/[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 1045 by ph10, Sun Sep 23 16:50:00 2012 UTC revision 1380 by ph10, Tue Oct 15 16:49:12 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 80  to check them every time. */ Line 80  to check them every time. */
80  /* Definitions to allow mutual recursion */  /* Definitions to allow mutual recursion */
81    
82  static int  static int
83    add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,    add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84      const pcre_uint32 *, unsigned int);      const pcre_uint32 *, unsigned int);
85    
86  static BOOL  static BOOL
87    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL,    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
88      int, int, int *, int *, branch_chain *, compile_data *, int *);      pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
89        compile_data *, int *);
90    
91    
92    
# Line 114  kicks in at the same number of forward r Line 115  kicks in at the same number of forward r
115  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
116  #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)  #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
117    
118    /* This value determines the size of the initial vector that is used for
119    remembering named groups during the pre-compile. It is allocated on the stack,
120    but if it is too small, it is expanded using malloc(), in a similar way to the
121    workspace. The value is the number of slots in the list. */
122    
123    #define NAMED_GROUP_LIST_SIZE  20
124    
125  /* The overrun tests check for a slightly smaller size so that they detect the  /* The overrun tests check for a slightly smaller size so that they detect the
126  overrun before it actually does run off the end of the data block. */  overrun before it actually does run off the end of the data block. */
127    
# Line 121  overrun before it actually does run off Line 129  overrun before it actually does run off
129    
130  /* Private flags added to firstchar and reqchar. */  /* Private flags added to firstchar and reqchar. */
131    
132  #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */  #define REQ_CASELESS    (1 << 0)        /* Indicates caselessness */
133  #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */  #define REQ_VARY        (1 << 1)        /* Reqchar followed non-literal item */
134    /* 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 451  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 483  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 503  static const char error_texts[] = Line 514  static const char error_texts[] =
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"    "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 642  static const pcre_uint8 ebcdic_chartab[] Line 659  static const pcre_uint8 ebcdic_chartab[]
659  #endif  #endif
660    
661    
662    /* This table is used to check whether auto-possessification is possible
663    between adjacent character-type opcodes. The left-hand (repeated) opcode is
664    used to select the row, and the right-hand opcode is use to select the column.
665    A value of 1 means that auto-possessification is OK. For example, the second
666    value in the first row means that \D+\d can be turned into \D++\d.
667    
668    The Unicode property types (\P and \p) have to be present to fill out the table
669    because of what their opcode values are, but the table values should always be
670    zero because property types are handled separately in the code. The last four
671    columns apply to items that cannot be repeated, so there is no need to have
672    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
673    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
674    
675    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
676    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
677    
678    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
679    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
680      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
681      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
682      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
683      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
684      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
685      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
686      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
687      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
688      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
689      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
690      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
691      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
692      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
693      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
694      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
695      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
696      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
697    };
698    
699    
700    /* This table is used to check whether auto-possessification is possible
701    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
702    left-hand (repeated) opcode is used to select the row, and the right-hand
703    opcode is used to select the column. The values are as follows:
704    
705      0   Always return FALSE (never auto-possessify)
706      1   Character groups are distinct (possessify if both are OP_PROP)
707      2   Check character categories in the same group (general or particular)
708      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
709    
710      4   Check left general category vs right particular category
711      5   Check right general category vs left particular category
712    
713      6   Left alphanum vs right general category
714      7   Left space vs right general category
715      8   Left word vs right general category
716    
717      9   Right alphanum vs left general category
718     10   Right space vs left general category
719     11   Right word vs left general category
720    
721     12   Left alphanum vs right particular category
722     13   Left space vs right particular category
723     14   Left word vs right particular category
724    
725     15   Right alphanum vs left particular category
726     16   Right space vs left particular category
727     17   Right word vs left particular category
728    */
729    
730    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
731    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
732      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
733      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
734      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
735      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
736      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
737      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
738      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
739      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
740      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
741      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
742      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
743    };
744    
745    /* This table is used to check whether auto-possessification is possible
746    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
747    specifies a general category and the other specifies a particular category. The
748    row is selected by the general category and the column by the particular
749    category. The value is 1 if the particular category is not part of the general
750    category. */
751    
752    static const pcre_uint8 catposstab[7][30] = {
753    /* Cc Cf Cn Co Cs Ll Lm Lo Lt Lu Mc Me Mn Nd Nl No Pc Pd Pe Pf Pi Po Ps Sc Sk Sm So Zl Zp Zs */
754      { 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* C */
755      { 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* L */
756      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* M */
757      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* N */
758      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1 },  /* P */
759      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1 },  /* S */
760      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0 }   /* Z */
761    };
762    
763    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
764    a general or particular category. The properties in each row are those
765    that apply to the character set in question. Duplication means that a little
766    unnecessary work is done when checking, but this keeps things much simpler
767    because they can all use the same code. For more details see the comment where
768    this table is used.
769    
770    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
771    "space", but from Perl 5.18 it's included, so both categories are treated the
772    same here. */
773    
774    static const pcre_uint8 posspropstab[3][4] = {
775      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
776      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
777      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
778    };
779    
780    /* This table is used when converting repeating opcodes into possessified
781    versions as a result of an explicit possessive quantifier such as ++. A zero
782    value means there is no possessified version - in those cases the item in
783    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
784    because all relevant opcodes are less than that. */
785    
786    static const pcre_uint8 opcode_possessify[] = {
787      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
788      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
789    
790      0,                       /* NOTI */
791      OP_POSSTAR, 0,           /* STAR, MINSTAR */
792      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
793      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
794      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
795      0,                       /* EXACT */
796      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
797    
798      OP_POSSTARI, 0,          /* STARI, MINSTARI */
799      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
800      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
801      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
802      0,                       /* EXACTI */
803      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
804    
805      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
806      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
807      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
808      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
809      0,                       /* NOTEXACT */
810      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
811    
812      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
813      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
814      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
815      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
816      0,                       /* NOTEXACTI */
817      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
818    
819      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
820      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
821      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
822      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
823      0,                       /* TYPEEXACT */
824      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
825    
826      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
827      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
828      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
829      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
830      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
831    
832      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
833      0, 0,                    /* REF, REFI */
834      0, 0,                    /* DNREF, DNREFI */
835      0, 0                     /* RECURSE, CALLOUT */
836    };
837    
838    
839    
840  /*************************************************  /*************************************************
# Line 663  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 747  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 781  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 790  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 < CHAR_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 803  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 828  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 838  else Line 1036  else
1036  #endif  #endif
1037            }            }
1038    
1039  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
1040          if (c > (utf ? 0x10ffff : 0xff))          if (c > (utf ? 0x10ffffU : 0xffU))
1041  #else  #elif defined COMPILE_PCRE16
1042  #ifdef COMPILE_PCRE16          if (c > (utf ? 0x10ffffU : 0xffffU))
1043          if (c > (utf ? 0x10ffff : 0xffff))  #elif defined COMPILE_PCRE32
1044  #endif          if (utf && c > 0x10ffffU)
1045  #endif  #endif
1046            {            {
1047            *errorcodeptr = ERR76;            *errorcodeptr = ERR76;
# Line 875  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 890  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 910  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 934  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 942  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 972  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 1023  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 1048  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  #endif            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1327  #endif  #endif
1328          }            }
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 1119  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 1159  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 1189  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 1220  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 1245  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 1297  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 1312  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 1337  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 1670  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 1686  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 1704  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 1730  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 1810  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 1850  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 1859  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 1894  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 1984  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 2020  for (;;) Line 1995  for (;;)
1995    
1996    
1997    
   
1998  /*************************************************  /*************************************************
1999  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
2000  *************************************************/  *************************************************/
# Line 2033  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 2044  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 2067  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 2097  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 2120  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 2172  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 2182  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 2217  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 2240  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 2326  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 2364  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        else
2365          {
2366          recurse_check *r = recurses;
2367          const pcre_uchar *endgroup = scode;
2368    
2369          do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2370          if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2371    
2372      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)        for (r = recurses; r != NULL; r = r->prev)
2373        if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;          if (r->group == scode) break;
2374          if (r != NULL) continue;   /* Mutual recursion */
2375          }
2376    
2377      /* Not a forward reference, test for completed backward reference */      /* Completed reference; scan the referenced group, remembering it on the
2378        stack chain to detect mutual recursions. */
2379    
2380      empty_branch = FALSE;      empty_branch = FALSE;
2381      scode = cd->start_code + GET(code, 1);      this_recurse.prev = recurses;
2382      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      this_recurse.group = scode;
   
     /* Completed backwards reference */  
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 2438  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 2480  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 2496  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 2543  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 2557  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 2591  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 2625  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 2637  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 2647  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  *         Complete a callout item                *      code += IMM2_SIZE;
 *************************************************/  
2853    
2854  /* A callout item contains the length of the next item in the pattern, which    list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
 we can't fill in till after we have reached the relevant point. This is used  
 for both automatic and manual callouts.  
