(p[n+1] == p[0] + n + 1) n++;      while(p[n+1] == p[0] + n + 1) n++;
4158      n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);      n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4159      }      }
4160    p += n + 1;    p += n + 1;
4161    }    }
4162  return n8;  return n8;
4163  }  }
4164    
4165    
4166    
# Line 3624  Arguments: Line 4175  Arguments:
4175    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4176    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4177    options       the options word    options       the options word
4178    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4179    p             points to row of 32-bit values, terminated by NOTACHAR    p             points to row of 32-bit values, terminated by NOTACHAR
4180    
4181  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4182                  the pointer to extra data is updated                  the pointer to extra data is updated
4183  */  */
4184    
4185  static int  static int
4186  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
4187    int options, compile_data *cd, const pcre_uint32 *p)    int options, compile_data *cd, const pcre_uint32 *p)
4188  {  {
4189  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
# Line 3644  while (p[0] < NOTACHAR) Line 4195  while (p[0] < NOTACHAR)
4195    while (p[1] == p[0] + 1) p++;    while (p[1] == p[0] + 1) p++;
4196    n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,    n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
4197      (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);      (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
4198    p++;    p++;
4199    }    }
4200  return n8;  return n8;
4201  }  }
4202    
4203    
4204    
# Line 3662  to find out the amount of memory needed, Line 4213  to find out the amount of memory needed,
4213  phase. The value of lengthptr distinguishes the two phases.  phase. The value of lengthptr distinguishes the two phases.
4214    
4215  Arguments:  Arguments:
4216    optionsptr     pointer to the option bits    optionsptr        pointer to the option bits
4217    codeptr        points to the pointer to the current code point    codeptr           points to the pointer to the current code point
4218    ptrptr         points to the current pattern pointer    ptrptr            points to the current pattern pointer
4219    errorcodeptr   points to error code variable    errorcodeptr      points to error code variable
4220    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr      place to put the first required character
4221    reqcharptr     set to the last literal character required, else < 0    firstcharflagsptr place to put the first character flags, or a negative number
4222    bcptr          points to current branch chain    reqcharptr        place to put the last required character
4223    cond_depth     conditional nesting depth    reqcharflagsptr   place to put the last required character flags, or a negative number
4224    cd             contains pointers to tables etc.    bcptr             points to current branch chain
4225    lengthptr      NULL during the real compile phase    cond_depth        conditional nesting depth
4226                   points to length accumulator during pre-compile phase    cd                contains pointers to tables etc.
4227      lengthptr         NULL during the real compile phase
4228                        points to length accumulator during pre-compile phase
4229    
4230  Returns:         TRUE on success  Returns:            TRUE on success
4231                   FALSE, with *errorcodeptr set non-zero on error                      FALSE, with *errorcodeptr set non-zero on error
4232  */  */
4233    
4234  static BOOL  static BOOL
4235  compile_branch(int *optionsptr, pcre_uchar **codeptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
4236    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,    const pcre_uchar **ptrptr, int *errorcodeptr,
4237    pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,    pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
4238      pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
4239      branch_chain *bcptr, int cond_depth,
4240    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
4241  {  {
4242  int repeat_type, op_type;  int repeat_type, op_type;
4243  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
4244  int bravalue = 0;  int bravalue = 0;
4245  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
4246  pcre_int32 firstchar, reqchar;  pcre_uint32 firstchar, reqchar;
4247  pcre_int32 zeroreqchar, zerofirstchar;  pcre_int32 firstcharflags, reqcharflags;
4248    pcre_uint32 zeroreqchar, zerofirstchar;
4249    pcre_int32 zeroreqcharflags, zerofirstcharflags;
4250  pcre_int32 req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
4251  int options = *optionsptr;               /* May change dynamically */  int options = *optionsptr;               /* May change dynamically */
4252  int after_manual_callout = 0;  int after_manual_callout = 0;
# Line 3717  dynamically as we process the pattern. * Line 4274  dynamically as we process the pattern. *
4274  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4275  /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */  /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
