/[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 964 by ph10, Fri May 4 13:03:39 2012 UTC revision 1382 by zherczeg, Fri Oct 18 07:55:07 2013 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2012 University of Cambridge             Copyright (c) 1997-2013 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When PCRE_DEBUG is defined, we need the pcre(16)_printint() function, which  /* When PCRE_DEBUG is defined, we need the pcre(16|32)_printint() function, which
57  is also used by pcretest. PCRE_DEBUG is not defined when building a production  is also used by pcretest. PCRE_DEBUG is not defined when building a production
58  library. We do not need to select pcre16_printint.c specially, because the  library. We do not need to select pcre16_printint.c specially, because the
59  COMPILE_PCREx macro will already be appropriately set. */  COMPILE_PCREx macro will already be appropriately set. */
# Line 68  COMPILE_PCREx macro will already be appr Line 68  COMPILE_PCREx macro will already be appr
68    
69  /* Macro for setting individual bits in class bitmaps. */  /* Macro for setting individual bits in class bitmaps. */
70    
71  #define SETBIT(a,b) a[b/8] |= (1 << (b%8))  #define SETBIT(a,b) a[(b)/8] |= (1 << ((b)&7))
72    
73  /* Maximum length value to check against when making sure that the integer that  /* Maximum length value to check against when making sure that the integer that
74  holds the compiled pattern length does not overflow. We make it a bit less than  holds the compiled pattern length does not overflow. We make it a bit less than
# Line 77  to check them every time. */ Line 77  to check them every time. */
77    
78  #define OFLOW_MAX (INT_MAX - 20)  #define OFLOW_MAX (INT_MAX - 20)
79    
80    /* Definitions to allow mutual recursion */
81    
82    static int
83      add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84        const pcre_uint32 *, unsigned int);
85    
86    static BOOL
87      compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
88        pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
89        compile_data *, int *);
90    
91    
92    
93  /*************************************************  /*************************************************
94  *      Code parameters and static tables         *  *      Code parameters and static tables         *
# Line 103  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 110  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    /* 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 440  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 472  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 490  static const char error_texts[] = Line 512  static const char error_texts[] =
512    "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"    "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
513    "invalid UTF-16 string\0"    "invalid UTF-16 string\0"
514    /* 75 */    /* 75 */
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"
517      "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 630  static const pcre_uint8 ebcdic_chartab[] Line 659  static const pcre_uint8 ebcdic_chartab[]
659  #endif  #endif
660    
661    
662  /* Definition to allow mutual recursion */  /* 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  static BOOL  
700    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,  /* This table is used to check whether auto-possessification is possible
701      int *, int *, branch_chain *, compile_data *, int *);  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    
# Line 657  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 741  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 a negative value which  positive value for a simple escape such as \n, or 0 for a data character which
945  encodes one of the more complicated things such as \d. A backreference to group  will be placed in chptr. A backreference to group n is returned as negative n.
946  n is returned as -(ESC_REF + n); ESC_REF is the highest ESC_xxx macro. When  When UTF-8 is enabled, a positive value greater than 255 may be returned in
947  UTF-8 is enabled, a positive value greater than 255 may be returned. On entry,  chptr. On entry, ptr is pointing at the \. On exit, it is on the final
948  ptr is pointing at the \. On exit, it is on the final character of the escape  character of the escape sequence.
 sequence.  
949    
950  Arguments:  Arguments:
951    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
952      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
956    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
957    
958  Returns:         zero or positive => a data character  Returns:         zero => a data character
959                   negative => a special escape sequence                   positive => a special escape sequence
960                     negative => a back reference
961                   on error, errorcodeptr is set                   on error, errorcodeptr is set
962  */  */
963    
964  static int  static int
965  check_escape(const pcre_uchar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
966    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. */
969  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
970  const pcre_uchar *ptr = *ptrptr + 1;  const pcre_uchar *ptr = *ptrptr + 1;
971  pcre_int32 c;  pcre_uint32 c;
972    int escape = 0;
973  int i;  int i;
974    
975  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
# Line 775  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 784  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) c = 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 */
994  else if (c < 'a' || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}  else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
995  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
996  #endif  #endif
997    
998  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
# Line 797  else if ((i = escapes[c - 0x48]) != 0) Line 1000  else if ((i = escapes[c - 0x48]) != 0)
1000  else  else
1001    {    {
1002    const pcre_uchar *oldptr;    const pcre_uchar *oldptr;
1003    BOOL braced, negated;    BOOL braced, negated, overflow;
1004      int s;
1005    
1006    switch (c)    switch (c)
1007      {      {
# Line 822  else Line 1026  else
1026          c = 0;          c = 0;
1027          for (i = 0; i < 4; ++i)          for (i = 0; i < 4; ++i)
1028            {            {
1029            register int cc = *(++ptr);            register pcre_uint32 cc = *(++ptr);
1030  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1031            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1032            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
# Line 831  else Line 1035  else
1035            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1036  #endif  #endif
1037            }            }
1038    
1039    #if defined COMPILE_PCRE8
1040            if (c > (utf ? 0x10ffffU : 0xffU))
1041    #elif defined COMPILE_PCRE16
1042            if (c > (utf ? 0x10ffffU : 0xffffU))
1043    #elif defined COMPILE_PCRE32
1044            if (utf && c > 0x10ffffU)
1045    #endif
1046              {
1047              *errorcodeptr = ERR76;
1048              }
1049            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1050          }          }
1051        }        }
1052      else      else
# Line 857  else Line 1073  else
1073      (3) For Oniguruma compatibility we also support \g followed by a name or a      (3) For Oniguruma compatibility we also support \g followed by a name or a
1074      number either in angle brackets or in single quotes. However, these are      number either in angle brackets or in single quotes. However, these are
1075      (possibly recursive) subroutine calls, _not_ backreferences. Just return      (possibly recursive) subroutine calls, _not_ backreferences. Just return
1076      the -ESC_g code (cf \k). */      the ESC_g code (cf \k). */
1077    
1078      case CHAR_g:      case CHAR_g:
1079      if (isclass) break;      if (isclass) break;
1080      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1081        {        {
1082        c = -ESC_g;        escape = ESC_g;
1083        break;        break;
1084        }        }
1085    
# Line 872  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          c = -ESC_k;          escape = ESC_k;
1096          break;          break;
1097          }          }
1098        braced = TRUE;        braced = TRUE;
# Line 892  else Line 1108  else
1108      else negated = FALSE;      else negated = FALSE;
1109    
1110      /* The integer range is limited by the machine's int representation. */      /* The integer range is limited by the machine's int representation. */
1111      c = 0;      s = 0;
1112        overflow = FALSE;
1113      while (IS_DIGIT(ptr[1]))      while (IS_DIGIT(ptr[1]))
1114        {        {
1115        if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */        if (s > INT_MAX / 10 - 1) /* Integer overflow */