2855    
2856  Arguments:    switch(base)
2857    previous_callout   points to previous callout item      {
2858    ptr                current pattern pointer      case OP_STAR:
2859    cd                 pointers to tables etc      list[0] = OP_CHAR;
2860        break;
2861    
2862  Returns:             nothing      case OP_STARI:
2863  */      list[0] = OP_CHARI;
2864        break;
2865    
2866  static void      case OP_NOTSTAR:
2867  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)      list[0] = OP_NOT;
2868  {      break;
 int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));  
 PUT(previous_callout, 2 + LINK_SIZE, length);  
 }  
2869    
2870        case OP_NOTSTARI:
2871        list[0] = OP_NOTI;
2872        break;
2873    
2874        case OP_TYPESTAR:
2875        list[0] = *code;
2876        code++;
2877        break;
2878        }
2879      c = list[0];
2880      }
2881    
2882  #ifdef SUPPORT_UCP  switch(c)
2883  /*************************************************    {
2884  *           Get othercase range                  *    case OP_NOT_DIGIT:
2885  *************************************************/    case OP_DIGIT:
2886      case OP_NOT_WHITESPACE:
2887      case OP_WHITESPACE:
2888      case OP_NOT_WORDCHAR:
2889      case OP_WORDCHAR:
2890      case OP_ANY:
2891      case OP_ALLANY:
2892      case OP_ANYNL:
2893      case OP_NOT_HSPACE:
2894      case OP_HSPACE:
2895      case OP_NOT_VSPACE:
2896      case OP_VSPACE:
2897      case OP_EXTUNI:
2898      case OP_EODN:
2899      case OP_EOD:
2900      case OP_DOLL:
2901      case OP_DOLLM:
2902      return code;
2903    
2904  /* This function is passed the start and end of a class range, in UTF-8 mode    case OP_CHAR:
2905  with UCP support. It searches up the characters, looking for ranges of    case OP_NOT:
2906  characters in the "other" case. Each call returns the next one, updating the    GETCHARINCTEST(chr, code);
2907  start address. A character with multiple other cases is returned on its own    list[2] = chr;
2908  with a special return value.    list[3] = NOTACHAR;
2909      return code;
2910    
2911  Arguments:    case OP_CHARI:
2912    cptr        points to starting character value; updated    case OP_NOTI:
2913    d           end value    list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2914    ocptr       where to put start of othercase range    GETCHARINCTEST(chr, code);
2915    odptr       where to put end of othercase range    list[2] = chr;
2916    
2917  Yield:        -1 when no more  #ifdef SUPPORT_UCP
2918                 0 when a range is returned    if (chr < 128 || (chr < 256 && !utf))
2919                >0 the CASESET offset for char with multiple other cases      list[3] = fcc[chr];
2920                  in this case, ocptr contains the original    else
2921  */      list[3] = UCD_OTHERCASE(chr);
2922    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2923      list[3] = (chr < 256) ? fcc[chr] : chr;
2924    #else
2925      list[3] = fcc[chr];
2926    #endif
2927    
2928  static int    /* The othercase might be the same value. */
 get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,  
   unsigned int *odptr)  
 {  
 unsigned int c, othercase, next;  
 int co;  
2929    
2930  /* Find the first character that has an other case. If it has multiple other    if (chr == list[3])
2931  cases, return its case offset value. */      list[3] = NOTACHAR;
2932      else
2933        list[4] = NOTACHAR;
2934      return code;
2935    
2936  for (c = *cptr; c <= d; c++)  #ifdef SUPPORT_UCP
2937    {    case OP_PROP:
2938    if ((co = UCD_CASESET(c)) != 0)    case OP_NOTPROP:
2939      if (code[0] != PT_CLIST)
2940      {      {
2941      *ocptr = c++;   /* Character that has the set */      list[2] = code[0];
2942      *cptr = c;      /* Rest of input range */      list[3] = code[1];
2943      return co;      return code + 2;
2944      }      }
   if ((othercase = UCD_OTHERCASE(c)) != c) break;  
   }  
2945    
2946  if (c > d) return -1;  /* Reached end of range */    /* Convert only if we have enough space. */
2947    
2948  *ocptr = othercase;    clist_src = PRIV(ucd_caseless_sets) + code[1];
2949  next = othercase + 1;    clist_dest = list + 2;
2950      code += 2;
2951    
2952  for (++c; c <= d; c++)    do {
2953    {       if (clist_dest >= list + 8)
2954    if (UCD_OTHERCASE(c) != next) break;         {
2955    next++;         /* Early return if there is not enough space. This should never
2956    }         happen, since all clists are shorter than 5 character now. */
2957           list[2] = code[0];
2958           list[3] = code[1];
2959           return code;
2960           }
2961         *clist_dest++ = *clist_src;
2962         }
2963      while(*clist_src++ != NOTACHAR);
2964    
2965  *odptr = next - 1;     /* End of othercase range */    /* All characters are stored. The terminating NOTACHAR
2966  *cptr = c;             /* Rest of input range */    is copied form the clist itself. */
2967  return 0;  
2968      list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
2969      return code;
2970    #endif
2971    
2972      case OP_NCLASS:
2973      case OP_CLASS:
2974    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2975      case OP_XCLASS:
2976    
2977      if (c == OP_XCLASS)
2978        end = code + GET(code, 0);
2979      else
2980    #endif
2981        end = code + 32 / sizeof(pcre_uchar);
2982    
2983      switch(*end)
2984        {
2985        case OP_CRSTAR:
2986        case OP_CRMINSTAR:
2987        case OP_CRQUERY:
2988        case OP_CRMINQUERY:
2989        case OP_CRPOSSTAR:
2990        case OP_CRPOSQUERY:
2991        list[1] = TRUE;
2992        end++;
2993        break;
2994    
2995        case OP_CRPLUS:
2996        case OP_CRMINPLUS:
2997        case OP_CRPOSPLUS:
2998        end++;
2999        break;
3000    
3001        case OP_CRRANGE:
3002        case OP_CRMINRANGE:
3003        case OP_CRPOSRANGE:
3004        list[1] = (GET2(end, 1) == 0);
3005        end += 1 + 2 * IMM2_SIZE;
3006        break;
3007        }
3008      list[2] = end - code;
3009      return end;
3010      }
3011    return NULL;    /* Opcode not accepted */
3012  }  }
3013    
3014    
3015    
3016  /*************************************************  /*************************************************
3017  *        Check a character and a property        *  *    Scan further character sets for match       *
3018  *************************************************/  *************************************************/
3019    
3020  /* This function is called by check_auto_possessive() when a property item  /* Checks whether the base and the current opcode have a common character, in
3021  is adjacent to a fixed character.  which case the base cannot be possessified.
3022    
3023  Arguments:  Arguments:
3024    c            the character    code        points to the byte code
3025    ptype        the property type    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3026    pdata        the data for the type    cd          static compile data
3027    negated      TRUE if it's a negated property (\P or \p{^)    base_list   the data list of the base opcode
3028    
3029  Returns:       TRUE if auto-possessifying is OK  Returns:      TRUE if the auto-possessification is possible
3030  */  */
3031    
3032  static BOOL  static BOOL
3033  check_char_prop(int c, int ptype, int pdata, BOOL negated)  compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3034      const pcre_uint32* base_list, const pcre_uchar *base_end)
3035  {  {
3036  const ucd_record *prop = GET_UCD(c);  pcre_uchar c;
3037  switch(ptype)  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_bits;
3043    pcre_uint32 chr;
3044    
3045    /* Note: the base_list[1] contains whether the current opcode has greedy
3046    (represented by a non-zero value) quantifier. This is a different from
3047    other character type lists, which stores here that the character iterator
3048    matches to an empty string (also represented by a non-zero value). */
3049    
3050    for(;;)
3051    {    {
3052    case PT_LAMP:    c = *code;
   return (prop->chartype == ucp_Lu ||  
           prop->chartype == ucp_Ll ||  
           prop->chartype == ucp_Lt) == negated;  
3053    
3054    case PT_GC:    /* Skip over callouts */
   return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;  
3055    
3056    case PT_PC:    if (c == OP_CALLOUT)
3057    return (pdata == prop->chartype) == negated;      {
3058        code += PRIV(OP_lengths)[c];
3059        continue;
3060        }
3061    
3062    case PT_SC:    if (c == OP_ALT)
3063    return (pdata == prop->script) == negated;      {
3064        do code += GET(code, 1); while (*code == OP_ALT);
3065        c = *code;
3066        }
3067    
3068    /* These are specials */    switch(c)
3069        {
3070        case OP_END:
3071        case OP_KETRPOS:
3072        /* TRUE only in greedy case. The non-greedy case could be replaced by
3073        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3074        uses more memory, which we cannot get at this stage.) */
3075    
3076    case PT_ALNUM:      return base_list[1] != 0;
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||  
           PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;  
3077    
3078    case PT_SPACE:    /* Perl space */      case OP_KET:
3079    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||      /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3080            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)      it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3081            == negated;      cannot be converted to a possessive form. */
3082    
3083        if (base_list[1] == 0) return FALSE;
3084    
3085        switch(*(code - GET(code, 1)))
3086          {
3087          case OP_ASSERT:
3088          case OP_ASSERT_NOT:
3089          case OP_ASSERTBACK:
3090          case OP_ASSERTBACK_NOT:
3091          case OP_ONCE:
3092          case OP_ONCE_NC:
3093          /* Atomic sub-patterns and assertions can always auto-possessify their
3094          last iterator. */
3095          return TRUE;
3096          }
3097    
3098    case PT_PXSPACE:  /* POSIX space */      code += PRIV(OP_lengths)[c];
3099    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||      continue;
           c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||  
           c == CHAR_FF || c == CHAR_CR)  
           == negated;  
3100    
3101    case PT_WORD:      case OP_ONCE:
3102    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||      case OP_ONCE_NC:
3103            PRIV(ucp_gentype)[prop->chartype] == ucp_N ||      case OP_BRA:
3104            c == CHAR_UNDERSCORE) == negated;      case OP_CBRA:
3105    }      next_code = code;
3106  return FALSE;      do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
 }  
 #endif  /* SUPPORT_UCP */  
3107    
3108        /* We do not support repeated brackets, because they can lead to
3109        infinite recursion. */
3110    
3111        if (*next_code != OP_KET) return FALSE;
3112    
3113  /*************************************************      next_code = code + GET(code, 1);
3114  *     Check if auto-possessifying is possible    *      code += PRIV(OP_lengths)[c];
 *************************************************/  
3115    
3116  /* This function is called for unlimited repeats of certain items, to see      while (*next_code == OP_ALT)
3117  whether the next thing could possibly match the repeated item. If not, it makes        {
3118  sense to automatically possessify the repeated item.        if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3119          code = next_code + 1 + LINK_SIZE;
3120          next_code += GET(next_code, 1);
3121          }
3122        continue;
3123    
3124  Arguments:      case OP_BRAZERO:
3125    previous      pointer to the repeated opcode      case OP_BRAMINZERO:
   utf           TRUE in UTF-8 / UTF-16 mode  
   ptr           next character in pattern  
   options       options bits  
   cd            contains pointers to tables etc.  
