/[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 903 by ph10, Sat Jan 21 16:37:17 2012 UTC revision 1370 by ph10, Wed Oct 9 10:18:26 2013 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2012 University of Cambridge             Copyright (c) 1997-2013 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When PCRE_DEBUG is defined, we need the pcre(16)_printint() function, which  /* When PCRE_DEBUG is defined, we need the pcre(16|32)_printint() function, which
57  is also used by pcretest. PCRE_DEBUG is not defined when building a production  is also used by pcretest. PCRE_DEBUG is not defined when building a production
58  library. We do not need to select pcre16_printint.c specially, because the  library. We do not need to select pcre16_printint.c specially, because the
59  COMPILE_PCREx macro will already be appropriately set. */  COMPILE_PCREx macro will already be appropriately set. */
# Line 68  COMPILE_PCREx macro will already be appr Line 68  COMPILE_PCREx macro will already be appr
68    
69  /* Macro for setting individual bits in class bitmaps. */  /* Macro for setting individual bits in class bitmaps. */
70    
71  #define SETBIT(a,b) a[b/8] |= (1 << (b%8))  #define SETBIT(a,b) a[(b)/8] |= (1 << ((b)&7))
72    
73  /* Maximum length value to check against when making sure that the integer that  /* Maximum length value to check against when making sure that the integer that
74  holds the compiled pattern length does not overflow. We make it a bit less than  holds the compiled pattern length does not overflow. We make it a bit less than
# Line 77  to check them every time. */ Line 77  to check them every time. */
77    
78  #define OFLOW_MAX (INT_MAX - 20)  #define OFLOW_MAX (INT_MAX - 20)
79    
80    /* Definitions to allow mutual recursion */
81    
82    static int
83      add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84        const pcre_uint32 *, unsigned int);
85    
86    static BOOL
87      compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
88        pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
89        compile_data *, int *);
90    
91    
92    
93  /*************************************************  /*************************************************
94  *      Code parameters and static tables         *  *      Code parameters and static tables         *
# Line 103  kicks in at the same number of forward r Line 115  kicks in at the same number of forward r
115  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
116  #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)  #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
117    
118    /* This value determines the size of the initial vector that is used for
119    remembering named groups during the pre-compile. It is allocated on the stack,
120    but if it is too small, it is expanded using malloc(), in a similar way to the
121    workspace. The value is the number of slots in the list. */
122    
123    #define NAMED_GROUP_LIST_SIZE  20
124    
125  /* The overrun tests check for a slightly smaller size so that they detect the  /* The overrun tests check for a slightly smaller size so that they detect the
126  overrun before it actually does run off the end of the data block. */  overrun before it actually does run off the end of the data block. */
127    
# Line 110  overrun before it actually does run off Line 129  overrun before it actually does run off
129    
130  /* Private flags added to firstchar and reqchar. */  /* Private flags added to firstchar and reqchar. */
131    
132  #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */  #define REQ_CASELESS    (1 << 0)        /* Indicates caselessness */
133  #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */  #define REQ_VARY        (1 << 1)        /* Reqchar followed non-literal item */
134    /* Negative values for the firstchar and reqchar flags */
135    #define REQ_UNSET       (-2)
136    #define REQ_NONE        (-1)
137    
138  /* Repeated character flags. */  /* Repeated character flags. */
139    
# Line 440  static const char error_texts[] = Line 462  static const char error_texts[] =
462    "POSIX collating elements are not supported\0"    "POSIX collating elements are not supported\0"
463    "this version of PCRE is compiled without UTF support\0"    "this version of PCRE is compiled without UTF support\0"
464    "spare error\0"  /** DEAD **/    "spare error\0"  /** DEAD **/
465    "character value in \\x{...} sequence is too large\0"    "character value in \\x{} or \\o{} is too large\0"
466    /* 35 */    /* 35 */
467    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
468    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
# Line 472  static const char error_texts[] = Line 494  static const char error_texts[] =
494    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
495    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
496    /* 60 */    /* 60 */
497    "(*VERB) not recognized\0"    "(*VERB) not recognized or malformed\0"
498    "number is too big\0"    "number is too big\0"
499    "subpattern name expected\0"    "subpattern name expected\0"
500    "digit expected after (?+\0"    "digit expected after (?+\0"
# Line 489  static const char error_texts[] = Line 511  static const char error_texts[] =
511    "too many forward references\0"    "too many forward references\0"
512    "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"    "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
513    "invalid UTF-16 string\0"    "invalid UTF-16 string\0"
514      /* 75 */
515      "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
516      "character value in \\u.... sequence is too large\0"
517      "invalid UTF-32 string\0"
518      "setting UTF is disabled by the application\0"
519      "non-hex character in \\x{} (closing brace missing?)\0"
520      /* 80 */
521      "non-octal character in \\o{} (closing brace missing?)\0"
522      "missing opening brace after \\o\0"
523    ;    ;
524    
525  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 628  static const pcre_uint8 ebcdic_chartab[] Line 659  static const pcre_uint8 ebcdic_chartab[]
659  #endif  #endif
660    
661    
662  /* Definition to allow mutual recursion */  /* This table is used to check whether auto-possessification is possible
663    between adjacent character-type opcodes. The left-hand (repeated) opcode is
664    used to select the row, and the right-hand opcode is use to select the column.