4276  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
4277    #ifndef COMPILE_PCRE32
4278  pcre_uchar utf_chars[6];  pcre_uchar utf_chars[6];
4279    #endif
4280  #else  #else
4281  BOOL utf = FALSE;  BOOL utf = FALSE;
4282  #endif  #endif
4283    
4284  /* Helper variables for OP_XCLASS opcode (for characters > 255). We define  /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
4285  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,
4286  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
4287  alternative calls for the different cases. */  alternative calls for the different cases. */
4288    
4289  pcre_uchar *class_uchardata;  pcre_uchar *class_uchardata;
# Line 3752  to take the zero repeat into account. Th Line 4311  to take the zero repeat into account. Th
4311  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
4312  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
4313    
4314  firstchar = reqchar = zerofirstchar = zeroreqchar = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
4315    firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
4316    
4317  /* The variable req_caseopt contains either the REQ_CASELESS value  /* The variable req_caseopt contains either the REQ_CASELESS value
4318  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 4338  for (;; ptr++)
4338    int recno;    int recno;
4339    int refsign;    int refsign;
4340    int skipbytes;    int skipbytes;
4341    int subreqchar;    pcre_uint32 subreqchar, subfirstchar;
4342    int subfirstchar;    pcre_int32 subreqcharflags, subfirstcharflags;
4343    int terminator;    int terminator;
4344    int mclength;    unsigned int mclength;
4345    int tempbracount;    unsigned int tempbracount;
4346    int ec; // FIXMEchpe pcre_uint32    pcre_uint32 ec;
4347    pcre_uchar mcbuffer[8];    pcre_uchar mcbuffer[8];
4348    
4349    /* Get next character in the pattern */    /* Get next character in the pattern */
# Line 3793  for (;; ptr++) Line 4353  for (;; ptr++)
4353    /* 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
4354    string. Nesting only happens one level deep. */    string. Nesting only happens one level deep. */
4355    
4356    if (c == 0 && nestptr != NULL)    if (c == CHAR_NULL && nestptr != NULL)
4357      {      {
4358      ptr = nestptr;      ptr = nestptr;
4359      nestptr = NULL;      nestptr = NULL;
# Line 3868  for (;; ptr++) Line 4428  for (;; ptr++)
4428    
4429    /* 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 */
4430    
4431    if (inescq && c != 0)    if (inescq && c != CHAR_NULL)
4432      {      {
4433      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4434        {        {
# Line 3893  for (;; ptr++) Line 4453  for (;; ptr++)
4453        }        }
4454      }      }
4455    
   /* Fill in length of a previous callout, except when the next thing is  
   a quantifier. */  
   
4456    is_quantifier =    is_quantifier =
4457      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4458      (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));      (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4459    
4460    if (!is_quantifier && previous_callout != NULL &&    /* Fill in length of a previous callout, except when the next thing is a
4461      quantifier or when processing a property substitution string in UCP mode. */
4462    
4463      if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4464         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
4465      {      {
4466      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 4476  for (;; ptr++)
4476      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
4477        {        {
4478        ptr++;        ptr++;
4479        while (*ptr != 0)        while (*ptr != CHAR_NULL)
4480          {          {
4481          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
4482          ptr++;          ptr++;
# Line 3924  for (;; ptr++) Line 4484  for (;; ptr++)
4484          if (utf) FORWARDCHAR(ptr);          if (utf) FORWARDCHAR(ptr);
4485  #endif  #endif
4486          }          }
4487        if (*ptr != 0) continue;        if (*ptr != CHAR_NULL) continue;
4488    
4489        /* Else fall through to handle end of string */        /* Else fall through to handle end of string */
4490        c = 0;        c = 0;
4491        }        }
4492      }      }
4493    
4494    /* No auto callout for quantifiers. */    /* No auto callout for quantifiers, or while processing property strings that
4495      are substituted for \w etc in UCP mode. */
4496    
4497    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4498      {      {
4499      previous_callout = code;      previous_callout = code;
4500      code = auto_callout(code, ptr, cd);      code = auto_callout(code, ptr, cd);
# Line 3946  for (;; ptr++) Line 4507  for (;; ptr++)
4507      case CHAR_VERTICAL_LINE:       /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
4508      case CHAR_RIGHT_PARENTHESIS:      case CHAR_RIGHT_PARENTHESIS:
4509      *firstcharptr = firstchar;      *firstcharptr = firstchar;
4510        *firstcharflagsptr = firstcharflags;
4511      *reqcharptr = reqchar;      *reqcharptr = reqchar;
4512        *reqcharflagsptr = reqcharflags;