3126    
3127  Returns:        TRUE if possessifying is wanted      next_code = code + 1;
3128  */      if (*next_code != OP_BRA && *next_code != OP_CBRA)
3129          return FALSE;
3130    
3131  static BOOL      do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
 check_auto_possessive(const pcre_uchar *previous, BOOL utf,  
   const pcre_uchar *ptr, int options, compile_data *cd)  
 {  
 pcre_int32 c, next;  
 int op_code = *previous++;  
3132    
3133  /* Skip whitespace and comments in extended mode */      /* We do not support repeated brackets, because they can lead to
3134        infinite recursion. */
3135        if (*next_code != OP_KET) return FALSE;
3136    
3137        /* The bracket content will be checked by the
3138        OP_BRA/OP_CBRA case above. */
3139        next_code += 1 + LINK_SIZE;
3140        if (!compare_opcodes(next_code, utf, cd, base_list, base_end)) return FALSE;
3141    
3142  if ((options & PCRE_EXTENDED) != 0)      code += PRIV(OP_lengths)[c];
3143    {      continue;
3144    for (;;)      }
3145    
3146      /* Check for a supported opcode, and load its properties. */
3147    
3148      code = get_chr_property_list(code, utf, cd->fcc, list);
3149      if (code == NULL) return FALSE;    /* Unsupported */
3150    
3151      /* If either opcode is a small character list, set pointers for comparing
3152      characters from that list with another list, or with a property. */
3153    
3154      if (base_list[0] == OP_CHAR)
3155      {      {
3156      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      chr_ptr = base_list + 2;
3157      if (*ptr == CHAR_NUMBER_SIGN)      list_ptr = list;
3158        }
3159      else if (list[0] == OP_CHAR)
3160        {
3161        chr_ptr = list + 2;
3162        list_ptr = base_list;
3163        }
3164    
3165      /* Some property combinations also acceptable. Unicode property opcodes are
3166      processed specially; the rest can be handled with a lookup table. */
3167    
3168      else
3169        {
3170        pcre_uint32 leftop, rightop;
3171    
3172        if (list[1] != 0) return FALSE;   /* Must match at least one character */
3173        leftop = base_list[0];
3174        rightop = list[0];
3175    
3176    #ifdef SUPPORT_UCP
3177        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3178        {        {
3179        ptr++;        if (rightop == OP_EOD) return TRUE;
3180        while (*ptr != 0)        if (rightop == OP_PROP || rightop == OP_NOTPROP)
3181          {          {
3182          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          int n;
3183          ptr++;          const pcre_uint8 *p;
3184  #ifdef SUPPORT_UTF          BOOL same = leftop == rightop;
3185          if (utf) FORWARDCHAR(ptr);          BOOL lisprop = leftop == OP_PROP;
3186  #endif          BOOL risprop = rightop == OP_PROP;
3187            BOOL bothprop = lisprop && risprop;
3188    
3189            /* There's a table that specifies how each combination is to be
3190            processed:
3191              0   Always return FALSE (never auto-possessify)
3192              1   Character groups are distinct (possessify if both are OP_PROP)
3193              2   Check character categories in the same group (general or particular)
3194              3   Return TRUE if the two opcodes are not the same
3195              ... see comments below
3196            */
3197    
3198            n = propposstab[base_list[2]][list[2]];
3199            switch(n)
3200              {
3201              case 0: return FALSE;
3202              case 1: return bothprop;
3203              case 2: return (base_list[3] == list[3]) != same;
3204              case 3: return !same;
3205    
3206              case 4:  /* Left general category, right particular category */
3207              return risprop && catposstab[base_list[3]][list[3]] == same;
3208    
3209              case 5:  /* Right general category, left particular category */
3210              return lisprop && catposstab[list[3]][base_list[3]] == same;
3211    
3212              /* This code is logically tricky. Think hard before fiddling with it.
3213              The posspropstab table has four entries per row. Each row relates to
3214              one of PCRE's special properties such as ALNUM or SPACE or WORD.
3215              Only WORD actually needs all four entries, but using repeats for the
3216              others means they can all use the same code below.
3217    
3218              The first two entries in each row are Unicode general categories, and
3219              apply always, because all the characters they include are part of the
3220              PCRE character set. The third and fourth entries are a general and a
3221              particular category, respectively, that include one or more relevant
3222              characters. One or the other is used, depending on whether the check
3223              is for a general or a particular category. However, in both cases the
3224              category contains more characters than the specials that are defined
3225              for the property being tested against. Therefore, it cannot be used
3226              in a NOTPROP case.
3227    
3228              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3229              Underscore is covered by ucp_P or ucp_Po. */
3230    
3231              case 6:  /* Left alphanum vs right general category */
3232              case 7:  /* Left space vs right general category */
3233              case 8:  /* Left word vs right general category */
3234              p = posspropstab[n-6];
3235              return risprop && lisprop ==
3236                (list[3] != p[0] &&
3237                 list[3] != p[1] &&
3238                (list[3] != p[2] || !lisprop));
3239    
3240              case 9:   /* Right alphanum vs left general category */
3241              case 10:  /* Right space vs left general category */
3242              case 11:  /* Right word vs left general category */
3243              p = posspropstab[n-9];
3244              return lisprop && risprop ==
3245                (base_list[3] != p[0] &&
3246                 base_list[3] != p[1] &&
3247                (base_list[3] != p[2] || !risprop));
3248    
3249              case 12:  /* Left alphanum vs right particular category */
3250              case 13:  /* Left space vs right particular category */
3251              case 14:  /* Left word vs right particular category */
3252              p = posspropstab[n-12];
3253              return risprop && lisprop ==
3254                (catposstab[p[0]][list[3]] &&
3255                 catposstab[p[1]][list[3]] &&
3256                (list[3] != p[3] || !lisprop));
3257    
3258              case 15:  /* Right alphanum vs left particular category */
3259              case 16:  /* Right space vs left particular category */
3260              case 17:  /* Right word vs left particular category */
3261              p = posspropstab[n-15];
3262              return lisprop && risprop ==
3263                (catposstab[p[0]][base_list[3]] &&
3264                 catposstab[p[1]][base_list[3]] &&
3265                (base_list[3] != p[3] || !risprop));
3266              }
3267          }          }
3268          return FALSE;
3269        }        }
     else break;  
     }  
   }  
3270    
3271  /* If the next item is one that we can handle, get its value. A non-negative      else
3272  value is a character, a negative value is an escape value. */  #endif  /* SUPPORT_UCP */
3273    
3274  if (*ptr == CHAR_BACKSLASH)      return leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3275    {             rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3276    int temperrorcode = 0;             autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3277    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;  
3278    
3279  /* Skip whitespace and comments in extended mode */    /* Control reaches here only if one of the items is a small character list.
3280      All characters are checked against the other side. */
3281    
3282  if ((options & PCRE_EXTENDED) != 0)    do
   {  
   for (;;)  
3283      {      {
3284      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      chr = *chr_ptr;
3285      if (*ptr == CHAR_NUMBER_SIGN)  
3286        switch(list_ptr[0])
3287        {        {
3288        ptr++;        case OP_CHAR:
3289        while (*ptr != 0)        ochr_ptr = list_ptr + 2;
3290          do
3291          {          {
3292          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (chr == *ochr_ptr) return FALSE;
3293          ptr++;          ochr_ptr++;
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
3294          }          }
3295        }        while(*ochr_ptr != NOTACHAR);
3296      else break;        break;
     }  
   }  
3297    
3298  /* If the next thing is itself optional, we have to give up. */        case OP_NOT:
3299          ochr_ptr = list_ptr + 2;
3300          do
3301            {
3302            if (chr == *ochr_ptr)
3303              break;
3304            ochr_ptr++;
3305            }
3306          while(*ochr_ptr != NOTACHAR);
3307          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3308          break;
3309    
3310  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||        /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3311    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)        set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
     return FALSE;  
3312    
3313  /* Now compare the next item with the previous opcode. First, handle cases when        case OP_DIGIT:
3314  the next item is a character. */        if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3315          break;
3316    
3317  if (next >= 0) switch(op_code)        case OP_NOT_DIGIT:
3318    {        if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3319    case OP_CHAR:        break;
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   return c != next;  
3320    
3321    /* For CHARI (caseless character) we must check the other case. If we have        case OP_WHITESPACE:
3322    Unicode property support, we can use it to test the other case of        if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3323    high-valued characters. */        break;
3324    
3325    case OP_CHARI:        case OP_NOT_WHITESPACE:
3326  #ifdef SUPPORT_UTF        if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3327    GETCHARTEST(c, previous);        break;
 #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 */  
3328    
3329    case OP_NOT:        case OP_WORDCHAR:
3330  #ifdef SUPPORT_UTF        if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3331    GETCHARTEST(c, previous);        break;
3332  #else  
3333    c = *previous;        case OP_NOT_WORDCHAR:
3334          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3335          break;
3336    
3337          case OP_HSPACE:
3338          switch(chr)
3339            {
3340            HSPACE_CASES: return FALSE;
3341            default: break;
3342            }
3343          break;
3344    
3345          case OP_NOT_HSPACE:
3346          switch(chr)
3347            {
3348            HSPACE_CASES: break;
3349            default: return FALSE;
3350            }
3351          break;
3352    
3353          case OP_ANYNL:
3354          case OP_VSPACE:
3355          switch(chr)
3356            {
3357            VSPACE_CASES: return FALSE;
3358            default: break;
3359            }
3360          break;
3361    
3362          case OP_NOT_VSPACE:
3363          switch(chr)
3364            {
3365            VSPACE_CASES: break;
3366            default: return FALSE;
3367            }
3368          break;
3369    
3370          case OP_DOLL:
3371          case OP_EODN:
3372          switch (chr)
3373            {
3374            case CHAR_CR:
3375            case CHAR_LF:
3376            case CHAR_VT:
3377            case CHAR_FF:
3378            case CHAR_NEL:
3379    #ifndef EBCDIC
3380            case 0x2028:
3381            case 0x2029:
3382    #endif  /* Not EBCDIC */
3383            return FALSE;
3384            }
3385          break;
3386    
3387          case OP_EOD:    /* Can always possessify before \z */
3388          break;
3389    
3390          case OP_PROP:
3391          case OP_NOTPROP:
3392          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3393                list_ptr[0] == OP_NOTPROP))
3394            return FALSE;
3395          break;
3396    
3397          case OP_NCLASS:
3398          if (chr > 255) return FALSE;
3399          /* Fall through */
3400    
3401          case OP_CLASS:
3402          if (chr > 255) break;
3403          class_bits = (pcre_uint8 *)((list_ptr == list ? code : base_end) - list_ptr[2]);
3404          if ((class_bits[chr >> 3] & (1 << (chr & 7))) != 0)
3405            return FALSE;
3406          break;
3407    
3408    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3409          case OP_XCLASS:
3410          if (list_ptr != list) return FALSE;   /* Class is first opcode */
3411          if (PRIV(xclass)(chr, code - list_ptr[2] + LINK_SIZE, utf))
3412            return FALSE;
3413          break;
3414  #endif  #endif
   return c == next;  
3415    
3416    case OP_NOTI:        default:
3417  #ifdef SUPPORT_UTF        return FALSE;
3418    GETCHARTEST(c, previous);        }
3419  #else  
3420    c = *previous;      chr_ptr++;
3421        }
3422      while(*chr_ptr != NOTACHAR);
3423    
3424      /* At least one character must be matched from this opcode. */
3425    
3426      if (list[1] == 0) return TRUE;
3427      }
3428    
3429    return FALSE;
3430    }
3431    
3432    
3433    
3434    /*************************************************
3435    *    Scan compiled regex for auto-possession     *
3436    *************************************************/
3437    
3438    /* Replaces single character iterations with their possessive alternatives
3439    if appropriate. This function modifies the compiled opcode!