1116          {          {
1117          c = -1;          overflow = TRUE;
1118          break;          break;
1119          }          }
1120        c = c * 10 + *(++ptr) - CHAR_0;        s = s * 10 + (int)(*(++ptr) - CHAR_0);
1121        }        }
1122      if (((unsigned int)c) > INT_MAX) /* Integer overflow */      if (overflow) /* Integer overflow */
1123        {        {
1124        while (IS_DIGIT(ptr[1]))        while (IS_DIGIT(ptr[1]))
1125          ptr++;          ptr++;
# Line 916  else Line 1133  else
1133        break;        break;
1134        }        }
1135    
1136      if (c == 0)      if (s == 0)
1137        {        {
1138        *errorcodeptr = ERR58;        *errorcodeptr = ERR58;
1139        break;        break;
# Line 924  else Line 1141  else
1141    
1142      if (negated)      if (negated)
1143        {        {
1144        if (c > bracount)        if (s > bracount)
1145          {          {
1146          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
1147          break;          break;
1148          }          }
1149        c = bracount - (c - 1);        s = bracount - (s - 1);
1150        }        }
1151    
1152      c = -(ESC_REF + c);      escape = -s;
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 954  else Line 1175  else
1175        {        {
1176        oldptr = ptr;        oldptr = ptr;
1177        /* The integer range is limited by the machine's int representation. */        /* The integer range is limited by the machine's int representation. */
1178        c -= CHAR_0;        s = (int)(c -CHAR_0);
1179          overflow = FALSE;
1180        while (IS_DIGIT(ptr[1]))        while (IS_DIGIT(ptr[1]))
1181          {          {
1182          if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */          if (s > INT_MAX / 10 - 1) /* Integer overflow */
1183            {            {
1184            c = -1;            overflow = TRUE;
1185            break;            break;
1186            }            }
1187          c = c * 10 + *(++ptr) - CHAR_0;          s = s * 10 + (int)(*(++ptr) - CHAR_0);
1188          }          }
1189        if (((unsigned int)c) > INT_MAX) /* Integer overflow */        if (overflow) /* Integer overflow */
1190          {          {
1191          while (IS_DIGIT(ptr[1]))          while (IS_DIGIT(ptr[1]))
1192            ptr++;            ptr++;
1193          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1194          break;          break;
1195          }          }
1196        if (c < 10 || c <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1197          {          {
1198          c = -(ESC_REF + c);          escape = -s;
1199          break;          break;
1200          }          }
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 1005  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          {          {
1275          c = 0;          c = 0;
1276          for (i = 0; i < 2; ++i)          for (i = 0; i < 2; ++i)
1277            {            {
1278            register int cc = *(++ptr);            register pcre_uint32 cc = *(++ptr);
1279  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1280            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1281            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
# Line 1030  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        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)        {
1299          if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1300          {          {
1301          register int 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
1310              if (c >= 0x10000000l) { overflow = TRUE; break; }
1311    #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  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
1322          if (c > (utf ? 0x10ffff : 0xff)) { c = -1; break; }            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1323  #else  #elif defined COMPILE_PCRE16
1324  #ifdef COMPILE_PCRE16            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1325          if (c > (utf ? 0x10ffff : 0xffff)) { c = -1; break; }  #elif defined COMPILE_PCRE32
1326              if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1327  #endif  #endif
1328  #endif            }
         }  
1329    
1330        if (c < 0)          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        int 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 1101  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 1141  else Line 1414  else
1414  newline". PCRE does not support \N{name}. However, it does support  newline". PCRE does not support \N{name}. However, it does support
1415  quantification such as \N{2,3}. */  quantification such as \N{2,3}. */
1416    
1417  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&  if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1418       !is_counted_repeat(ptr+2))       !is_counted_repeat(ptr+2))
1419    *errorcodeptr = ERR37;    *errorcodeptr = ERR37;
1420    
1421  /* If PCRE_UCP is set, we change the values for \d etc. */  /* If PCRE_UCP is set, we change the values for \d etc. */
1422    
1423  if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)  if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1424    c -= (ESC_DU - ESC_D);    escape += (ESC_DU - ESC_D);
1425    
1426  /* Set the pointer to the final character before returning. */  /* Set the pointer to the final character before returning. */
1427    
1428  *ptrptr = ptr;  *ptrptr = ptr;
1429  return c;  *chptr = c;
1430    return escape;
1431  }  }
1432    
1433    
# Line 1171  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  int c, i, bot, top;  pcre_uchar c;
1460    int i, bot, top;
1461  const pcre_uchar *ptr = *ptrptr;  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 1202  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 1227  top = PRIV(utt_size); Line 1504  top = PRIV(utt_size);
1504    
1505  while (bot < top)  while (bot < top)
1506    {    {
1507      int r;
1508    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1509    c = 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 (c == 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 (c > 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 1279  int max = -1; Line 1557  int max = -1;
1557  /* Read the minimum value and do a paranoid check: a negative value indicates  /* Read the minimum value and do a paranoid check: a negative value indicates
1558  an integer overflow. */  an integer overflow. */
1559    
1560  while (IS_DIGIT(*p)) min = min * 10 + *p++ - CHAR_0;  while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);
1561  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1562    {    {
1563    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 1294  if (*p == CHAR_RIGHT_CURLY_BRACKET) max Line 1572  if (*p == CHAR_RIGHT_CURLY_BRACKET) max
1572    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1573      {      {
1574      max = 0;      max = 0;
1575      while(IS_DIGIT(*p)) max = max * 10 + *p++ - CHAR_0;      while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);
1576      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1577        {        {
1578        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 1319  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 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)  
       {  
       int 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 == ptr - thisname &&  
           STRNCMP_UC_UC(name, thisname, 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 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 1652  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 1668  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 1686  and doing the check at the end; a flag s Line 1668  and doing the check at the end; a flag s
1668    
1669  Arguments:  Arguments:
1670    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1671    utf      TRUE in UTF-8 / UTF-16 mode    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1672    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1673    cd       the "compile data" structure    cd       the "compile data" structure
1674    
# Line 1712  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 1792  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 1832  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  #ifdef SUPPORT_UTF  #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]);
# Line 1841  for (;;) Line 1823  for (;;)
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) cc += 2;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1827          cc += 2;
1828      cc += 1 + IMM2_SIZE + 1;      cc += 1 + IMM2_SIZE + 1;
1829      break;      break;
1830    
# Line 1876  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  
     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 1966  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 2002  for (;;) Line 1995  for (;;)
1995    
1996    
1997    
   