4513      *codeptr = code;      *codeptr = code;
4514      *ptrptr = ptr;      *ptrptr = ptr;
4515      if (lengthptr != NULL)      if (lengthptr != NULL)
# Line 3970  for (;; ptr++) Line 4533  for (;; ptr++)
4533      previous = NULL;      previous = NULL;
4534      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
4535        {        {
4536        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;        if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4537        *code++ = OP_CIRCM;        *code++ = OP_CIRCM;
4538        }        }
4539      else *code++ = OP_CIRC;      else *code++ = OP_CIRC;
# Line 3985  for (;; ptr++) Line 4548  for (;; ptr++)
4548      repeats. The value of reqchar doesn't change either. */      repeats. The value of reqchar doesn't change either. */
4549    
4550      case CHAR_DOT:      case CHAR_DOT:
4551      if (firstchar == REQ_UNSET) firstchar = REQ_NONE;      if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4552      zerofirstchar = firstchar;      zerofirstchar = firstchar;
4553        zerofirstcharflags = firstcharflags;
4554      zeroreqchar = reqchar;      zeroreqchar = reqchar;
4555        zeroreqcharflags = reqcharflags;
4556      previous = code;      previous = code;
4557      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
4558      break;      break;
# Line 4061  for (;; ptr++) Line 4626  for (;; ptr++)
4626          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4627        {        {
4628        *code++ = negate_class? OP_ALLANY : OP_FAIL;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
4629        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;        if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4630        zerofirstchar = firstchar;        zerofirstchar = firstchar;
4631          zerofirstcharflags = firstcharflags;
4632        break;        break;
4633        }        }
4634    
# Line 4097  for (;; ptr++) Line 4663  for (;; ptr++)
4663      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
4664      loop, c contains the first byte of the character. */      loop, c contains the first byte of the character. */
4665    
4666      if (c != 0) do      if (c != CHAR_NULL) do
4667        {        {
4668        const pcre_uchar *oldptr;        const pcre_uchar *oldptr;
4669    
4670  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #ifdef SUPPORT_UTF
4671        if (utf && HAS_EXTRALEN(c))        if (utf && HAS_EXTRALEN(c))
4672          {                           /* Braces are required because the */          {                           /* Braces are required because the */
4673          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
# Line 4112  for (;; ptr++) Line 4678  for (;; ptr++)
4678        /* In the pre-compile phase, accumulate the length of any extra        /* In the pre-compile phase, accumulate the length of any extra
4679        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
4680        contain a zillion > 255 characters no longer overwrite the work space        contain a zillion > 255 characters no longer overwrite the work space
4681        (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,
4682        however. */        however. */
4683    
4684        if (lengthptr != NULL && class_uchardata > class_uchardata_base)        if (lengthptr != NULL && class_uchardata > class_uchardata_base)
# Line 4176  for (;; ptr++) Line 4742  for (;; ptr++)
4742          alpha. This relies on the fact that the class table starts with          alpha. This relies on the fact that the class table starts with
4743          alpha, lower, upper as the first 3 entries. */          alpha, lower, upper as the first 3 entries. */
4744    
4745          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
4746            posix_class = 0;            posix_class = 0;
4747    
4748          /* 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 4819  for (;; ptr++)
4819    
4820        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
4821          {          {
4822          escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options, TRUE);          escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options,
4823              TRUE);
4824          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
4825            if (escape == 0) c = ec;
         if (escape == 0)  
           c = ec;  
4826          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 */
4827          else if (escape == ESC_N)            /* \N is not supported in a class */          else if (escape == ESC_N)          /* \N is not supported in a class */
4828            {            {
4829            *errorcodeptr = ERR71;            *errorcodeptr = ERR71;
4830            goto FAILED;            goto FAILED;
# Line 4316  for (;; ptr++) Line 4880  for (;; ptr++)
4880              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
4881              continue;              continue;
4882    
4883              /* 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
4884              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
4885              class. Luckily, the value of CHAR_VT is 0x0b in both ASCII and              previously set by something earlier in the character class.