3440    
3441    Arguments:
3442      code        points to start of the byte code
3443      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3444      cd          static compile data
3445    
3446    Returns:      nothing
3447    */
3448    
3449    static void
3450    auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3451    {
3452    register pcre_uchar c, d;
3453    const pcre_uchar *end;
3454    pcre_uchar *repeat_code;
3455    pcre_uint32 list[8];
3456    
3457    for (;;)
3458      {
3459      c = *code;
3460    
3461      if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3462        {
3463        c -= get_repeat_base(c) - OP_STAR;
3464        end = (c <= OP_MINUPTO) ?
3465          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3466        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3467    
3468        if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3469          {
3470          switch(c)
3471            {
3472            case OP_STAR:
3473            *code += OP_POSSTAR - OP_STAR;
3474            break;
3475    
3476            case OP_MINSTAR:
3477            *code += OP_POSSTAR - OP_MINSTAR;
3478            break;
3479    
3480            case OP_PLUS:
3481            *code += OP_POSPLUS - OP_PLUS;
3482            break;
3483    
3484            case OP_MINPLUS:
3485            *code += OP_POSPLUS - OP_MINPLUS;
3486            break;
3487    
3488            case OP_QUERY:
3489            *code += OP_POSQUERY - OP_QUERY;
3490            break;
3491    
3492            case OP_MINQUERY:
3493            *code += OP_POSQUERY - OP_MINQUERY;
3494            break;
3495    
3496            case OP_UPTO:
3497            *code += OP_POSUPTO - OP_UPTO;
3498            break;
3499    
3500            case OP_MINUPTO:
3501            *code += OP_MINUPTO - OP_UPTO;
3502            break;
3503            }
3504          }
3505        c = *code;
3506        }
3507      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3508        {
3509    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3510        if (c == OP_XCLASS)
3511          repeat_code = code + 1 + GET(code, 1);
3512        else
3513  #endif  #endif
3514    if (c == next) return TRUE;        repeat_code = code + 1 + (32 / sizeof(pcre_uchar));
3515  #ifdef SUPPORT_UTF  
3516    if (utf)      d = *repeat_code;
3517        if (d >= OP_CRSTAR && d <= OP_CRMINRANGE)
3518          {
3519          /* end must not be NULL. */
3520          end = get_chr_property_list(code, utf, cd->fcc, list);
3521    
3522          list[1] = d == OP_CRSTAR || d == OP_CRPLUS || d == OP_CRQUERY ||
3523            d == OP_CRRANGE;
3524    
3525          if (compare_opcodes(end, utf, cd, list, end))
3526            {
3527            switch (d)
3528              {
3529              case OP_CRSTAR:
3530              *repeat_code = OP_CRPOSSTAR;
3531              break;
3532    
3533              case OP_CRPLUS:
3534              *repeat_code = OP_CRPOSPLUS;
3535              break;
3536    
3537              case OP_CRQUERY:
3538              *repeat_code = OP_CRPOSQUERY;
3539              break;
3540    
3541              case OP_CRRANGE:
3542              *repeat_code = OP_CRPOSRANGE;
3543              break;
3544              }
3545            }
3546          }
3547        }
3548    
3549      switch(c)
3550      {      {
3551      unsigned int othercase;      case OP_END:
3552      if (next < 128) othercase = cd->fcc[next]; else      return;
3553  #ifdef SUPPORT_UCP  
3554      othercase = UCD_OTHERCASE((unsigned int)next);      case OP_TYPESTAR:
3555  #else      case OP_TYPEMINSTAR:
3556      othercase = NOTACHAR;      case OP_TYPEPLUS:
3557        case OP_TYPEMINPLUS:
3558        case OP_TYPEQUERY:
3559        case OP_TYPEMINQUERY:
3560        case OP_TYPEPOSSTAR:
3561        case OP_TYPEPOSPLUS:
3562        case OP_TYPEPOSQUERY:
3563        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3564        break;
3565    
3566        case OP_TYPEUPTO:
3567        case OP_TYPEMINUPTO:
3568        case OP_TYPEEXACT:
3569        case OP_TYPEPOSUPTO:
3570        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3571          code += 2;
3572        break;
3573    
3574    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3575        case OP_XCLASS:
3576        code += GET(code, 1);
3577        break;
3578  #endif  #endif
3579      return (unsigned int)c == othercase;  
3580        case OP_MARK:
3581        case OP_PRUNE_ARG:
3582        case OP_SKIP_ARG:
3583        case OP_THEN_ARG:
3584        code += code[1];
3585        break;
3586      }      }
3587    
3588      /* Add in the fixed length from the table */
3589    
3590      code += PRIV(OP_lengths)[c];
3591    
3592      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3593      a multi-byte character. The length in the table is a minimum, so we have to
3594      arrange to skip the extra bytes. */
3595    
3596    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3597      if (utf) switch(c)
3598        {
3599        case OP_CHAR:
3600        case OP_CHARI:
3601        case OP_NOT:
3602        case OP_NOTI:
3603        case OP_STAR:
3604        case OP_MINSTAR:
3605        case OP_PLUS:
3606        case OP_MINPLUS:
3607        case OP_QUERY:
3608        case OP_MINQUERY:
3609        case OP_UPTO:
3610        case OP_MINUPTO:
3611        case OP_EXACT:
3612        case OP_POSSTAR:
3613        case OP_POSPLUS:
3614        case OP_POSQUERY:
3615        case OP_POSUPTO:
3616        case OP_STARI:
3617        case OP_MINSTARI:
3618        case OP_PLUSI:
3619        case OP_MINPLUSI:
3620        case OP_QUERYI:
3621        case OP_MINQUERYI:
3622        case OP_UPTOI:
3623        case OP_MINUPTOI:
3624        case OP_EXACTI:
3625        case OP_POSSTARI:
3626        case OP_POSPLUSI:
3627        case OP_POSQUERYI:
3628        case OP_POSUPTOI:
3629        case OP_NOTSTAR:
3630        case OP_NOTMINSTAR:
3631        case OP_NOTPLUS:
3632        case OP_NOTMINPLUS:
3633        case OP_NOTQUERY:
3634        case OP_NOTMINQUERY:
3635        case OP_NOTUPTO:
3636        case OP_NOTMINUPTO:
3637        case OP_NOTEXACT:
3638        case OP_NOTPOSSTAR:
3639        case OP_NOTPOSPLUS:
3640        case OP_NOTPOSQUERY:
3641        case OP_NOTPOSUPTO:
3642        case OP_NOTSTARI:
3643        case OP_NOTMINSTARI:
3644        case OP_NOTPLUSI:
3645        case OP_NOTMINPLUSI:
3646        case OP_NOTQUERYI:
3647        case OP_NOTMINQUERYI:
3648        case OP_NOTUPTOI:
3649        case OP_NOTMINUPTOI:
3650        case OP_NOTEXACTI:
3651        case OP_NOTPOSSTARI:
3652        case OP_NOTPOSPLUSI:
3653        case OP_NOTPOSQUERYI:
3654        case OP_NOTPOSUPTOI:
3655        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3656        break;
3657        }
3658    #else
3659      (void)(utf);  /* Keep compiler happy by referencing function argument */
3660    #endif
3661      }
3662    }
3663    
3664    
3665    
3666    /*************************************************
3667    *           Check for POSIX class syntax         *
3668    *************************************************/
3669    
3670    /* This function is called when the sequence "[:" or "[." or "[=" is
3671    encountered in a character class. It checks whether this is followed by a
3672    sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3673    reach an unescaped ']' without the special preceding character, return FALSE.