1998  /*************************************************  /*************************************************
1999  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
2000  *************************************************/  *************************************************/
# Line 2015  length. Line 2007  length.
2007    
2008  Arguments:  Arguments:
2009    code        points to start of expression    code        points to start of expression
2010    utf         TRUE in UTF-8 / UTF-16 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2011    number      the required bracket number or negative to find a lookbehind    number      the required bracket number or negative to find a lookbehind
2012    
2013  Returns:      pointer to the opcode for the bracket, or NULL if not found  Returns:      pointer to the opcode for the bracket, or NULL if not found
# Line 2026  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 2049  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 2079  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        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2075          || code[1 + IMM2_SIZE] == OP_NOTPROP) 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 2102  for (;;) Line 2091  for (;;)
2091    a multi-byte character. The length in the table is a minimum, so we have to    a multi-byte character. The length in the table is a minimum, so we have to
2092    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2093    
2094  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2095      if (utf) switch(c)      if (utf) switch(c)
2096        {        {
2097        case OP_CHAR:        case OP_CHAR:
# Line 2154  instance of OP_RECURSE. Line 2143  instance of OP_RECURSE.
2143    
2144  Arguments:  Arguments:
2145    code        points to start of expression    code        points to start of expression
2146    utf         TRUE in UTF-8 / UTF-16 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2147    
2148  Returns:      pointer to the opcode for OP_RECURSE, or NULL if not found  Returns:      pointer to the opcode for OP_RECURSE, or NULL if not found
2149  */  */
# Line 2164  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 2199  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        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2192          || code[1 + IMM2_SIZE] == OP_NOTPROP) 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 2222  for (;;) Line 2208  for (;;)
2208      by a multi-byte character. The length in the table is a minimum, so we have      by a multi-byte character. The length in the table is a minimum, so we have
2209      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2210    
2211  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2212      if (utf) switch(c)      if (utf) switch(c)
2213        {        {
2214        case OP_CHAR:        case OP_CHAR:
# Line 2308  bracket whose current branch will alread Line 2294  bracket whose current branch will alread
2294  Arguments:  Arguments:
2295    code        points to start of search    code        points to start of search
2296    endcode     points to where to stop    endcode     points to where to stop
2297    utf         TRUE if in UTF-8 / UTF-16 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 2346  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 reference; scan the referenced group, remembering it on the
2378        stack chain to detect mutual recursions. */
2379    
2380      /* Completed backwards reference */      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 2420  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 2462  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 2478  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 2525  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      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2575        || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;        code += 2;
2576      break;      break;
2577    
2578      /* End of branch */      /* End of branch */
# Line 2539  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  #ifdef SUPPORT_UTF  #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 2573  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 2607  Arguments: Line 2678  Arguments:
2678    code        points to start of the recursion    code        points to start of the recursion
2679    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2680    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2681    utf         TRUE if in UTF-8 / UTF-16 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2682    cd          pointers to tables etc    cd          pointers to tables etc
2683    
2684  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
# Line 2619  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 2629  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 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        code += IMM2_SIZE;
2853    
2854      list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2855    
2856      switch(base)
2857        {
2858        case OP_STAR:
2859        list[0] = OP_CHAR;
2860        break;
2861    
2862        case OP_STARI:
2863        list[0] = OP_CHARI;
2864        break;
2865    
2866        case OP_NOTSTAR:
2867        list[0] = OP_NOT;
2868        break;
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    switch(c)
2883      {
2884      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      case OP_CHAR:
2905      case OP_NOT:
2906      GETCHARINCTEST(chr, code);
2907      list[2] = chr;
2908      list[3] = NOTACHAR;
2909      return code;
2910    
2911      case OP_CHARI:
2912      case OP_NOTI:
2913      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2914      GETCHARINCTEST(chr, code);
2915      list[2] = chr;
2916    
2917    #ifdef SUPPORT_UCP
2918      if (chr < 128 || (chr < 256 && !utf))
2919        list[3] = fcc[chr];
2920      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      /* The othercase might be the same value. */
2929    
2930      if (chr == list[3])
2931        list[3] = NOTACHAR;
2932      else
2933        list[4] = NOTACHAR;
2934      return code;
2935    
2936    #ifdef SUPPORT_UCP
2937      case OP_PROP:
2938      case OP_NOTPROP:
2939      if (code[0] != PT_CLIST)
2940        {
2941        list[2] = code[0];
2942        list[3] = code[1];
2943        return code + 2;
2944        }
2945    
2946      /* Convert only if we have enough space. */
2947    
2948      clist_src = PRIV(ucd_caseless_sets) + code[1];
2949      clist_dest = list + 2;
2950      code += 2;
2951    
2952      do {
2953         if (clist_dest >= list + 8)
2954           {
2955           /* 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      /* All characters are stored. The terminating NOTACHAR
2966      is copied form the clist itself. */
2967    
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) - 1;
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    *    Scan further character sets for match       *
3018    *************************************************/
3019    
3020    /* Checks whether the base and the current opcode have a common character, in
3021    which case the base cannot be possessified.
3022    
3023    Arguments:
3024      code        points to the byte code
3025      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3026      cd          static compile data
3027      base_list   the data list of the base opcode
3028    
3029    Returns:      TRUE if the auto-possessification is possible
3030    */
3031    
3032    static BOOL
3033    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    pcre_uchar c;
3037    pcre_uint32 list[8];
3038    const pcre_uint32 *chr_ptr;
3039    const pcre_uint32 *ochr_ptr;
3040    const pcre_uint32 *list_ptr;
3041    const pcre_uchar *next_code;
3042    const pcre_uint8 *class_bitset;
3043    const pcre_uint32 *set1, *set2, *set_end;
3044    pcre_uint32 chr;
3045    BOOL accepted, invert_bits;
3046    
3047    /* Note: the base_list[1] contains whether the current opcode has greedy
3048    (represented by a non-zero value) quantifier. This is a different from
3049    other character type lists, which stores here that the character iterator
3050    matches to an empty string (also represented by a non-zero value). */
3051    
3052    for(;;)
3053      {
3054      c = *code;
3055    
3056      /* Skip over callouts */
3057    
3058      if (c == OP_CALLOUT)
3059        {
3060        code += PRIV(OP_lengths)[c];
3061        continue;
3062        }
3063    
3064      if (c == OP_ALT)
3065        {
3066        do code += GET(code, 1); while (*code == OP_ALT);
3067        c = *code;
3068        }
3069    
3070      switch(c)
3071        {
3072        case OP_END:
3073        case OP_KETRPOS:
3074        /* TRUE only in greedy case. The non-greedy case could be replaced by
3075        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3076        uses more memory, which we cannot get at this stage.) */
3077    
3078        return base_list[1] != 0;
3079    
3080        case OP_KET:
3081        /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3082        it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3083        cannot be converted to a possessive form. */
3084    
3085        if (base_list[1] == 0) return FALSE;
3086    
3087        switch(*(code - GET(code, 1)))
3088          {
3089          case OP_ASSERT:
3090          case OP_ASSERT_NOT:
3091          case OP_ASSERTBACK:
3092          case OP_ASSERTBACK_NOT:
3093          case OP_ONCE:
3094          case OP_ONCE_NC:
3095          /* Atomic sub-patterns and assertions can always auto-possessify their
3096          last iterator. */
3097          return TRUE;
3098          }
3099    
3100        code += PRIV(OP_lengths)[c];
3101        continue;
3102    
3103        case OP_ONCE:
3104        case OP_ONCE_NC:
3105        case OP_BRA:
3106        case OP_CBRA:
3107        next_code = code;
3108        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3109    
3110        /* We do not support repeated brackets, because they can lead to
3111        infinite recursion. */
3112    
3113        if (*next_code != OP_KET) return FALSE;
3114    
3115        next_code = code + GET(code, 1);
3116        code += PRIV(OP_lengths)[c];
3117    
3118        while (*next_code == OP_ALT)
3119          {
3120          if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3121          code = next_code + 1 + LINK_SIZE;
3122          next_code += GET(next_code, 1);
3123          }
3124        continue;
3125    
3126        case OP_BRAZERO:
3127        case OP_BRAMINZERO:
3128    
3129        next_code = code + 1;