4886              EBCDIC, so we lazily just adjust the appropriate bit. */              Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
4887                we could just adjust the appropriate bit. From PCRE 8.34 we no
4888                longer treat \s and \S specially. */
4889    
4890              case ESC_s:              case ESC_s:
4891              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];  
4892              continue;              continue;
4893    
4894              case ESC_S:              case ESC_S:
4895              should_flip_negation = TRUE;              should_flip_negation = TRUE;
4896              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 */  
4897              continue;              continue;
4898    
4899              /* The rest apply in both UCP and non-UCP cases. */              /* The rest apply in both UCP and non-UCP cases. */
4900    
4901              case ESC_h:              case ESC_h:
4902              (void)add_list_to_class(classbits, &class_uchardata, options, cd,              (void)add_list_to_class(classbits, &class_uchardata, options, cd,
4903                PRIV(hspace_list), NOTACHAR);                PRIV(hspace_list), NOTACHAR);
4904              continue;              continue;
4905    
4906              case ESC_H:              case ESC_H:
4907              (void)add_not_list_to_class(classbits, &class_uchardata, options,              (void)add_not_list_to_class(classbits, &class_uchardata, options,
4908                cd, PRIV(hspace_list));                cd, PRIV(hspace_list));
4909              continue;              continue;
4910    
4911              case ESC_v:              case ESC_v:
4912              (void)add_list_to_class(classbits, &class_uchardata, options, cd,              (void)add_list_to_class(classbits, &class_uchardata, options, cd,
4913                PRIV(vspace_list), NOTACHAR);                PRIV(vspace_list), NOTACHAR);
4914              continue;              continue;
4915    
4916              case ESC_V:              case ESC_V:
4917              (void)add_not_list_to_class(classbits, &class_uchardata, options,              (void)add_not_list_to_class(classbits, &class_uchardata, options,
4918                cd, PRIV(vspace_list));                cd, PRIV(vspace_list));
4919              continue;              continue;
4920    
4921  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
# Line 4360  for (;; ptr++) Line 4923  for (;; ptr++)
4923              case ESC_P:              case ESC_P:
4924                {                {
4925                BOOL negated;                BOOL negated;
4926                int pdata;                unsigned int ptype = 0, pdata = 0;
4927                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                if (!get_ucp(&ptr, &negated, &ptype, &pdata, errorcodeptr))
4928                if (ptype < 0) goto FAILED;                  goto FAILED;
4929                *class_uchardata++ = ((escape == ESC_p) != negated)?                *class_uchardata++ = ((escape == ESC_p) != negated)?
4930                  XCL_PROP : XCL_NOTPROP;                  XCL_PROP : XCL_NOTPROP;
4931                *class_uchardata++ = ptype;                *class_uchardata++ = ptype;
# Line 4390  for (;; ptr++) Line 4953  for (;; ptr++)
4953    
4954          /* Fall through if the escape just defined a single character (c >= 0).          /* Fall through if the escape just defined a single character (c >= 0).
4955          This may be greater than 256. */          This may be greater than 256. */
4956    
4957          escape = 0;          escape = 0;
4958    
4959          }   /* End of backslash handling */          }   /* End of backslash handling */
# Line 4416  for (;; ptr++) Line 4979  for (;; ptr++)
4979    
4980        if (!inescq && ptr[1] == CHAR_MINUS)        if (!inescq && ptr[1] == CHAR_MINUS)
4981          {          {
4982          int d;          pcre_uint32 d;
4983          ptr += 2;          ptr += 2;
4984          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
4985    
# Line 4431  for (;; ptr++) Line 4994  for (;; ptr++)
4994            inescq = TRUE;            inescq = TRUE;
4995            break;            break;
4996            }            }
4997    
4998          /* 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
4999          back the pointer and jump to handle the character that preceded it. */          back the pointer and jump to handle the character that preceded it. */
5000    
5001          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))          if (*ptr == CHAR_NULL || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
5002            {            {
5003            ptr = oldptr;            ptr = oldptr;
5004            goto CLASS_SINGLE_CHARACTER;            goto CLASS_SINGLE_CHARACTER;
5005            }            }
5006    
5007          /* Otherwise, we have a potential range; pick up the next character */          /* Otherwise, we have a potential range; pick up the next character */
5008    
5009  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
# Line 4487  for (;; ptr++) Line 5050  for (;; ptr++)
5050          /* We have found a character range, so single character optimizations          /* We have found a character range, so single character optimizations