3674    
3675    Originally, this function only recognized a sequence of letters between the
3676    terminators, but it seems that Perl recognizes any sequence of characters,
3677    though of course unknown POSIX names are subsequently rejected. Perl gives an
3678    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3679    didn't consider this to be a POSIX class. Likewise for [:1234:].
3680    
3681    The problem in trying to be exactly like Perl is in the handling of escapes. We
3682    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3683    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3684    below handles the special case of \], but does not try to do any other escape
3685    processing. This makes it different from Perl for cases such as [:l\ower:]
3686    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3687    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
3688    I think.
3689    
3690    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3691    It seems that the appearance of a nested POSIX class supersedes an apparent
3692    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3693    a digit.
3694    
3695    In Perl, unescaped square brackets may also appear as part of class names. For
3696    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3697    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3698    seem right at all. PCRE does not allow closing square brackets in POSIX class
3699    names.
3700    
3701    Arguments:
3702      ptr      pointer to the initial [
3703      endptr   where to return the end pointer
3704    
3705    Returns:   TRUE or FALSE
3706    */
3707    
3708    static BOOL
3709    check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3710    {
3711    pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
3712    terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
3713    for (++ptr; *ptr != CHAR_NULL; ptr++)
3714      {
3715      if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3716        ptr++;
3717      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3718    else    else
3719  #endif  /* SUPPORT_UTF */      {
3720    return (c == TABLE_GET((unsigned int)next, cd->fcc, next));  /* Non-UTF-8 mode */      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3721          {
3722          *endptr = ptr;
3723          return TRUE;
3724          }
3725        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3726             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3727              ptr[1] == CHAR_EQUALS_SIGN) &&
3728            check_posix_syntax(ptr, endptr))
3729          return FALSE;
3730        }
3731      }
3732    return FALSE;
3733    }
3734    
   /* 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. */  
3735    
   case OP_DIGIT:  
   return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;  
3736    
   case OP_NOT_DIGIT:  
   return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;  
3737    
3738    case OP_WHITESPACE:  /*************************************************
3739    return next > 255 || (cd->ctypes[next] & ctype_space) == 0;  *          Check POSIX class name                *
3740    *************************************************/
3741    
3742    case OP_NOT_WHITESPACE:  /* This function is called to check the name given in a POSIX-style class entry
3743    return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;  such as [:alnum:].
3744    
3745    case OP_WORDCHAR:  Arguments:
3746    return next > 255 || (cd->ctypes[next] & ctype_word) == 0;    ptr        points to the first letter
3747      len        the length of the name
3748    
3749    case OP_NOT_WORDCHAR:  Returns:     a value representing the name, or -1 if unknown
3750    return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;  */
3751    
3752    static int
3753    check_posix_name(const pcre_uchar *ptr, int len)
3754    {
3755    const char *pn = posix_names;
3756    register int yield = 0;
3757    while (posix_name_lengths[yield] != 0)
3758      {
3759      if (len == posix_name_lengths[yield] &&
3760        STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3761      pn += posix_name_lengths[yield] + 1;
3762      yield++;
3763      }
3764    return -1;
3765    }
3766    
3767    
3768    /*************************************************
3769    *    Adjust OP_RECURSE items in repeated group   *
3770    *************************************************/
3771    
3772    /* OP_RECURSE items contain an offset from the start of the regex to the group
3773    that is referenced. This means that groups can be replicated for fixed
3774    repetition simply by copying (because the recursion is allowed to refer to
3775    earlier groups that are outside the current group). However, when a group is
3776    optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3777    inserted before it, after it has been compiled. This means that any OP_RECURSE
3778    items within it that refer to the group itself or any contained groups have to
3779    have their offsets adjusted. That one of the jobs of this function. Before it
3780    is called, the partially compiled regex must be temporarily terminated with
3781    OP_END.
3782    
3783    This function has been extended with the possibility of forward references for
3784    recursions and subroutine calls. It must also check the list of such references
3785    for the group we are dealing with. If it finds that one of the recursions in
3786    the current group is on this list, it adjusts the offset in the list, not the
3787    value in the reference (which is a group number).
3788    
3789    Arguments:
3790      group      points to the start of the group
3791      adjust     the amount by which the group is to be moved
3792      utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
3793      cd         contains pointers to tables etc.
3794      save_hwm   the hwm forward reference pointer at the start of the group
3795    
3796    Returns:     nothing
3797    */
3798    
3799    static void
3800    adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3801      pcre_uchar *save_hwm)
3802    {
3803    pcre_uchar *ptr = group;
3804    
3805    while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3806      {
3807      int offset;
3808      pcre_uchar *hc;
3809    
3810      /* See if this recursion is on the forward reference list. If so, adjust the
3811      reference. */
3812    
3813    case OP_HSPACE:    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
   case OP_NOT_HSPACE:  
   switch(next)  
3814      {      {
3815      HSPACE_CASES:      offset = (int)GET(hc, 0);
3816      return op_code == OP_NOT_HSPACE;      if (cd->start_code + offset == ptr + 1)
3817          {
3818      default:        PUT(hc, 0, offset + adjust);
3819      return op_code != OP_NOT_HSPACE;        break;
3820          }
3821      }      }
3822    
3823    case OP_ANYNL:    /* Otherwise, adjust the recursion offset if it's after the start of this
3824    case OP_VSPACE:    group. */
3825    case OP_NOT_VSPACE:  
3826    switch(next)    if (hc >= cd->hwm)
3827      {      {
3828      VSPACE_CASES:      offset = (int)GET(ptr, 1);
3829      return op_code == OP_NOT_VSPACE;      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
   
     default:  
     return op_code != OP_NOT_VSPACE;  
3830      }      }
3831    
3832  #ifdef SUPPORT_UCP    ptr += 1 + LINK_SIZE;
   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;  
3833    }    }
3834    }
3835    
3836    
 /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP  
 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:  
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   switch(-next)  
     {  
     case ESC_d:  
     return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;  
   
     case ESC_D:  
     return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;  
   
     case ESC_s:  
     return c > 255 || (cd->ctypes[c] & ctype_space) == 0;  
   
     case ESC_S:  
     return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;  
   
     case ESC_w:  
     return c > 255 || (cd->ctypes[c] & ctype_word) == 0;  
3837    
3838      case ESC_W:  /*************************************************
3839      return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;  *        Insert an automatic callout point       *
3840    *************************************************/
3841    
3842      case ESC_h:  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
3843      case ESC_H:  callout points before each pattern item.
     switch(c)  
       {  
       HSPACE_CASES:  
       return -next != ESC_h;  
   
       default:  
       return -next == ESC_h;  
       }  
3844    
3845      case ESC_v:  Arguments:
3846      case ESC_V:    code           current code pointer
3847      switch(c)    ptr            current pattern pointer
3848        {    cd             pointers to tables etc
       VSPACE_CASES:  
       return -next != ESC_v;  
   
       default:  
       return -next == ESC_v;  
       }  
3849    
3850      /* When PCRE_UCP is set, these values get generated for \d etc. Find  Returns:         new code pointer
3851      their substitutions and process them. The result will always be either  */
     -ESC_p or -ESC_P. Then fall through to process those values. */  
3852    
3853  #ifdef SUPPORT_UCP  static pcre_uchar *
3854      case ESC_du:  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
3855      case ESC_DU:  {
3856      case ESC_wu:  *code++ = OP_CALLOUT;
3857      case ESC_WU:  *code++ = 255;
3858      case ESC_su:  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
3859      case ESC_SU:  PUT(code, LINK_SIZE, 0);                       /* Default length */
3860        {  return code + 2 * LINK_SIZE;
3861        int temperrorcode = 0;  }
       ptr = substitutes[-next - ESC_DU];  
       next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);  
       if (temperrorcode != 0) return FALSE;  
       ptr++;    /* For compatibility */  
       }  
     /* Fall through */  
3862    
     case ESC_p:  
     case ESC_P:  
       {  
       int ptype, pdata, errorcodeptr;  
       BOOL negated;  
3863    
       ptr--;      /* Make ptr point at the p or P */  
       ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);  
       if (ptype < 0) return FALSE;  
       ptr++;      /* Point past the final curly ket */  
3864    
3865        /* If the property item is optional, we have to give up. (When generated  /*************************************************
3866        from \d etc by PCRE_UCP, this test will have been applied much earlier,  *         Complete a callout item                *
3867        to the original \d etc. At this point, ptr will point to a zero byte. */  *************************************************/
3868    
3869        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  /* A callout item contains the length of the next item in the pattern, which
3870          STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)  we can't fill in till after we have reached the relevant point. This is used
3871            return FALSE;  for both automatic and manual callouts.
3872    
3873        /* Do the property check. */  Arguments:
3874      previous_callout   points to previous callout item
3875      ptr                current pattern pointer
3876      cd                 pointers to tables etc
3877    
3878        return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);  Returns:             nothing
3879        }  */
 #endif  
3880    
3881      default:  static void
3882      return FALSE;  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
3883      }  {
3884    int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
3885    PUT(previous_callout, 2 + LINK_SIZE, length);
3886    }
3887    
   /* In principle, support for Unicode properties should be integrated here as  
   well. It means re-organizing the above code so as to get hold of the property  
   values before switching on the op-code. However, I wonder how many patterns  
   combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,  
   these op-codes are never generated.) */  
3888    
   case OP_DIGIT:  
   return next == -ESC_D || next == -ESC_s || next == -ESC_W ||  
          next == -ESC_h || next == -ESC_v || next == -ESC_R;  
3889    
3890    case OP_NOT_DIGIT:  #ifdef SUPPORT_UCP
3891    return next == -ESC_d;  /*************************************************
3892    *           Get othercase range                  *
3893    *************************************************/
3894    
3895    case OP_WHITESPACE:  /* This function is passed the start and end of a class range, in UTF-8 mode
3896    return next == -ESC_S || next == -ESC_d || next == -ESC_w;  with UCP support. It searches up the characters, looking for ranges of
3897    characters in the "other" case. Each call returns the next one, updating the
3898    start address. A character with multiple other cases is returned on its own
3899    with a special return value.