3130        if (*next_code != OP_BRA && *next_code != OP_CBRA)
3131          return FALSE;
3132    
3133        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3134    
3135        /* We do not support repeated brackets, because they can lead to
3136        infinite recursion. */
3137        if (*next_code != OP_KET) return FALSE;
3138    
3139        /* The bracket content will be checked by the
3140        OP_BRA/OP_CBRA case above. */
3141        next_code += 1 + LINK_SIZE;
3142        if (!compare_opcodes(next_code, utf, cd, base_list, base_end)) return FALSE;
3143    
3144        code += PRIV(OP_lengths)[c];
3145        continue;
3146        }
3147    
3148      /* Check for a supported opcode, and load its properties. */
3149    
3150      code = get_chr_property_list(code, utf, cd->fcc, list);
3151      if (code == NULL) return FALSE;    /* Unsupported */
3152    
3153      /* If either opcode is a small character list, set pointers for comparing
3154      characters from that list with another list, or with a property. */
3155    
3156      if (base_list[0] == OP_CHAR)
3157        {
3158        chr_ptr = base_list + 2;
3159        list_ptr = list;
3160        }
3161      else if (list[0] == OP_CHAR)
3162        {
3163        chr_ptr = list + 2;
3164        list_ptr = base_list;
3165        }
3166    
3167      /* Character bitsets can also be compared to certain opcodes. */
3168    
3169      else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3170    #ifdef COMPILE_PCRE8
3171          /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3172          || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3173    #endif
3174          )
3175        {
3176    #ifdef COMPILE_PCRE8
3177        if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3178    #else
3179        if (base_list[0] == OP_CLASS)
3180    #endif
3181          {
3182          set1 = (pcre_uint32 *)(base_end - base_list[2]);
3183          list_ptr = list;
3184          }
3185        else
3186          {
3187          set1 = (pcre_uint32 *)(code - list[2]);
3188          list_ptr = base_list;
3189          }
3190    
3191        invert_bits = FALSE;
3192        switch(list_ptr[0])
3193          {
3194          case OP_CLASS:
3195          case OP_NCLASS:
3196          set2 = (pcre_uint32 *)
3197            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3198          break;
3199    
3200          /* OP_XCLASS cannot be supported here, because its bitset
3201          is not necessarily complete. E.g: [a-\0x{200}] is stored
3202          as a character range, and the appropriate bits are not set. */
3203    
3204          case OP_NOT_DIGIT:
3205            invert_bits = TRUE;
3206            /* Fall through */
3207          case OP_DIGIT:
3208            set2 = (pcre_uint32 *)(cd->cbits + cbit_digit);
3209            break;
3210    
3211          case OP_NOT_WHITESPACE:
3212            invert_bits = TRUE;
3213            /* Fall through */
3214          case OP_WHITESPACE:
3215            set2 = (pcre_uint32 *)(cd->cbits + cbit_space);
3216            break;
3217    
3218          case OP_NOT_WORDCHAR:
3219            invert_bits = TRUE;
3220            /* Fall through */
3221          case OP_WORDCHAR:
3222            set2 = (pcre_uint32 *)(cd->cbits + cbit_word);
3223            break;
3224    
3225          default:
3226          return FALSE;
3227          }
3228    
3229        /* Compare 4 bytes to improve speed. */
3230        set_end = set1 + (32 / 4);
3231        if (invert_bits)
3232          {
3233          do
3234            {
3235            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3236            }
3237          while (set1 < set_end);
3238          }
3239        else
3240          {
3241          do
3242            {
3243            if ((*set1++ & *set2++) != 0) return FALSE;
3244            }
3245          while (set1 < set_end);
3246          }
3247    
3248        if (list[1] == 0) return TRUE;
3249        /* Might be an empty repeat. */
3250        continue;
3251        }
3252    
3253      /* Some property combinations also acceptable. Unicode property opcodes are
3254      processed specially; the rest can be handled with a lookup table. */
3255    
3256      else
3257        {
3258        pcre_uint32 leftop, rightop;
3259    
3260        leftop = base_list[0];
3261        rightop = list[0];
3262    
3263    #ifdef SUPPORT_UCP
3264        accepted = FALSE; /* Always set in non-unicode case. */
3265        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3266          {
3267          if (rightop == OP_EOD)
3268            accepted = TRUE;
3269          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3270            {
3271            int n;
3272            const pcre_uint8 *p;
3273            BOOL same = leftop == rightop;
3274            BOOL lisprop = leftop == OP_PROP;
3275            BOOL risprop = rightop == OP_PROP;
3276            BOOL bothprop = lisprop && risprop;
3277    
3278            /* There's a table that specifies how each combination is to be
3279            processed:
3280              0   Always return FALSE (never auto-possessify)
3281              1   Character groups are distinct (possessify if both are OP_PROP)
3282              2   Check character categories in the same group (general or particular)
3283              3   Return TRUE if the two opcodes are not the same
3284              ... see comments below
3285            */
3286    
3287            n = propposstab[base_list[2]][list[2]];
3288            switch(n)
3289              {
3290              case 0: break;
3291              case 1: accepted = bothprop; break;
3292              case 2: accepted = (base_list[3] == list[3]) != same; break;
3293              case 3: accepted = !same; break;
3294    
3295              case 4:  /* Left general category, right particular category */
3296              accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3297              break;
3298    
3299              case 5:  /* Right general category, left particular category */
3300              accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3301              break;
3302    
3303              /* This code is logically tricky. Think hard before fiddling with it.
3304              The posspropstab table has four entries per row. Each row relates to
3305              one of PCRE's special properties such as ALNUM or SPACE or WORD.
3306              Only WORD actually needs all four entries, but using repeats for the
3307              others means they can all use the same code below.
3308    
3309              The first two entries in each row are Unicode general categories, and
3310              apply always, because all the characters they include are part of the
3311              PCRE character set. The third and fourth entries are a general and a
3312              particular category, respectively, that include one or more relevant
3313              characters. One or the other is used, depending on whether the check
3314              is for a general or a particular category. However, in both cases the
3315              category contains more characters than the specials that are defined
3316              for the property being tested against. Therefore, it cannot be used
3317              in a NOTPROP case.
3318    
3319              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3320              Underscore is covered by ucp_P or ucp_Po. */
3321    
3322              case 6:  /* Left alphanum vs right general category */
3323              case 7:  /* Left space vs right general category */
3324              case 8:  /* Left word vs right general category */
3325              p = posspropstab[n-6];
3326              accepted = risprop && lisprop ==
3327                (list[3] != p[0] &&
3328                 list[3] != p[1] &&
3329                (list[3] != p[2] || !lisprop));
3330              break;
3331    
3332              case 9:   /* Right alphanum vs left general category */
3333              case 10:  /* Right space vs left general category */
3334              case 11:  /* Right word vs left general category */
3335              p = posspropstab[n-9];
3336              accepted = lisprop && risprop ==
3337                (base_list[3] != p[0] &&
3338                 base_list[3] != p[1] &&
3339                (base_list[3] != p[2] || !risprop));
3340              break;
3341    
3342              case 12:  /* Left alphanum vs right particular category */
3343              case 13:  /* Left space vs right particular category */
3344              case 14:  /* Left word vs right particular category */
3345              p = posspropstab[n-12];
3346              accepted = risprop && lisprop ==
3347                (catposstab[p[0]][list[3]] &&
3348                 catposstab[p[1]][list[3]] &&
3349                (list[3] != p[3] || !lisprop));
3350              break;
3351    
3352              case 15:  /* Right alphanum vs left particular category */
3353              case 16:  /* Right space vs left particular category */
3354              case 17:  /* Right word vs left particular category */
3355              p = posspropstab[n-15];
3356              accepted = lisprop && risprop ==
3357                (catposstab[p[0]][base_list[3]] &&
3358                 catposstab[p[1]][base_list[3]] &&
3359                (base_list[3] != p[3] || !risprop));
3360              break;
3361              }
3362            }
3363          }
3364    
3365        else
3366    #endif  /* SUPPORT_UCP */
3367    
3368        accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3369               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3370               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3371    
3372        if (!accepted)
3373          return FALSE;
3374    
3375        if (list[1] == 0) return TRUE;
3376        /* Might be an empty repeat. */
3377        continue;
3378        }
3379    
3380      /* Control reaches here only if one of the items is a small character list.
3381      All characters are checked against the other side. */
3382    
3383      do
3384        {
3385        chr = *chr_ptr;
3386    
3387        switch(list_ptr[0])
3388          {
3389          case OP_CHAR:
3390          ochr_ptr = list_ptr + 2;
3391          do
3392            {
3393            if (chr == *ochr_ptr) return FALSE;
3394            ochr_ptr++;
3395            }
3396          while(*ochr_ptr != NOTACHAR);
3397          break;
3398    
3399          case OP_NOT:
3400          ochr_ptr = list_ptr + 2;
3401          do
3402            {
3403            if (chr == *ochr_ptr)
3404              break;
3405            ochr_ptr++;
3406            }
3407          while(*ochr_ptr != NOTACHAR);