5051          cannot be done anymore. Any value greater than 1 indicates that there          cannot be done anymore. Any value greater than 1 indicates that there
5052          is more than one character. */          is more than one character. */
5053    
5054          class_one_char = 2;          class_one_char = 2;
5055    
5056          /* 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. */
5057    
5058          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
5059    
5060          class_has_8bitchar +=          class_has_8bitchar +=
5061            add_to_class(classbits, &class_uchardata, options, cd, c, d);            add_to_class(classbits, &class_uchardata, options, cd, c, d);
5062    
5063          continue;   /* Go get the next char in the class */          continue;   /* Go get the next char in the class */
5064          }          }
5065    
5066        /* 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
5067        char, or after \ that introduces a single character or for an apparent        char, or after \ that introduces a single character or for an apparent
5068        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
5069        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
5070        with a zillion characters in it. */        with a zillion characters in it. */
5071    
5072        CLASS_SINGLE_CHARACTER:        CLASS_SINGLE_CHARACTER:
# Line 4522  for (;; ptr++) Line 5085  for (;; ptr++)
5085          {          {
5086          ptr++;          ptr++;
5087          zeroreqchar = reqchar;          zeroreqchar = reqchar;
5088            zeroreqcharflags = reqcharflags;
5089    
5090          if (negate_class)          if (negate_class)
5091            {            {
5092  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
             // FIXMEchpe pcreuint32?  
5093            int d;            int d;
5094  #endif  #endif
5095            if (firstchar == REQ_UNSET) firstchar = REQ_NONE;            if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5096            zerofirstchar = firstchar;            zerofirstchar = firstchar;
5097              zerofirstcharflags = firstcharflags;
5098    
5099            /* For caseless UTF-8 mode when UCP support is available, check            /* For caseless UTF-8 mode when UCP support is available, check
5100            whether this character has more than one other case. If so, generate            whether this character has more than one other case. If so, generate
5101            a special OP_NOTPROP item instead of OP_NOTI. */            a special OP_NOTPROP item instead of OP_NOTI. */
5102    
5103  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
5104            if (utf && (options & PCRE_CASELESS) != 0 &&            if (utf && (options & PCRE_CASELESS) != 0 &&
5105                (d = UCD_CASESET(c)) != 0)                (d = UCD_CASESET(c)) != 0)
5106              {              {
5107              *code++ = OP_NOTPROP;              *code++ = OP_NOTPROP;
5108              *code++ = PT_CLIST;              *code++ = PT_CLIST;
5109              *code++ = d;              *code++ = d;
5110              }              }
5111            else            else
5112  #endif  #endif
5113            /* Char has only one other case, or UCP not available */            /* Char has only one other case, or UCP not available */
5114    
# Line 4557  for (;; ptr++) Line 5121  for (;; ptr++)
5121  #endif  #endif
5122                *code++ = c;                *code++ = c;
5123              }              }
5124    
5125            /* We are finished with this character class */            /* We are finished with this character class */
5126    
5127            goto END_CLASS;            goto END_CLASS;
5128            }            }
5129    
# Line 4577  for (;; ptr++) Line 5141  for (;; ptr++)
5141            }            }
5142          goto ONE_CHAR;          goto ONE_CHAR;
5143          }       /* End of 1-char optimization */          }       /* End of 1-char optimization */
5144    
5145        /* 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
5146        has been generated. Add this character to the class. */        has been generated. Add this character to the class. */
5147    
5148        class_has_8bitchar +=        class_has_8bitchar +=
5149          add_to_class(classbits, &class_uchardata, options, cd, c, c);          add_to_class(classbits, &class_uchardata, options, cd, c, c);
5150        }        }
5151    
# Line 4589  for (;; ptr++) Line 5153  for (;; ptr++)
5153      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
5154      string. */      string. */
5155    
5156      while (((c = *(++ptr)) != 0 ||      while (((c = *(++ptr)) != CHAR_NULL ||
5157             (nestptr != NULL &&             (nestptr != NULL &&
5158               (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&               (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != CHAR_NULL)) &&
5159             (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));             (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
5160    
5161      /* Check for missing terminating ']' */      /* Check for missing terminating ']' */
5162    