3900    
3901    case OP_NOT_WHITESPACE:  Arguments:
3902    return next == -ESC_s || next == -ESC_h || next == -ESC_v || next == -ESC_R;    cptr        points to starting character value; updated
3903      d           end value
3904      ocptr       where to put start of othercase range
3905      odptr       where to put end of othercase range
3906    
3907    case OP_HSPACE:  Yield:        -1 when no more
3908    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||                 0 when a range is returned
3909           next == -ESC_w || next == -ESC_v || next == -ESC_R;                >0 the CASESET offset for char with multiple other cases
3910                    in this case, ocptr contains the original
3911    */
3912    
3913    case OP_NOT_HSPACE:  static int
3914    return next == -ESC_h;  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
3915      pcre_uint32 *odptr)
3916    {
3917    pcre_uint32 c, othercase, next;
3918    unsigned int co;
3919    
3920    /* Can't have \S in here because VT matches \S (Perl anomaly) */  /* Find the first character that has an other case. If it has multiple other
3921    case OP_ANYNL:  cases, return its case offset value. */
   case OP_VSPACE:  
   return next == -ESC_V || next == -ESC_d || next == -ESC_w;  
3922    
3923    case OP_NOT_VSPACE:  for (c = *cptr; c <= d; c++)
3924    return next == -ESC_v || next == -ESC_R;    {
3925      if ((co = UCD_CASESET(c)) != 0)
3926        {
3927        *ocptr = c++;   /* Character that has the set */
3928        *cptr = c;      /* Rest of input range */
3929        return (int)co;
3930        }
3931      if ((othercase = UCD_OTHERCASE(c)) != c) break;
3932      }
3933    
3934    case OP_WORDCHAR:  if (c > d) return -1;  /* Reached end of range */
   return next == -ESC_W || next == -ESC_s || next == -ESC_h ||  
          next == -ESC_v || next == -ESC_R;  
3935    
3936    case OP_NOT_WORDCHAR:  *ocptr = othercase;
3937    return next == -ESC_w || next == -ESC_d;  next = othercase + 1;
3938    
3939    default:  for (++c; c <= d; c++)
3940    return FALSE;    {
3941      if (UCD_OTHERCASE(c) != next) break;
3942      next++;
3943    }    }
3944    
3945  /* Control does not reach here */  *odptr = next - 1;     /* End of othercase range */
3946    *cptr = c;             /* Rest of input range */
3947    return 0;
3948  }  }
3949    #endif  /* SUPPORT_UCP */
3950    
3951    
3952    
# Line 3381  switch(op_code) Line 3955  switch(op_code)
3955  *************************************************/  *************************************************/
3956    
3957  /* This function packages up the logic of adding a character or range of  /* This function packages up the logic of adding a character or range of
3958  characters to a class. The character values in the arguments will be within the  characters to a class. The character values in the arguments will be within the
3959  valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is  valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
3960  mutually recursive with the function immediately below.  mutually recursive with the function immediately below.
3961    
3962  Arguments:  Arguments:
3963    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
3964    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
3965    options       the options word    options       the options word
3966    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
3967    start         start of range character    start         start of range character
3968    end           end of range character    end           end of range character
3969    
3970  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
3971                  the pointer to extra data is updated                  the pointer to extra data is updated
3972  */  */
3973    
3974  static int  static int
3975  add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,  add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
3976    compile_data *cd, unsigned int start, unsigned int end)    compile_data *cd, pcre_uint32 start, pcre_uint32 end)
3977  {  {
3978  unsigned int c;  pcre_uint32 c;
3979  int n8 = 0;  int n8 = 0;
3980    
3981  /* If caseless matching is required, scan the range and process alternate  /* If caseless matching is required, scan the range and process alternate
3982  cases. In Unicode, there are 8-bit characters that have alternate cases that  cases. In Unicode, there are 8-bit characters that have alternate cases that
3983  are greater than 255 and vice-versa. Sometimes we can just extend the original  are greater than 255 and vice-versa. Sometimes we can just extend the original
3984  range. */  range. */
3985    
3986  if ((options & PCRE_CASELESS) != 0)  if ((options & PCRE_CASELESS) != 0)
3987    {    {
3988  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3989    if ((options & PCRE_UTF8) != 0)    if ((options & PCRE_UTF8) != 0)
3990      {      {
3991      int rc;      int rc;
3992      unsigned int oc, od;      pcre_uint32 oc, od;
3993    
3994      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
3995      c = start;      c = start;
3996    
3997      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
3998        {        {
3999        /* Handle a single character that has more than one other case. */        /* Handle a single character that has more than one other case. */
4000    
4001        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4002          PRIV(ucd_caseless_sets) + rc, oc);          PRIV(ucd_caseless_sets) + rc, oc);
4003    
4004        /* Do nothing if the other case range is within the original range. */        /* Do nothing if the other case range is within the original range. */
4005    
4006        else if (oc >= start && od <= end) continue;        else if (oc >= start && od <= end) continue;
4007    
4008        /* Extend the original range if there is overlap, noting that if oc < c, we        /* Extend the original range if there is overlap, noting that if oc < c, we
4009        can't have od > end because a subrange is always shorter than the basic        can't have od > end because a subrange is always shorter than the basic
4010        range. Otherwise, use a recursive call to add the additional range. */        range. Otherwise, use a recursive call to add the additional range. */
4011    
4012        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4013        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
4014        else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);        else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
# Line 3444  if ((options & PCRE_CASELESS) != 0) Line 4018  if ((options & PCRE_CASELESS) != 0)
4018  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
4019    
4020    /* Not UTF-mode, or no UCP */    /* Not UTF-mode, or no UCP */
4021    
4022    for (c = start; c <= end && c < 256; c++)    for (c = start; c <= end && c < 256; c++)
4023      {      {
4024      SETBIT(classbits, cd->fcc[c]);      SETBIT(classbits, cd->fcc[c]);
4025      n8++;      n8++;
4026      }      }
4027    }    }
4028    
4029  /* Now handle the original range. Adjust the final value according to the bit  /* Now handle the original range. Adjust the final value according to the bit
4030  length - this means that the same lists of (e.g.) horizontal spaces can be used  length - this means that the same lists of (e.g.) horizontal spaces can be used
4031  in all cases. */  in all cases. */
4032    
4033  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
4034  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4035    if ((options & PCRE_UTF8) == 0)    if ((options & PCRE_UTF8) == 0)
4036  #endif  #endif
4037    if (end > 0xff) end = 0xff;    if (end > 0xff) end = 0xff;
 #endif  
4038    
4039  #ifdef COMPILE_PCRE16  #elif defined COMPILE_PCRE16
4040  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4041    if ((options & PCRE_UTF16) == 0)    if ((options & PCRE_UTF16) == 0)
4042  #endif  #endif
4043    if (end > 0xffff) end = 0xffff;    if (end > 0xffff) end = 0xffff;
4044  #endif  
4045    #endif /* COMPILE_PCRE[8|16] */
4046    
4047  /* If all characters are less than 256, use the bit map. Otherwise use extra  /* If all characters are less than 256, use the bit map. Otherwise use extra
4048  data. */  data. */
# Line 3477  if (end < 0x100) Line 4051  if (end < 0x100)
4051    {    {
4052    for (c = start; c <= end; c++)    for (c = start; c <= end; c++)
4053      {      {
4054      n8++;      n8++;
4055      SETBIT(classbits, c);      SETBIT(classbits, c);
4056      }      }
4057    }    }
4058    
4059  else  else
4060    {    {
4061    pcre_uchar *uchardata = *uchardptr;    pcre_uchar *uchardata = *uchardptr;
4062    
4063  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4064    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
4065      {      {
4066      if (start < end)      if (start < end)
4067        {        {
4068        *uchardata++ = XCL_RANGE;        *uchardata++ = XCL_RANGE;
4069        uchardata += PRIV(ord2utf)(start, uchardata);        uchardata += PRIV(ord2utf)(start, uchardata);
4070        uchardata += PRIV(ord2utf)(end, uchardata);        uchardata += PRIV(ord2utf)(end, uchardata);
4071        }        }
4072      else if (start == end)      else if (start == end)
4073        {        {
4074        *uchardata++ = XCL_SINGLE;        *uchardata++ = XCL_SINGLE;
4075        uchardata += PRIV(ord2utf)(start, uchardata);        uchardata += PRIV(ord2utf)(start, uchardata);
4076        }        }
4077      }      }
4078    else    else
4079  #endif  /* SUPPORT_UTF */  #endif  /* SUPPORT_UTF */
4080    
4081    /* Without UTF support, character values are constrained by the bit length,    /* Without UTF support, character values are constrained by the bit length,
4082    and can only be > 256 for 16-bit and 32-bit libraries. */    and can only be > 256 for 16-bit and 32-bit libraries. */
4083    
4084  #ifdef COMPILE_PCRE8  #ifdef COMPILE_PCRE8
4085      {}      {}
4086  #else  #else
4087    if (start < end)    if (start < end)
4088      {      {
4089      *uchardata++ = XCL_RANGE;      *uchardata++ = XCL_RANGE;
# Line 3520  else Line 4094  else
4094      {      {
4095      *uchardata++ = XCL_SINGLE;      *uchardata++ = XCL_SINGLE;
4096      *uchardata++ = start;      *uchardata++ = start;
4097      }      }
4098  #endif  #endif
4099    
4100    *uchardptr = uchardata;   /* Updata extra data pointer */    *uchardptr = uchardata;   /* Updata extra data pointer */
4101    }    }
4102    
4103  return n8;    /* Number of 8-bit characters */  return n8;    /* Number of 8-bit characters */
4104  }  }
4105    
4106    
4107    
4108    
4109  /*************************************************  /*************************************************
4110  *        Add a list of characters to a class     *  *        Add a list of characters to a class     *
4111  *************************************************/  *************************************************/
4112    
4113  /* This function is used for adding a list of case-equivalent characters to a  /* This function is used for adding a list of case-equivalent characters to a
4114  class, and also for adding a list of horizontal or vertical whitespace. If the  class, and also for adding a list of horizontal or vertical whitespace. If the
4115  list is in order (which it should be), ranges of characters are detected and  list is in order (which it should be), ranges of characters are detected and