3408          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3409          break;
3410    
3411          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3412          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3413    
3414          case OP_DIGIT:
3415          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3416          break;
3417    
3418          case OP_NOT_DIGIT:
3419          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3420          break;
3421    
3422          case OP_WHITESPACE:
3423          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3424          break;
3425    
3426          case OP_NOT_WHITESPACE:
3427          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3428          break;
3429    
3430          case OP_WORDCHAR:
3431          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3432          break;
3433    
3434          case OP_NOT_WORDCHAR:
3435          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3436          break;
3437    
3438          case OP_HSPACE:
3439          switch(chr)
3440            {
3441            HSPACE_CASES: return FALSE;
3442            default: break;
3443            }
3444          break;
3445    
3446          case OP_NOT_HSPACE:
3447          switch(chr)
3448            {
3449            HSPACE_CASES: break;
3450            default: return FALSE;
3451            }
3452          break;
3453    
3454          case OP_ANYNL:
3455          case OP_VSPACE:
3456          switch(chr)
3457            {
3458            VSPACE_CASES: return FALSE;
3459            default: break;
3460            }
3461          break;
3462    
3463          case OP_NOT_VSPACE:
3464          switch(chr)
3465            {
3466            VSPACE_CASES: break;
3467            default: return FALSE;
3468            }
3469          break;
3470    
3471          case OP_DOLL:
3472          case OP_EODN:
3473          switch (chr)
3474            {
3475            case CHAR_CR:
3476            case CHAR_LF:
3477            case CHAR_VT:
3478            case CHAR_FF:
3479            case CHAR_NEL:
3480    #ifndef EBCDIC
3481            case 0x2028:
3482            case 0x2029:
3483    #endif  /* Not EBCDIC */
3484            return FALSE;
3485            }
3486          break;
3487    
3488          case OP_EOD:    /* Can always possessify before \z */
3489          break;
3490    
3491          case OP_PROP:
3492          case OP_NOTPROP:
3493          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3494                list_ptr[0] == OP_NOTPROP))
3495            return FALSE;
3496          break;
3497    
3498          case OP_NCLASS:
3499          if (chr > 255) return FALSE;
3500          /* Fall through */
3501    
3502          case OP_CLASS:
3503          if (chr > 255) break;
3504          class_bitset = (pcre_uint8 *)
3505            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3506          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3507          break;
3508    
3509    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3510          case OP_XCLASS:
3511          if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3512              list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3513          break;
3514    #endif
3515    
3516          default:
3517          return FALSE;
3518          }
3519    
3520        chr_ptr++;
3521        }
3522      while(*chr_ptr != NOTACHAR);
3523    
3524      /* At least one character must be matched from this opcode. */
3525    
3526      if (list[1] == 0) return TRUE;
3527      }
3528    
3529    return FALSE;
3530    }
3531    
3532    
3533    
3534    /*************************************************
3535    *    Scan compiled regex for auto-possession     *
3536    *************************************************/
3537    
3538    /* Replaces single character iterations with their possessive alternatives
3539    if appropriate. This function modifies the compiled opcode!
3540    
3541    Arguments:
3542      code        points to start of the byte code
3543      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3544      cd          static compile data
3545    
3546    Returns:      nothing
3547    */
3548    
3549    static void
3550    auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3551    {
3552    register pcre_uchar c;
3553    const pcre_uchar *end;
3554    pcre_uchar *repeat_opcode;
3555    pcre_uint32 list[8];
3556    
3557    for (;;)
3558      {
3559      c = *code;
3560    
3561      if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3562        {
3563        c -= get_repeat_base(c) - OP_STAR;
3564        end = (c <= OP_MINUPTO) ?
3565          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3566        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3567    
3568        if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3569          {
3570          switch(c)
3571            {
3572            case OP_STAR:
3573            *code += OP_POSSTAR - OP_STAR;
3574            break;
3575    
3576            case OP_MINSTAR:
3577            *code += OP_POSSTAR - OP_MINSTAR;
3578            break;
3579    
3580            case OP_PLUS:
3581            *code += OP_POSPLUS - OP_PLUS;
3582            break;
3583    
3584            case OP_MINPLUS:
3585            *code += OP_POSPLUS - OP_MINPLUS;
3586            break;
3587    
3588            case OP_QUERY:
3589            *code += OP_POSQUERY - OP_QUERY;
3590            break;
3591    
3592            case OP_MINQUERY:
3593            *code += OP_POSQUERY - OP_MINQUERY;
3594            break;
3595    
3596            case OP_UPTO:
3597            *code += OP_POSUPTO - OP_UPTO;
3598            break;
3599    
3600            case OP_MINUPTO:
3601            *code += OP_MINUPTO - OP_UPTO;
3602            break;
3603            }
3604          }
3605        c = *code;
3606        }
3607      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3608        {
3609    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3610        if (c == OP_XCLASS)
3611          repeat_opcode = code + GET(code, 1);
3612        else
3613    #endif
3614          repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3615    
3616        c = *repeat_opcode;
3617        if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3618          {
3619          /* end must not be NULL. */
3620          end = get_chr_property_list(code, utf, cd->fcc, list);
3621    
3622          list[1] = (c & 1) == 0;
3623    
3624          if (compare_opcodes(end, utf, cd, list, end))
3625            {
3626            switch (c)
3627              {
3628              case OP_CRSTAR:
3629              case OP_CRMINSTAR:
3630              *repeat_opcode = OP_CRPOSSTAR;
3631              break;
3632    
3633              case OP_CRPLUS:
3634              case OP_CRMINPLUS:
3635              *repeat_opcode = OP_CRPOSPLUS;
3636              break;
3637    
3638              case OP_CRQUERY:
3639              case OP_CRMINQUERY:
3640              *repeat_opcode = OP_CRPOSQUERY;
3641              break;
3642    
3643              case OP_CRRANGE:
3644              case OP_CRMINRANGE:
3645              *repeat_opcode = OP_CRPOSRANGE;
3646              break;
3647              }
3648            }
3649          }
3650        c = *code;
3651        }
3652    
3653      switch(c)
3654        {
3655        case OP_END:
3656        return;
3657    
3658        case OP_TYPESTAR:
3659        case OP_TYPEMINSTAR:
3660        case OP_TYPEPLUS:
3661        case OP_TYPEMINPLUS:
3662        case OP_TYPEQUERY:
3663        case OP_TYPEMINQUERY:
3664        case OP_TYPEPOSSTAR:
3665        case OP_TYPEPOSPLUS:
3666        case OP_TYPEPOSQUERY:
3667        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3668        break;
3669    
3670        case OP_TYPEUPTO:
3671        case OP_TYPEMINUPTO:
3672        case OP_TYPEEXACT:
3673        case OP_TYPEPOSUPTO:
3674        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3675          code += 2;
3676        break;
3677    
3678    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3679        case OP_XCLASS:
3680        code += GET(code, 1);
3681        break;
3682    #endif
3683    
3684        case OP_MARK:
3685        case OP_PRUNE_ARG:
3686        case OP_SKIP_ARG:
3687        case OP_THEN_ARG:
3688        code += code[1];
3689        break;
3690        }
3691    
3692      /* Add in the fixed length from the table */
3693    
3694      code += PRIV(OP_lengths)[c];
3695    
3696      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3697      a multi-byte character. The length in the table is a minimum, so we have to
3698      arrange to skip the extra bytes. */
3699    
3700    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3701      if (utf) switch(c)
3702        {
3703        case OP_CHAR:
3704        case OP_CHARI:
3705        case OP_NOT:
3706        case OP_NOTI:
3707        case OP_STAR:
3708        case OP_MINSTAR:
3709        case OP_PLUS:
3710        case OP_MINPLUS:
3711        case OP_QUERY:
3712        case OP_MINQUERY:
3713        case OP_UPTO:
3714        case OP_MINUPTO:
3715        case OP_EXACT:
3716        case OP_POSSTAR:
3717        case OP_POSPLUS:
3718        case OP_POSQUERY:
3719        case OP_POSUPTO:
3720        case OP_STARI:
3721        case OP_MINSTARI:
3722        case OP_PLUSI:
3723        case OP_MINPLUSI:
3724        case OP_QUERYI:
3725        case OP_MINQUERYI:
3726        case OP_UPTOI:
3727        case OP_MINUPTOI:
3728        case OP_EXACTI:
3729        case OP_POSSTARI:
3730        case OP_POSPLUSI:
3731        case OP_POSQUERYI:
3732        case OP_POSUPTOI:
3733        case OP_NOTSTAR:
3734        case OP_NOTMINSTAR:
3735        case OP_NOTPLUS:
3736        case OP_NOTMINPLUS:
3737        case OP_NOTQUERY:
3738        case OP_NOTMINQUERY:
3739        case OP_NOTUPTO:
3740        case OP_NOTMINUPTO:
3741        case OP_NOTEXACT:
3742        case OP_NOTPOSSTAR:
3743        case OP_NOTPOSPLUS:
3744        case OP_NOTPOSQUERY:
3745        case OP_NOTPOSUPTO:
3746        case OP_NOTSTARI:
3747        case OP_NOTMINSTARI:
3748        case OP_NOTPLUSI:
3749        case OP_NOTMINPLUSI:
3750        case OP_NOTQUERYI:
3751        case OP_NOTMINQUERYI:
3752        case OP_NOTUPTOI:
3753        case OP_NOTMINUPTOI:
3754        case OP_NOTEXACTI:
3755        case OP_NOTPOSSTARI:
3756        case OP_NOTPOSPLUSI:
3757        case OP_NOTPOSQUERYI:
3758        case OP_NOTPOSUPTOI:
3759        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3760        break;
3761        }
3762    #else
3763      (void)(utf);  /* Keep compiler happy by referencing function argument */
3764    #endif
3765      }