665    A value of 1 means that auto-possessification is OK. For example, the second
666    value in the first row means that \D+\d can be turned into \D++\d.
667    
668    The Unicode property types (\P and \p) have to be present to fill out the table
669    because of what their opcode values are, but the table values should always be
670    zero because property types are handled separately in the code. The last four
671    columns apply to items that cannot be repeated, so there is no need to have
672    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
673    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
674    
675    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
676    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
677    
678    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
679    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
680      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
681      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
682      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
683      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
684      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
685      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
686      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
687      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
688      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
689      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
690      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
691      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
692      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
693      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
694      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
695      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
696      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
697    };
698    
699  static BOOL  
700    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,  /* This table is used to check whether auto-possessification is possible
701      int *, int *, branch_chain *, compile_data *, int *);  between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
702    left-hand (repeated) opcode is used to select the row, and the right-hand
703    opcode is used to select the column. The values are as follows:
704    
705      0   Always return FALSE (never auto-possessify)
706      1   Character groups are distinct (possessify if both are OP_PROP)
707      2   Check character categories in the same group (general or particular)
708      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
709    
710      4   Check left general category vs right particular category
711      5   Check right general category vs left particular category
712    
713      6   Left alphanum vs right general category
714      7   Left space vs right general category
715      8   Left word vs right general category
716    
717      9   Right alphanum vs left general category
718     10   Right space vs left general category
719     11   Right word vs left general category
720    
721     12   Left alphanum vs right particular category
722     13   Left space vs right particular category
723     14   Left word vs right particular category
724    
725     15   Right alphanum vs left particular category
726     16   Right space vs left particular category
727     17   Right word vs left particular category
728    */
729    
730    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
731    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
732      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
733      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
734      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
735      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
736      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
737      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
738      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
739      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
740      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
741      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
742      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
743    };
744    
745    /* This table is used to check whether auto-possessification is possible
746    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
747    specifies a general category and the other specifies a particular category. The
748    row is selected by the general category and the column by the particular
749    category. The value is 1 if the particular category is not part of the general
750    category. */
751    
752    static const pcre_uint8 catposstab[7][30] = {
753    /* Cc Cf Cn Co Cs Ll Lm Lo Lt Lu Mc Me Mn Nd Nl No Pc Pd Pe Pf Pi Po Ps Sc Sk Sm So Zl Zp Zs */
754      { 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* C */
755      { 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* L */
756      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* M */
757      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* N */
758      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1 },  /* P */
759      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1 },  /* S */
760      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0 }   /* Z */
761    };
762    
763    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
764    a general or particular category. The properties in each row are those
765    that apply to the character set in question. Duplication means that a little
766    unnecessary work is done when checking, but this keeps things much simpler
767    because they can all use the same code. For more details see the comment where
768    this table is used.
769    
770    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
771    "space", but from Perl 5.18 it's included, so both categories are treated the
772    same here. */
773    
774    static const pcre_uint8 posspropstab[3][4] = {
775      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
776      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
777      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
778    };
779    
780    
781    
# Line 655  find_error_text(int n) Line 798  find_error_text(int n)
798  const char *s = error_texts;  const char *s = error_texts;
799  for (; n > 0; n--)  for (; n > 0; n--)
800    {    {
801    while (*s++ != 0) {};    while (*s++ != CHAR_NULL) {};
802    if (*s == 0) return "Error text not found (please report)";    if (*s == CHAR_NULL) return "Error text not found (please report)";
803    }    }
804  return s;  return s;
805  }  }
806    
807    
808    
809  /*************************************************  /*************************************************
810  *           Expand the workspace                 *  *           Expand the workspace                 *
811  *************************************************/  *************************************************/
# Line 739  return (*p == CHAR_RIGHT_CURLY_BRACKET); Line 883  return (*p == CHAR_RIGHT_CURLY_BRACKET);
883  *************************************************/  *************************************************/
884    
885  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
886  positive value for a simple escape such as \n, or a negative value which  positive value for a simple escape such as \n, or 0 for a data character which
887  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.
888  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
889  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
890  ptr is pointing at the \. On exit, it is on the final character of the escape  character of the escape sequence.
 sequence.  