4116  handled appropriately. This function is mutually recursive with the function  handled appropriately. This function is mutually recursive with the function
# Line 3546  Arguments: Line 4120  Arguments:
4120    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4121    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4122    options       the options word    options       the options word
4123    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4124    p             points to row of 32-bit values, terminated by NOTACHAR    p             points to row of 32-bit values, terminated by NOTACHAR
4125    except        character to omit; this is used when adding lists of    except        character to omit; this is used when adding lists of
4126                    case-equivalent characters to avoid including the one we                    case-equivalent characters to avoid including the one we
4127                    already know about                    already know about
4128    
4129  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4130                  the pointer to extra data is updated                  the pointer to extra data is updated
4131  */  */
# Line 3565  while (p[0] < NOTACHAR) Line 4139  while (p[0] < NOTACHAR)
4139    {    {
4140    int n = 0;    int n = 0;
4141    if (p[0] != except)    if (p[0] != except)
4142      {      {
4143      while(p[n+1] == p[0] + n + 1) n++;      while(p[n+1] == p[0] + n + 1) n++;
4144      n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);      n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4145      }      }
4146    p += n + 1;    p += n + 1;
4147    }    }
4148  return n8;  return n8;
4149  }  }
4150    
4151    
4152    
# Line 3587  Arguments: Line 4161  Arguments:
4161    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4162    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4163    options       the options word    options       the options word
4164    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4165    p             points to row of 32-bit values, terminated by NOTACHAR    p             points to row of 32-bit values, terminated by NOTACHAR
4166    
4167  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4168                  the pointer to extra data is updated                  the pointer to extra data is updated
4169  */  */
4170    
4171  static int  static int
4172  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
4173    int options, compile_data *cd, const pcre_uint32 *p)    int options, compile_data *cd, const pcre_uint32 *p)
4174  {  {
4175    BOOL utf = (options & PCRE_UTF8) != 0;
4176  int n8 = 0;  int n8 = 0;
4177  if (p[0] > 0)  if (p[0] > 0)
4178    n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);    n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
4179  while (p[0] < NOTACHAR)  while (p[0] < NOTACHAR)
4180    {    {
4181    while (p[1] == p[0] + 1) p++;    while (p[1] == p[0] + 1) p++;
4182    n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,    n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
4183      (p[1] == NOTACHAR)? 0x10ffff : p[1] - 1);      (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
4184    p++;    p++;
4185    }    }
4186  return n8;  return n8;
4187  }  }
4188    
4189    
4190    
# Line 3624  to find out the amount of memory needed, Line 4199  to find out the amount of memory needed,
4199  phase. The value of lengthptr distinguishes the two phases.  phase. The value of lengthptr distinguishes the two phases.
4200    
4201  Arguments:  Arguments:
4202    optionsptr     pointer to the option bits    optionsptr        pointer to the option bits
4203    codeptr        points to the pointer to the current code point    codeptr           points to the pointer to the current code point
4204    ptrptr         points to the current pattern pointer    ptrptr            points to the current pattern pointer
4205    errorcodeptr   points to error code variable    errorcodeptr      points to error code variable
4206    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr      place to put the first required character
4207    reqcharptr     set to the last literal character required, else < 0    firstcharflagsptr place to put the first character flags, or a negative number
4208    bcptr          points to current branch chain    reqcharptr        place to put the last required character
4209    cond_depth     conditional nesting depth    reqcharflagsptr   place to put the last required character flags, or a negative number
4210    cd             contains pointers to tables etc.    bcptr             points to current branch chain
4211    lengthptr      NULL during the real compile phase    cond_depth        conditional nesting depth
4212                   points to length accumulator during pre-compile phase    cd                contains pointers to tables etc.
4213      lengthptr         NULL during the real compile phase
4214                        points to length accumulator during pre-compile phase
4215    
4216  Returns:         TRUE on success  Returns:            TRUE on success
4217                   FALSE, with *errorcodeptr set non-zero on error                      FALSE, with *errorcodeptr set non-zero on error
4218  */  */
4219    
4220  static BOOL  static BOOL
4221  compile_branch(int *optionsptr, pcre_uchar **codeptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
4222    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,    const pcre_uchar **ptrptr, int *errorcodeptr,
4223    pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,    pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
4224      pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
4225      branch_chain *bcptr, int cond_depth,
4226    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
4227  {  {
4228  int repeat_type, op_type;  int repeat_type, op_type;
4229  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
4230  int bravalue = 0;  int bravalue = 0;
4231  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
4232  pcre_int32 firstchar, reqchar;  pcre_uint32 firstchar, reqchar;
4233  pcre_int32 zeroreqchar, zerofirstchar;  pcre_int32 firstcharflags, reqcharflags;
4234    pcre_uint32 zeroreqchar, zerofirstchar;
4235    pcre_int32 zeroreqcharflags, zerofirstcharflags;
4236  pcre_int32 req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
4237  int options = *optionsptr;               /* May change dynamically */  int options = *optionsptr;               /* May change dynamically */
4238  int after_manual_callout = 0;  int after_manual_callout = 0;
4239  int length_prevgroup = 0;  int length_prevgroup = 0;
4240  register int c;  register pcre_uint32 c;
4241    int escape;
4242  register pcre_uchar *code = *codeptr;  register pcre_uchar *code = *codeptr;
4243  pcre_uchar *last_code = code;  pcre_uchar *last_code = code;
4244  pcre_uchar *orig_code = code;  pcre_uchar *orig_code = code;
# Line 3676  must not do this for other options (e.g. Line 4258  must not do this for other options (e.g.
4258  dynamically as we process the pattern. */  dynamically as we process the pattern. */
4259    
4260  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4261  /* PCRE_UTF16 has the same value as PCRE_UTF8. */  /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
4262  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
4263    #ifndef COMPILE_PCRE32
4264  pcre_uchar utf_chars[6];  pcre_uchar utf_chars[6];
4265    #endif
4266  #else  #else
4267  BOOL utf = FALSE;  BOOL utf = FALSE;
4268  #endif  #endif
4269    
4270  /* Helper variables for OP_XCLASS opcode (for characters > 255). */  /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
4271    class_uchardata always so that it can be passed to add_to_class() always,
4272    though it will not be used in non-UTF 8-bit cases. This avoids having to supply
4273    alternative calls for the different cases. */
4274    
4275    pcre_uchar *class_uchardata;
4276  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4277  BOOL xclass;  BOOL xclass;
 pcre_uchar *class_uchardata;  
4278  pcre_uchar *class_uchardata_base;  pcre_uchar *class_uchardata_base;
4279  #endif  #endif
4280    
# Line 3710  to take the zero repeat into account. Th Line 4297  to take the zero repeat into account. Th
4297  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
4298  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
4299    
4300  firstchar = reqchar = zerofirstchar = zeroreqchar = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
4301    firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
4302    
4303  /* The variable req_caseopt contains either the REQ_CASELESS value  /* The variable req_caseopt contains either the REQ_CASELESS value
4304  or zero, according to the current setting of the caseless flag. The  or zero, according to the current setting of the caseless flag. The
# Line 3736  for (;; ptr++) Line 4324  for (;; ptr++)
4324    int recno;    int recno;
4325    int refsign;    int refsign;
4326    int skipbytes;    int skipbytes;
4327    int subreqchar;    pcre_uint32 subreqchar, subfirstchar;
4328    int subfirstchar;    pcre_int32 subreqcharflags, subfirstcharflags;
4329    int terminator;    int terminator;
4330    int mclength;    unsigned int mclength;
4331    int tempbracount;    unsigned int tempbracount;
4332      pcre_uint32 ec;
4333    pcre_uchar mcbuffer[8];    pcre_uchar mcbuffer[8];
4334    
4335    /* Get next character in the pattern */    /* Get next character in the pattern */
# Line 3750  for (;; ptr++) Line 4339  for (;; ptr++)
4339    /* If we are at the end of a nested substitution, revert to the outer level    /* If we are at the end of a nested substitution, revert to the outer level
4340    string. Nesting only happens one level deep. */    string. Nesting only happens one level deep. */
4341    
4342    if (c == 0 && nestptr != NULL)    if (c == CHAR_NULL && nestptr != NULL)
4343      {      {
4344      ptr = nestptr;      ptr = nestptr;
4345      nestptr = NULL;      nestptr = NULL;
# Line 3825  for (;; ptr++) Line 4414  for (;; ptr++)
4414    
4415    /* If in \Q...\E, check for the end; if not, we have a literal */    /* If in \Q...\E, check for the end; if not, we have a literal */
4416    
4417    if (inescq && c != 0)    if (inescq && c != CHAR_NULL)
4418      {      {
4419      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4420        {        {
# Line 3850  for (;; ptr++) Line 4439  for (;; ptr++)
4439        }        }
4440      }      }
4441    
   /* Fill in length of a previous callout, except when the next thing is  
   a quantifier. */  
   
4442    is_quantifier =    is_quantifier =
4443      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4444      (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));      (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4445    
4446    if (!is_quantifier && previous_callout != NULL &&    /* Fill in length of a previous callout, except when the next thing is a
4447      quantifier or when processing a property substitution string in UCP mode. */
4448    
4449      if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4450         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
4451      {      {
4452      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
# Line 3873  for (;; ptr++) Line 4462  for (;; ptr++)
4462      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