3766    }
3767    
3768    
3769    
3770    /*************************************************
3771    *           Check for POSIX class syntax         *
3772    *************************************************/
3773    
3774    /* This function is called when the sequence "[:" or "[." or "[=" is
3775    encountered in a character class. It checks whether this is followed by a
3776    sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3777    reach an unescaped ']' without the special preceding character, return FALSE.
3778    
3779    Originally, this function only recognized a sequence of letters between the
3780    terminators, but it seems that Perl recognizes any sequence of characters,
3781    though of course unknown POSIX names are subsequently rejected. Perl gives an
3782    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3783    didn't consider this to be a POSIX class. Likewise for [:1234:].
3784    
3785    The problem in trying to be exactly like Perl is in the handling of escapes. We
3786    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3787    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3788    below handles the special case of \], but does not try to do any other escape
3789    processing. This makes it different from Perl for cases such as [:l\ower:]
3790    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3791    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
3792    I think.
3793    
3794    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3795    It seems that the appearance of a nested POSIX class supersedes an apparent
3796    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3797    a digit.
3798    
3799    In Perl, unescaped square brackets may also appear as part of class names. For
3800    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3801    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3802    seem right at all. PCRE does not allow closing square brackets in POSIX class
3803    names.
3804    
3805    Arguments:
3806      ptr      pointer to the initial [
3807      endptr   where to return the end pointer
3808    
3809    Returns:   TRUE or FALSE
3810    */
3811    
3812    static BOOL
3813    check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3814    {
3815    pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
3816    terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
3817    for (++ptr; *ptr != CHAR_NULL; ptr++)
3818      {
3819      if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3820        ptr++;
3821      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3822      else
3823        {
3824        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3825          {
3826          *endptr = ptr;
3827          return TRUE;
3828          }
3829        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3830             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3831              ptr[1] == CHAR_EQUALS_SIGN) &&
3832            check_posix_syntax(ptr, endptr))
3833          return FALSE;
3834        }
3835      }
3836    return FALSE;
3837    }
3838    
3839    
3840    
3841    
3842    /*************************************************
3843    *          Check POSIX class name                *
3844    *************************************************/
3845    
3846    /* This function is called to check the name given in a POSIX-style class entry
3847    such as [:alnum:].
3848    
3849    Arguments:
3850      ptr        points to the first letter
3851      len        the length of the name
3852    
3853    Returns:     a value representing the name, or -1 if unknown
3854    */
3855    
3856    static int
3857    check_posix_name(const pcre_uchar *ptr, int len)
3858    {
3859    const char *pn = posix_names;
3860    register int yield = 0;
3861    while (posix_name_lengths[yield] != 0)
3862      {
3863      if (len == posix_name_lengths[yield] &&
3864        STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3865      pn += posix_name_lengths[yield] + 1;
3866      yield++;
3867      }
3868    return -1;
3869    }
3870    
3871    
3872    /*************************************************
3873    *    Adjust OP_RECURSE items in repeated group   *
3874    *************************************************/
3875    
3876    /* OP_RECURSE items contain an offset from the start of the regex to the group
3877    that is referenced. This means that groups can be replicated for fixed
3878    repetition simply by copying (because the recursion is allowed to refer to
3879    earlier groups that are outside the current group). However, when a group is
3880    optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3881    inserted before it, after it has been compiled. This means that any OP_RECURSE
3882    items within it that refer to the group itself or any contained groups have to
3883    have their offsets adjusted. That one of the jobs of this function. Before it
3884    is called, the partially compiled regex must be temporarily terminated with
3885    OP_END.
3886    
3887    This function has been extended with the possibility of forward references for
3888    recursions and subroutine calls. It must also check the list of such references
3889    for the group we are dealing with. If it finds that one of the recursions in
3890    the current group is on this list, it adjusts the offset in the list, not the
3891    value in the reference (which is a group number).
3892    
3893    Arguments:
3894      group      points to the start of the group
3895      adjust     the amount by which the group is to be moved
3896      utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
3897      cd         contains pointers to tables etc.
3898      save_hwm   the hwm forward reference pointer at the start of the group
3899    
3900    Returns:     nothing
3901    */
3902    
3903    static void
3904    adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3905      pcre_uchar *save_hwm)
3906    {
3907    pcre_uchar *ptr = group;
3908    
3909    while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3910      {
3911      int offset;
3912      pcre_uchar *hc;
3913    
3914      /* See if this recursion is on the forward reference list. If so, adjust the
3915      reference. */
3916    
3917      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3918        {
3919        offset = (int)GET(hc, 0);
3920        if (cd->start_code + offset == ptr + 1)
3921          {
3922          PUT(hc, 0, offset + adjust);
3923          break;
3924          }
3925        }
3926    
3927      /* Otherwise, adjust the recursion offset if it's after the start of this
3928      group. */
3929    
3930      if (hc >= cd->hwm)
3931        {
3932        offset = (int)GET(ptr, 1);
3933        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3934        }
3935    
3936      ptr += 1 + LINK_SIZE;
3937      }
3938    }
3939    
3940    
3941    
3942    /*************************************************
3943    *        Insert an automatic callout point       *
3944    *************************************************/
3945    
3946    /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
3947    callout points before each pattern item.
3948    
3949    Arguments:
3950      code           current code pointer
3951      ptr            current pattern pointer
3952      cd             pointers to tables etc
3953    
3954    Returns:         new code pointer
3955    */
3956    
3957    static pcre_uchar *
3958    auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
3959    {
3960    *code++ = OP_CALLOUT;
3961    *code++ = 255;
3962    PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
3963    PUT(code, LINK_SIZE, 0);                       /* Default length */
3964    return code + 2 * LINK_SIZE;
3965    }
3966    
3967    
3968    
# Line 2858  PUT(previous_callout, 2 + LINK_SIZE, len Line 3997  PUT(previous_callout, 2 + LINK_SIZE, len
3997  *************************************************/  *************************************************/
3998    
3999  /* This function is passed the start and end of a class range, in UTF-8 mode  /* This function is passed the start and end of a class range, in UTF-8 mode
4000  with UCP support. It searches up the characters, looking for internal ranges of  with UCP support. It searches up the characters, looking for ranges of
4001  characters in the "other" case. Each call returns the next one, updating the  characters in the "other" case. Each call returns the next one, updating the
4002  start address.  start address. A character with multiple other cases is returned on its own
4003    with a special return value.
4004    
4005  Arguments:  Arguments:
4006    cptr        points to starting character value; updated    cptr        points to starting character value; updated
# Line 2868  Arguments: Line 4008  Arguments:
4008    ocptr       where to put start of othercase range    ocptr       where to put start of othercase range
4009    odptr       where to put end of othercase range    odptr       where to put end of othercase range
4010    
4011  Yield:        TRUE when range returned; FALSE when no more  Yield:        -1 when no more
4012                   0 when a range is returned
4013                  >0 the CASESET offset for char with multiple other cases
4014                    in this case, ocptr contains the original
4015  */  */
4016    
4017  static BOOL  static int
4018  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4019    unsigned int *odptr)    pcre_uint32 *odptr)
4020  {  {
4021  unsigned int c, othercase, next;  pcre_uint32 c, othercase, next;
4022    unsigned int co;
4023    
4024    /* Find the first character that has an other case. If it has multiple other
4025    cases, return its case offset value. */
4026    
4027  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
4028    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }    {
4029      if ((co = UCD_CASESET(c)) != 0)
4030        {
4031        *ocptr = c++;   /* Character that has the set */
4032        *cptr = c;      /* Rest of input range */
4033        return (int)co;
4034        }
4035      if ((othercase = UCD_OTHERCASE(c)) != c) break;
4036      }
4037    
4038  if (c > d) return FALSE;  if (c > d) return -1;  /* Reached end of range */
4039    
4040  *ocptr = othercase;  *ocptr = othercase;
4041  next = othercase + 1;  next = othercase + 1;
# Line 2891  for (++c; c <= d; c++) Line 4046  for (++c; c <= d; c++)
4046    next++;    next++;
4047    }    }
4048    
4049  *odptr = next - 1;  *odptr = next - 1;     /* End of othercase range */
4050  *cptr = c;  *cptr = c;             /* Rest of input range */
4051    return 0;
 return TRUE;  
 }  
   