891    
892  Arguments:  Arguments:
893    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
894      chptr          points to a returned data character
895    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
896    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
897    options        the options bits    options        the options bits
898    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
899    
900  Returns:         zero or positive => a data character  Returns:         zero => a data character
901                   negative => a special escape sequence                   positive => a special escape sequence
902                     negative => a back reference
903                   on error, errorcodeptr is set                   on error, errorcodeptr is set
904  */  */
905    
906  static int  static int
907  check_escape(const pcre_uchar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
908    int options, BOOL isclass)    int bracount, int options, BOOL isclass)
909  {  {
910  /* PCRE_UTF16 has the same value as PCRE_UTF8. */  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
911  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
912  const pcre_uchar *ptr = *ptrptr + 1;  const pcre_uchar *ptr = *ptrptr + 1;
913  pcre_int32 c;  pcre_uint32 c;
914    int escape = 0;
915  int i;  int i;
916    
917  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
# Line 773  ptr--;                            /* Set Line 919  ptr--;                            /* Set
919    
920  /* If backslash is at the end of the pattern, it's an error. */  /* If backslash is at the end of the pattern, it's an error. */
921    
922  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
923    
924  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
925  in a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
# Line 782  Otherwise further processing may be requ Line 928  Otherwise further processing may be requ
928  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
929  /* Not alphanumeric */  /* Not alphanumeric */
930  else if (c < CHAR_0 || c > CHAR_z) {}  else if (c < CHAR_0 || c > CHAR_z) {}
931  else if ((i = escapes[c - CHAR_0]) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0)
932      { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
933    
934  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
935  /* Not alphanumeric */  /* Not alphanumeric */
936  else if (c < 'a' || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}  else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
937  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; }
938  #endif  #endif
939    
940  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
# Line 795  else if ((i = escapes[c - 0x48]) != 0) Line 942  else if ((i = escapes[c - 0x48]) != 0)
942  else  else
943    {    {
944    const pcre_uchar *oldptr;    const pcre_uchar *oldptr;
945    BOOL braced, negated;    BOOL braced, negated, overflow;
946      int s;
947    
948    switch (c)    switch (c)
949      {      {
# Line 820  else Line 968  else
968          c = 0;          c = 0;
969          for (i = 0; i < 4; ++i)          for (i = 0; i < 4; ++i)
970            {            {
971            register int cc = *(++ptr);            register pcre_uint32 cc = *(++ptr);
972  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
973            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
974            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 829  else Line 977  else
977            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
978  #endif  #endif
979            }            }
980    
981    #if defined COMPILE_PCRE8
982            if (c > (utf ? 0x10ffffU : 0xffU))
983    #elif defined COMPILE_PCRE16
984            if (c > (utf ? 0x10ffffU : 0xffffU))
985    #elif defined COMPILE_PCRE32
986            if (utf && c > 0x10ffffU)
987    #endif
988              {
989              *errorcodeptr = ERR76;
990              }
991            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
992          }          }
993        }        }
994      else      else
# Line 855  else Line 1015  else
1015      (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
1016      number either in angle brackets or in single quotes. However, these are      number either in angle brackets or in single quotes. However, these are
1017      (possibly recursive) subroutine calls, _not_ backreferences. Just return      (possibly recursive) subroutine calls, _not_ backreferences. Just return
1018      the -ESC_g code (cf \k). */      the ESC_g code (cf \k). */
1019    
1020      case CHAR_g:      case CHAR_g:
1021      if (isclass) break;      if (isclass) break;
1022      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1023        {        {
1024        c = -ESC_g;        escape = ESC_g;
1025        break;        break;
1026        }        }
1027    
# Line 870  else Line 1030  else
1030      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1031        {        {
1032        const pcre_uchar *p;        const pcre_uchar *p;
1033        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)        for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
1034          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1035        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)        if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1036          {          {
1037          c = -ESC_k;          escape = ESC_k;
1038          break;          break;
1039          }          }
1040        braced = TRUE;        braced = TRUE;
# Line 890  else Line 1050  else
1050      else negated = FALSE;      else negated = FALSE;
1051    
1052      /* The integer range is limited by the machine's int representation. */      /* The integer range is limited by the machine's int representation. */
1053      c = 0;      s = 0;
1054        overflow = FALSE;
1055      while (IS_DIGIT(ptr[1]))      while (IS_DIGIT(ptr[1]))
1056        {        {
1057        if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */        if (s > INT_MAX / 10 - 1) /* Integer overflow */
1058          {          {
1059          c = -1;          overflow = TRUE;
1060          break;          break;
1061          }          }
1062        c = c * 10 + *(++ptr) - CHAR_0;        s = s * 10 + (int)(*(++ptr) - CHAR_0);
1063        }        }
1064      if (((unsigned int)c) > INT_MAX) /* Integer overflow */      if (overflow) /* Integer overflow */
1065        {        {
1066        while (IS_DIGIT(ptr[1]))        while (IS_DIGIT(ptr[1]))
1067          ptr++;          ptr++;
# Line 914  else Line 1075  else
1075        break;        break;
1076        }        }
1077    
1078      if (c == 0)      if (s == 0)
1079        {        {
1080        *errorcodeptr = ERR58;        *errorcodeptr = ERR58;
1081        break;        break;
# Line 922  else Line 1083  else
1083    
1084      if (negated)      if (negated)
1085        {        {
1086        if (c > bracount)        if (s > bracount)
1087          {          {
1088          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
1089          break;          break;
1090          }          }
1091        c = bracount - (c - 1);        s = bracount - (s - 1);
1092        }        }
1093    
1094      c = -(ESC_REF + c);      escape = -s;
1095      break;      break;
1096    
1097      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1098      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. Perl has changed
1099      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1100        recommended to avoid the ambiguities in the old syntax.