4463        {        {
4464        ptr++;        ptr++;
4465        while (*ptr != 0)        while (*ptr != CHAR_NULL)
4466          {          {
4467          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
4468          ptr++;          ptr++;
# Line 3881  for (;; ptr++) Line 4470  for (;; ptr++)
4470          if (utf) FORWARDCHAR(ptr);          if (utf) FORWARDCHAR(ptr);
4471  #endif  #endif
4472          }          }
4473        if (*ptr != 0) continue;        if (*ptr != CHAR_NULL) continue;
4474    
4475        /* Else fall through to handle end of string */        /* Else fall through to handle end of string */
4476        c = 0;        c = 0;
4477        }        }
4478      }      }
4479    
4480    /* No auto callout for quantifiers. */    /* No auto callout for quantifiers, or while processing property strings that
4481      are substituted for \w etc in UCP mode. */
4482    
4483    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4484      {      {
4485      previous_callout = code;      previous_callout = code;
4486      code = auto_callout(code, ptr, cd);      code = auto_callout(code, ptr, cd);
# Line 3903  for (;; ptr++) Line 4493  for (;; ptr++)
4493      case CHAR_VERTICAL_LINE:       /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
4494      case CHAR_RIGHT_PARENTHESIS:      case CHAR_RIGHT_PARENTHESIS:
4495      *firstcharptr = firstchar;      *firstcharptr = firstchar;
4496        *firstcharflagsptr = firstcharflags;
4497      *reqcharptr = reqchar;      *reqcharptr = reqchar;
4498        *reqcharflagsptr = reqcharflags;
4499      *codeptr = code;      *codeptr = code;
4500      *ptrptr = ptr;      *ptrptr = ptr;
4501      if (lengthptr != NULL)      if (lengthptr != NULL)
# Line 3927  for (;; ptr++) Line 4519  for (;; ptr++)
4519      previous = NULL;      previous = NULL;
4520      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
4521        {        {
4522        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;        if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4523        *code++ = OP_CIRCM;        *code++ = OP_CIRCM;
4524        }        }
4525      else *code++ = OP_CIRC;      else *code++ = OP_CIRC;
# Line 3942  for (;; ptr++) Line 4534  for (;; ptr++)
4534      repeats. The value of reqchar doesn't change either. */      repeats. The value of reqchar doesn't change either. */
4535    
4536      case CHAR_DOT:      case CHAR_DOT:
4537      if (firstchar == REQ_UNSET) firstchar = REQ_NONE;      if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4538      zerofirstchar = firstchar;      zerofirstchar = firstchar;
4539        zerofirstcharflags = firstcharflags;
4540      zeroreqchar = reqchar;      zeroreqchar = reqchar;
4541        zeroreqcharflags = reqcharflags;
4542      previous = code;      previous = code;
4543      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
4544      break;      break;
# Line 4018  for (;; ptr++) Line 4612  for (;; ptr++)
4612          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4613        {        {
4614        *code++ = negate_class? OP_ALLANY : OP_FAIL;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
4615        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;        if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4616        zerofirstchar = firstchar;        zerofirstchar = firstchar;
4617          zerofirstcharflags = firstcharflags;
4618        break;        break;
4619        }        }
4620    
# Line 4054  for (;; ptr++) Line 4649  for (;; ptr++)
4649      means that an initial ] is taken as a data character. At the start of the      means that an initial ] is taken as a data character. At the start of the
4650      loop, c contains the first byte of the character. */      loop, c contains the first byte of the character. */
4651    
4652      if (c != 0) do      if (c != CHAR_NULL) do
4653        {        {
4654        const pcre_uchar *oldptr;        const pcre_uchar *oldptr;
4655    
# Line 4069  for (;; ptr++) Line 4664  for (;; ptr++)
4664        /* In the pre-compile phase, accumulate the length of any extra        /* In the pre-compile phase, accumulate the length of any extra
4665        data and reset the pointer. This is so that very large classes that        data and reset the pointer. This is so that very large classes that
4666        contain a zillion > 255 characters no longer overwrite the work space        contain a zillion > 255 characters no longer overwrite the work space
4667        (which is on the stack). We have to remember that there was XCLASS data,        (which is on the stack). We have to remember that there was XCLASS data,
4668        however. */        however. */
4669    
4670        if (lengthptr != NULL && class_uchardata > class_uchardata_base)        if (lengthptr != NULL && class_uchardata > class_uchardata_base)
# Line 4210  for (;; ptr++) Line 4805  for (;; ptr++)
4805    
4806        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
4807          {          {
4808          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options,
4809              TRUE);
4810          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
4811            if (escape == 0) c = ec;
4812          if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */          else if (escape == ESC_b) c = CHAR_BS; /* \b is backspace in a class */
4813          else if (-c == ESC_N)            /* \N is not supported in a class */          else if (escape == ESC_N)          /* \N is not supported in a class */
4814            {            {
4815            *errorcodeptr = ERR71;            *errorcodeptr = ERR71;
4816            goto FAILED;            goto FAILED;
4817            }            }
4818          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (escape == ESC_Q)            /* Handle start of quoted string */
4819            {            {
4820            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
4821              {              {
# Line 4228  for (;; ptr++) Line 4824  for (;; ptr++)
4824            else inescq = TRUE;            else inescq = TRUE;
4825            continue;            continue;
4826            }            }
4827          else if (-c == ESC_E) continue;  /* Ignore orphan \E */          else if (escape == ESC_E) continue;  /* Ignore orphan \E */
4828    
4829          if (c < 0)          else
4830            {            {
4831            register const pcre_uint8 *cbits = cd->cbits;            register const pcre_uint8 *cbits = cd->cbits;
4832            /* Every class contains at least two < 256 characters. */            /* Every class contains at least two < 256 characters. */
# Line 4238  for (;; ptr++) Line 4834  for (;; ptr++)
4834            /* Every class contains at least two characters. */            /* Every class contains at least two characters. */
4835            class_one_char += 2;            class_one_char += 2;
4836    
4837            switch (-c)            switch (escape)
4838              {              {
4839  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4840              case ESC_du:     /* These are the values given for \d etc */              case ESC_du:     /* These are the values given for \d etc */
# Line 4248  for (;; ptr++) Line 4844  for (;; ptr++)
4844              case ESC_su:     /* of the default ASCII testing. */              case ESC_su:     /* of the default ASCII testing. */
4845              case ESC_SU:              case ESC_SU:
4846              nestptr = ptr;              nestptr = ptr;
4847              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */              ptr = substitutes[escape - ESC_DU] - 1;  /* Just before substitute */
4848              class_has_8bitchar--;                /* Undo! */              class_has_8bitchar--;                /* Undo! */
4849              continue;              continue;
4850  #endif  #endif
# Line 4270  for (;; ptr++) Line 4866  for (;; ptr++)
4866              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
4867              continue;              continue;
4868    
4869              /* Perl 5.004 onwards omits VT from \s, but we must preserve it              /* Perl 5.004 onwards omitted VT from \s, but restored it at Perl
4870              if it was previously set by something earlier in the character              5.18. Before PCRE 8.34, we had to preserve the VT bit if it was
4871              class. Luckily, the value of CHAR_VT is 0x0b in both ASCII and              previously set by something earlier in the character class.
4872              EBCDIC, so we lazily just adjust the appropriate bit. */              Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
4873                we could just adjust the appropriate bit. From PCRE 8.34 we no
4874                longer treat \s and \S specially. */
4875    
4876              case ESC_s:              case ESC_s:
4877              classbits[0] |= cbits[cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
             classbits[1] |= cbits[cbit_space+1] & ~0x08;  
             for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];  
4878              continue;              continue;
4879    
4880              case ESC_S:              case ESC_S:
4881              should_flip_negation = TRUE;              should_flip_negation = TRUE;
4882              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
             classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */  
4883              continue;              continue;
4884    
4885              /* The rest apply in both UCP and non-UCP cases. */              /* The rest apply in both UCP and non-UCP cases. */
4886    
4887              case ESC_h:              case ESC_h:
4888              (void)add_list_to_class(classbits, &class_uchardata, options, cd,              (void)add_list_to_class(classbits, &class_uchardata, options, cd,
4889                PRIV(hspace_list), NOTACHAR);                PRIV(hspace_list), NOTACHAR);
4890              continue;              continue;
4891    
4892              case ESC_H:              case ESC_H:
4893              (void)add_not_list_to_class(classbits, &class_uchardata, options,              (void)add_not_list_to_class(classbits, &class_uchardata, options,
4894                cd, PRIV(hspace_list));                cd, PRIV(hspace_list));
4895              continue;              continue;
4896    
4897              case ESC_v:              case ESC_v:
4898              (void)add_list_to_class(classbits, &class_uchardata, options, cd,              (void)add_list_to_class(classbits, &class_uchardata, options, cd,
4899                PRIV(vspace_list), NOTACHAR);                PRIV(vspace_list), NOTACHAR);
4900              continue;              continue;
4901    
4902              case ESC_V:              case ESC_V:
4903              (void)add_not_list_to_class(classbits, &class_uchardata, options,              (void)add_not_list_to_class(classbits, &class_uchardata, options,
4904                cd, PRIV(vspace_list));                cd, PRIV(vspace_list));
4905              continue;              continue;
4906    
4907  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
# Line 4314  for (;; ptr++) Line 4909  for (;; ptr++)
4909              case ESC_P:              case ESC_P:
4910                {                {
4911                BOOL negated;                BOOL negated;
4912