   
   
 /*************************************************  
 *        Check a character and a property        *  
 *************************************************/  
   
 /* This function is called by check_auto_possessive() when a property item  
 is adjacent to a fixed character.  
   
 Arguments:  
   c            the character  
   ptype        the property type  
   pdata        the data for the type  
   negated      TRUE if it's a negated property (\P or \p{^)  
   
 Returns:       TRUE if auto-possessifying is OK  
 */  
   
 static BOOL  
 check_char_prop(int c, int ptype, int pdata, BOOL negated)  
 {  
 const ucd_record *prop = GET_UCD(c);  
 switch(ptype)  
   {  
   case PT_LAMP:  
   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;  
   
   case PT_PC:  
   return (pdata == prop->chartype) == negated;  
   
   case PT_SC:  
   return (pdata == prop->script) == negated;  
   
   /* These are specials */  
   
   case PT_ALNUM:  
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||  
           PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;  
   
   case PT_SPACE:    /* Perl space */  
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||  
           c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)  
           == negated;  
   
   case PT_PXSPACE:  /* POSIX space */  
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||  
           c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||  
           c == CHAR_FF || c == CHAR_CR)  
           == negated;  
   
   case PT_WORD:  
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||  
           PRIV(ucp_gentype)[prop->chartype] == ucp_N ||  
           c == CHAR_UNDERSCORE) == negated;  
   }  
 return FALSE;  
4052  }  }
4053  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
4054    
4055    
4056    
4057  /*************************************************  /*************************************************
4058  *     Check if auto-possessifying is possible    *  *        Add a character or range to a class     *
4059  *************************************************/  *************************************************/
4060    
4061  /* This function is called for unlimited repeats of certain items, to see  /* This function packages up the logic of adding a character or range of
4062  whether the next thing could possibly match the repeated item. If not, it makes  characters to a class. The character values in the arguments will be within the
4063  sense to automatically possessify the repeated item.  valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
4064    mutually recursive with the function immediately below.
4065    
4066  Arguments:  Arguments:
4067    previous      pointer to the repeated opcode    classbits     the bit map for characters < 256
4068    utf           TRUE in UTF-8 / UTF-16 mode    uchardptr     points to the pointer for extra data
4069    ptr           next character in pattern    options       the options word
   options       options bits  
4070    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4071      start         start of range character
4072      end           end of range character
4073    
4074  Returns:        TRUE if possessifying is wanted  Returns:        the number of < 256 characters added
4075                    the pointer to extra data is updated
4076  */  */
4077    
4078  static BOOL  static int
4079  check_auto_possessive(const pcre_uchar *previous, BOOL utf,  add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4080    const pcre_uchar *ptr, int options, compile_data *cd)    compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4081  {  {
4082  pcre_int32 c, next;  pcre_uint32 c;
4083  int op_code = *previous++;  int n8 = 0;
4084    
4085  /* Skip whitespace and comments in extended mode */  /* If caseless matching is required, scan the range and process alternate
4086    cases. In Unicode, there are 8-bit characters that have alternate cases that
4087    are greater than 255 and vice-versa. Sometimes we can just extend the original
4088    range. */
4089    
4090  if ((options & PCRE_EXTENDED) != 0)  if ((options & PCRE_CASELESS) != 0)
4091    {    {
4092    for (;;)  #ifdef SUPPORT_UCP
4093      if ((options & PCRE_UTF8) != 0)
4094      {      {
4095      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      int rc;
4096      if (*ptr == CHAR_NUMBER_SIGN)      pcre_uint32 oc, od;
       {  
       ptr++;  
       while (*ptr != 0)  
         {  
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
       }  
     else break;  
     }  
   }  
   
 /* If the next item is one that we can handle, get its value. A non-negative  
 value is a character, a negative value is an escape value. */  
   
 if (*ptr == CHAR_BACKSLASH)  
   {  
   int temperrorcode = 0;  
   next = check_escape(&ptr, &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)  
   {  
 #ifdef SUPPORT_UTF  
   if (utf) { GETCHARINC(next, ptr); } else  
 #endif  
   next = *ptr++;  
   }  
 else return FALSE;  
4097    
4098  /* Skip whitespace and comments in extended mode */      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
4099        c = start;
4100    
4101  if ((options & PCRE_EXTENDED) != 0)      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
   {  
   for (;;)  
     {  
     while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;  
     if (*ptr == CHAR_NUMBER_SIGN)  
4102        {        {
4103        ptr++;        /* Handle a single character that has more than one other case. */
       while (*ptr != 0)  
         {  
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
       }  
     else break;  
     }  
   }  
   
 /* If the next thing is itself optional, we have to give up. */  
4104    
4105  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4106    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)          PRIV(ucd_caseless_sets) + rc, oc);
     return FALSE;  
   
 /* Now compare the next item with the previous opcode. First, handle cases when  
 the next item is a character. */  
   
 if (next >= 0) switch(op_code)  
   {  
   case OP_CHAR:  
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   return c != next;  
4107    
4108    /* For CHARI (caseless character) we must check the other case. If we have        /* Do nothing if the other case range is within the original range. */
   Unicode property support, we can use it to test the other case of  
   high-valued characters. */  
4109    
4110    case OP_CHARI:        else if (oc >= start && od <= end) continue;
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   if (c == next) return FALSE;  
 #ifdef SUPPORT_UTF  
   if (utf)  
     {  
     unsigned int othercase;  
     if (next < 128) othercase = cd->fcc[next]; else  
 #ifdef SUPPORT_UCP  
     othercase = UCD_OTHERCASE((unsigned int)next);  
 #else  
     othercase = NOTACHAR;  
 #endif  
     return (unsigned int)c != othercase;  
     }  
   else  
 #endif  /* SUPPORT_UTF */  
   return (c != TABLE_GET((unsigned int)next, cd->fcc, next));  /* Non-UTF-8 mode */  
4111    
4112    case OP_NOT:        /* Extend the original range if there is overlap, noting that if oc < c, we
4113  #ifdef SUPPORT_UTF        can't have od > end because a subrange is always shorter than the basic
4114    GETCHARTEST(c, previous);        range. Otherwise, use a recursive call to add the additional range. */
 #else  
   c = *previous;  
 #endif  
   return c == next;  
4115    
4116    case OP_NOTI:        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4117  #ifdef SUPPORT_UTF        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
4118    GETCHARTEST(c, previous);        else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
4119  #else        }
   c = *previous;  
 #endif  
   if (c == next) return TRUE;  
 #ifdef SUPPORT_UTF  
   if (utf)  
     {  
     unsigned int othercase;  
     if (next < 128) othercase = cd->fcc[next]; else  
 #ifdef SUPPORT_UCP  
     othercase = UCD_OTHERCASE((unsigned int)next);  
 #else  
     othercase = NOTACHAR;  
 #endif  
     return (unsigned int)c == othercase;  
4120      }      }
4121    else    else
4122  #endif  /* SUPPORT_UTF */  #endif  /* SUPPORT_UCP */
   return (c == TABLE_GET((unsigned int)next, cd->fcc, next));  /* Non-UTF-8 mode */  
   
   /* 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. */  
   
   case OP_DIGIT:  
   return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;  
   
   case OP_NOT_DIGIT:  
   return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;  
   
   case OP_WHITESPACE:  
   return next > 255 || (cd->ctypes[next] & ctype_space) == 0;  
   
   case OP_NOT_WHITESPACE:  
   return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;  
   
   case OP_WORDCHAR:  
   return next > 255 || (cd->ctypes[next] & ctype_word) == 0;  
   
   case OP_NOT_WORDCHAR:  
   return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;  
4123    
4124    case OP_HSPACE:    /* Not UTF-mode, or no UCP */
   case OP_NOT_HSPACE:  
   switch(next)  
     {  
     case 0x09:  
     case 0x20:  
     case 0xa0:  
     case 0x1680:  
     case 0x180e:  
     case 0x2000:  
     case 0x2001:  
     case 0x2002:  
     case 0x2003:  
     case 0x2004:  
     case 0x2005:  
     case 0x2006:  
     case 0x2007:  
     case 0x2008:  
     case 0x2009:  
     case 0x200A:  
     case 0x202f:  
     case 0x205f:  
     case 0x3000:  
     return op_code == OP_NOT_HSPACE;  
     default:  
     return op_code != OP_NOT_HSPACE;  
     }  
4125    
4126    case OP_ANYNL:    for (c = start; c <= end && c < 256; c++)
   case OP_VSPACE:  
   case OP_NOT_VSPACE:  
   switch(next)  
4127      {      {
4128      case 0x0a:      SETBIT(classbits, cd->fcc[c]);
4129      case 0x0b:      n8++;
     case 0x0c:  
     case 0x0d:  
     case 0x85:  
     case 0x2028:  
     case 0x2029:  
     return op_code == OP_NOT_VSPACE;  
     default:  
     return op_code != OP_NOT_VSPACE;  
4130      }      }
   
 #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;  
4131    }    }
4132    
4133    /* Now handle the original range. Adjust the final value according to the bit
4134    length - this means that the same lists of (e.g.) horizontal spaces can be used
4135    in all cases. */
4136    
4137  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP  #if defined COMPILE_PCRE8
 is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are  
 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. */  
   