5163      if (c == 0)      if (c == CHAR_NULL)
5164        {        {
5165        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
5166        goto FAILED;        goto FAILED;
5167        }        }
5168    
5169      /* 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
5170      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
5171      phase, class_uchardata gets emptied to prevent workspace overflow, so it      phase, class_uchardata gets emptied to prevent workspace overflow, so it
5172      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
5173      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
5174      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
5175      instead of just doing a test on class_uchardata below. */      instead of just doing a test on class_uchardata below. */
5176    
5177  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5178      if (class_uchardata > class_uchardata_base) xclass = TRUE;      if (class_uchardata > class_uchardata_base) xclass = TRUE;
5179  #endif  #endif
# Line 4618  for (;; ptr++) Line 5182  for (;; ptr++)
5182      setting, whatever the repeat count. Any reqchar setting must remain      setting, whatever the repeat count. Any reqchar setting must remain
5183      unchanged after any kind of repeat. */      unchanged after any kind of repeat. */
5184    
5185      if (firstchar == REQ_UNSET) firstchar = REQ_NONE;      if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5186      zerofirstchar = firstchar;      zerofirstchar = firstchar;
5187        zerofirstcharflags = firstcharflags;
5188      zeroreqchar = reqchar;      zeroreqchar = reqchar;
5189        zeroreqcharflags = reqcharflags;
5190    
5191      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
5192      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 5242  for (;; ptr++)
5242        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
5243        }        }
5244      code += 32 / sizeof(pcre_uchar);      code += 32 / sizeof(pcre_uchar);
5245    
5246      END_CLASS:      END_CLASS:
5247      break;      break;
5248    
# Line 4715  for (;; ptr++) Line 5281  for (;; ptr++)
5281      if (repeat_min == 0)      if (repeat_min == 0)
5282        {        {
5283        firstchar = zerofirstchar;    /* Adjust for zero repeat */        firstchar = zerofirstchar;    /* Adjust for zero repeat */
5284          firstcharflags = zerofirstcharflags;
5285        reqchar = zeroreqchar;        /* Ditto */        reqchar = zeroreqchar;        /* Ditto */
5286          reqcharflags = zeroreqcharflags;
5287        }        }
5288    
5289      /* Remember whether this is a variable length repeat */      /* Remember whether this is a variable length repeat */
# Line 4818  for (;; ptr++) Line 5386  for (;; ptr++)
5386          {          {
5387          c = code[-1];          c = code[-1];
5388          if (*previous <= OP_CHARI && repeat_min > 1)          if (*previous <= OP_CHARI && repeat_min > 1)
5389            reqchar = c | req_caseopt | cd->req_varyopt;            {
5390          }            reqchar = c;
5391              reqcharflags = req_caseopt | cd->req_varyopt;
5392        /* 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;  
5393          }          }
5394    
5395        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 5409  for (;; ptr++)
5409        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
5410        c = *previous;        c = *previous;
5411    
       if (!possessive_quantifier &&  
           repeat_max < 0 &&  
           check_auto_possessive(previous, utf, ptr + 1, options, cd))  
         {  
         repeat_type = 0;    /* Force greedy */  
         possessive_quantifier = TRUE;  
         }  
   
5412        OUTPUT_SINGLE_REPEAT:        OUTPUT_SINGLE_REPEAT:
5413        if (*previous == OP_PROP || *previous == OP_NOTPROP)        if (*previous == OP_PROP || *previous == OP_NOTPROP)
5414          {          {
# Line 4875  for (;; ptr++) Line 5425  for (;; ptr++)
5425    
5426        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
5427    
       /*--------------------------------------------------------------------*/  
       /* This code is obsolete from release 8.00; the restriction was finally  
       removed: */  
   
       /* All real repeats make it impossible to handle partial matching (maybe  
       one day we will be able to remove this restriction). */  
   
       /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */  
       /*--------------------------------------------------------------------*/  
   
5428        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
5429    
5430        repeat_type += op_type;        repeat_type += op_type;
# Line 5017  for (;; ptr++) Line 5557  for (;; ptr++)
5557      /* If previous was a character class or a back reference, we put the repeat      /* If previous was a character class or a back reference, we put the repeat
5558      stuff after it, but just skip the item if the repeat was {0,0}. */      stuff after it, but just skip the item if the repeat was {0,0}. */
5559    
5560      else if (*previous == OP_CLASS ||      else if (*previous == OP_CLASS || *previous == OP_NCLASS ||
              *previous == OP_NCLASS ||  
5561  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8