 switch(op_code)  
   {  
   case OP_CHAR:  
   case OP_CHARI:  
4138  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4139    GETCHARTEST(c, previous);    if ((options & PCRE_UTF8) == 0)
 #else  
   c = *previous;  
4140  #endif  #endif
4141    switch(-next)    if (end > 0xff) end = 0xff;
     {  
     case ESC_d:  
     return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;  
4142    
4143      case ESC_D:  #elif defined COMPILE_PCRE16
4144      return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;  #ifdef SUPPORT_UTF
4145      if ((options & PCRE_UTF16) == 0)
4146    #endif
4147      if (end > 0xffff) end = 0xffff;
4148    
4149      case ESC_s:  #endif /* COMPILE_PCRE[8|16] */
     return c > 255 || (cd->ctypes[c] & ctype_space) == 0;  
4150    
4151      case ESC_S:  /* If all characters are less than 256, use the bit map. Otherwise use extra
4152      return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;  data. */
4153    
4154      case ESC_w:  if (end < 0x100)
4155      return c > 255 || (cd->ctypes[c] & ctype_word) == 0;    {
4156      for (c = start; c <= end; c++)
4157        {
4158        n8++;
4159        SETBIT(classbits, c);
4160        }
4161      }
4162    
4163      case ESC_W:  else
4164      return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;    {
4165      pcre_uchar *uchardata = *uchardptr;
4166    
4167      case ESC_h:  #ifdef SUPPORT_UTF
4168      case ESC_H:    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
4169      switch(c)      {
4170        if (start < end)
4171        {        {
4172        case 0x09:        *uchardata++ = XCL_RANGE;
4173        case 0x20:        uchardata += PRIV(ord2utf)(start, uchardata);
4174        case 0xa0:        uchardata += PRIV(ord2utf)(end, uchardata);
       case 0x1680:  
       case 0x180e:  
       case 0x2000:  
       case 0x2001:  
       case 0x2002:  
       case 0x2003:  
       case 0x2004:  
       case 0x2005:  
       case 0x2006:  
       case 0x2007:  
       case 0x2008:  
       case 0x2009:  
       case 0x200A:  
       case 0x202f:  
       case 0x205f:  
       case 0x3000:  
       return -next != ESC_h;  
       default:  
       return -next == ESC_h;  
4175        }        }
4176        else if (start == end)
     case ESC_v:  
     case ESC_V:  
     switch(c)  
4177        {        {
4178        case 0x0a:        *uchardata++ = XCL_SINGLE;
4179        case 0x0b:        uchardata += PRIV(ord2utf)(start, uchardata);
       case 0x0c:  
       case 0x0d:  
       case 0x85:  
       case 0x2028:  
       case 0x2029:  
       return -next != ESC_v;  
       default:  
       return -next == ESC_v;  
       }  
   
     /* When PCRE_UCP is set, these values get generated for \d etc. Find  
     their substitutions and process them. The result will always be either  
     -ESC_p or -ESC_P. Then fall through to process those values. */  
   
 #ifdef SUPPORT_UCP  
     case ESC_du:  
     case ESC_DU:  
     case ESC_wu:  
     case ESC_WU:  
     case ESC_su:  
     case ESC_SU:  
       {  
       int temperrorcode = 0;  
       ptr = substitutes[-next - ESC_DU];  
       next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);  
       if (temperrorcode != 0) return FALSE;  
       ptr++;    /* For compatibility */  
4180        }        }
4181      /* Fall through */      }
4182      else
4183    #endif  /* SUPPORT_UTF */
4184    
4185      case ESC_p:    /* Without UTF support, character values are constrained by the bit length,
4186      case ESC_P:    and can only be > 256 for 16-bit and 32-bit libraries. */
       {  
       int ptype, pdata, errorcodeptr;  
       BOOL negated;  
4187    
4188        ptr--;      /* Make ptr point at the p or P */  #ifdef COMPILE_PCRE8
4189        ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);      {}
4190        if (ptype < 0) return FALSE;  #else
4191        ptr++;      /* Point past the final curly ket */    if (start < end)
4192        {
4193        *uchardata++ = XCL_RANGE;
4194        *uchardata++ = start;
4195        *uchardata++ = end;
4196        }
4197      else if (start == end)
4198        {
4199        *uchardata++ = XCL_SINGLE;
4200        *uchardata++ = start;
4201        }
4202    #endif
4203    
4204        /* If the property item is optional, we have to give up. (When generated    *uchardptr = uchardata;   /* Updata extra data pointer */
4205        from \d etc by PCRE_UCP, this test will have been applied much earlier,    }
       to the original \d etc. At this point, ptr will point to a zero byte. */  
4206    
4207        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  return n8;    /* Number of 8-bit characters */
4208          STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)  }
           return FALSE;  
4209    
       /* Do the property check. */  
4210    
       return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);  
       }  
 #endif  
4211    
     default:  
     return FALSE;  
     }  
4212    
4213    /* In principle, support for Unicode properties should be integrated here as  /*************************************************
4214    well. It means re-organizing the above code so as to get hold of the property  *        Add a list of characters to a class     *
4215    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.) */  
4216    
4217    case OP_DIGIT:  /* This function is used for adding a list of case-equivalent characters to a
4218    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||  class, and also for adding a list of horizontal or vertical whitespace. If the
4219           next == -ESC_h || next == -ESC_v || next == -ESC_R;  list is in order (which it should be), ranges of characters are detected and
4220    handled appropriately. This function is mutually recursive with the function
4221    above.
4222    
4223    case OP_NOT_DIGIT:  Arguments:
4224    return next == -ESC_d;    classbits     the bit map for characters < 256
4225      uchardptr     points to the pointer for extra data
4226      options       the options word
4227      cd            contains pointers to tables etc.
4228      p             points to row of 32-bit values, terminated by NOTACHAR
4229      except        character to omit; this is used when adding lists of
4230                      case-equivalent characters to avoid including the one we
4231                      already know about
4232    
4233    case OP_WHITESPACE:  Returns:        the number of < 256 characters added
4234    return next == -ESC_S || next == -ESC_d || next == -ESC_w;                  the pointer to extra data is updated
4235    */
4236    
4237    case OP_NOT_WHITESPACE:  static int
4238    return next == -ESC_s || next == -ESC_h || next == -ESC_v || next == -ESC_R;  add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4239      compile_data *cd, const pcre_uint32 *p, unsigned int except)
4240    {
4241    int n8 = 0;
4242    while (p[0] < NOTACHAR)
4243      {
4244      int n = 0;
4245      if (p[0] != except)
4246        {
4247        while(p[n+1] == p[0] + n + 1) n++;
4248        n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4249        }
4250      p += n + 1;
4251      }
4252    return n8;
4253    }
4254    
   case OP_HSPACE:  
   return next == -ESC_S || next == -ESC_H || next == -ESC_d ||  
          next == -ESC_w || next == -ESC_v || next == -ESC_R;  
4255    
   case OP_NOT_HSPACE:  
   return next == -ESC_h;  
4256    
4257    /* Can't have \S in here because VT matches \S (Perl anomaly) */  /*************************************************
4258    case OP_ANYNL:  *    Add characters not in a list to a class     *
4259    case OP_VSPACE:  *************************************************/
   return next == -ESC_V || next == -ESC_d || next == -ESC_w;  
4260    
4261    case OP_NOT_VSPACE:  /* This function is used for adding the complement of a list of horizontal or
4262    return next == -ESC_v || next == -ESC_R;  vertical whitespace to a class. The list must be in order.
4263    
4264    case OP_WORDCHAR:  Arguments:
4265    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||    classbits     the bit map for characters < 256
4266           next == -ESC_v || next == -ESC_R;    uchardptr     points to the pointer for extra data
4267      options       the options word
4268      cd            contains pointers to tables etc.
4269      p             points to row of 32-bit values, terminated by NOTACHAR
4270    
4271    case OP_NOT_WORDCHAR:  Returns:        the number of < 256 characters added
4272    return next == -ESC_w || next == -ESC_d;                  the pointer to extra data is updated
4273    */
4274    
4275    default:  static int
4276    return FALSE;  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
4277      int options, compile_data *cd, const pcre_uint32 *p)
4278    {
4279    BOOL utf = (options & PCRE_UTF8) != 0;
4280    int n8 = 0;
4281    if (p[0] > 0)
4282      n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
4283    while (p[0] < NOTACHAR)
4284      {
4285      while (p[1] == p[0] + 1) p++;
4286      n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
4287        (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
4288      p++;
4289    }    }
4290    return n8;
 /* Control does not reach here */  
4291  }  }
4292    
4293    
# Line 3398  to find out the amount of memory needed, Line 4303  to find out the amount of memory needed,
4303  phase. The value of lengthptr distinguishes the two phases.  phase. The value of lengthptr distinguishes the two phases.
4304    
4305  Arguments:  Arguments:
4306    optionsptr     pointer to the option bits    optionsptr        pointer to the option bits
4307    codeptr        points to the pointer to the current code point    codeptr           points to the pointer to the current code point
4308    ptrptr         points to the current pattern pointer    ptrptr            points to the current pattern pointer
4309    errorcodeptr   points to error code variable    errorcodeptr      points to error code variable
4310    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr      place to put the first required character
4311    reqcharptr     set to the last literal character required, else < 0    firstcharflagsptr place to put the first character flags, or a negative number
4312    bcptr          points to current branch chain    reqcharptr        place to put the last required character
4313    cond_depth     conditional nesting depth    reqcharflagsptr   place to put the last required character flags, or a negative number
4314    cd             contains pointers to tables etc.    bcptr             points to current branch chain
4315    lengthptr      NULL during the real compile phase    cond_depth        conditional nesting depth
4316           &nbs