1101    
1102      Outside a character class, the digits are read as a decimal number. If the      Outside a character class, the digits are read as a decimal number. If the
1103      number is less than 10, or if there are that many previous extracting      number is less than 8 (used to be 10), or if there are that many previous
1104      left brackets, then it is a back reference. Otherwise, up to three octal      extracting left brackets, then it is a back reference. Otherwise, up to
1105      digits are read to form an escaped byte. Thus \123 is likely to be octal      three octal digits are read to form an escaped byte. Thus \123 is likely to
1106      123 (cf \0123, which is octal 012 followed by the literal 3). If the octal      be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1107      value is greater than 377, the least significant 8 bits are taken. Inside a      the octal value is greater than 377, the least significant 8 bits are
1108      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1109    
1110        Inside a character class, \ followed by a digit is always either a literal
1111        8 or 9 or an octal number. */
1112    
1113      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1114      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
# Line 952  else Line 1117  else
1117        {        {
1118        oldptr = ptr;        oldptr = ptr;
1119        /* The integer range is limited by the machine's int representation. */        /* The integer range is limited by the machine's int representation. */
1120        c -= CHAR_0;        s = (int)(c -CHAR_0);
1121          overflow = FALSE;
1122        while (IS_DIGIT(ptr[1]))        while (IS_DIGIT(ptr[1]))
1123          {          {
1124          if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */          if (s > INT_MAX / 10 - 1) /* Integer overflow */
1125            {            {
1126            c = -1;            overflow = TRUE;
1127            break;            break;
1128            }            }
1129          c = c * 10 + *(++ptr) - CHAR_0;          s = s * 10 + (int)(*(++ptr) - CHAR_0);
1130          }          }
1131        if (((unsigned int)c) > INT_MAX) /* Integer overflow */        if (overflow) /* Integer overflow */
1132          {          {
1133          while (IS_DIGIT(ptr[1]))          while (IS_DIGIT(ptr[1]))
1134            ptr++;            ptr++;
1135          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1136          break;          break;
1137          }          }
1138        if (c < 10 || c <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1139          {          {
1140          c = -(ESC_REF + c);          escape = -s;
1141          break;          break;
1142          }          }
1143        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1144        }        }
1145    
1146      /* Handle an octal number following \. If the first digit is 8 or 9, Perl      /* Handle a digit following \ when the number is not a back reference. If
1147      generates a binary zero byte and treats the digit as a following literal.      the first digit is 8 or 9, Perl used to generate a binary zero byte and
1148      Thus we have to pull back the pointer by one. */      then treat the digit as a following literal. At least by Perl 5.18 this
1149        changed so as not to insert the binary zero. */
1150      if ((c = *ptr) >= CHAR_8)  
1151        {      if ((c = *ptr) >= CHAR_8) break;
1152        ptr--;  
1153        c = 0;      /* Fall through with a digit less than 8 */
       break;  
       }  
1154    
1155      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
1156      larger first octal digit. The original code used just to take the least      larger first octal digit. The original code used just to take the least
# Line 1002  else Line 1166  else
1166      if (!utf && c > 0xff) *errorcodeptr = ERR51;      if (!utf && c > 0xff) *errorcodeptr = ERR51;
1167  #endif  #endif
1168      break;      break;
1169    
1170        /* \o is a relatively new Perl feature, supporting a more general way of
1171        specifying character codes in octal. The only supported form is \o{ddd}. */
1172    
1173        case CHAR_o:
1174        if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1175          {
1176          ptr += 2;
1177          c = 0;
1178          overflow = FALSE;
1179          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1180            {
1181            register pcre_uint32 cc = *ptr++;
1182            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1183    #ifdef COMPILE_PCRE32
1184            if (c >= 0x10000000l) { overflow = TRUE; break; }
1185    #endif
1186            c = (c << 3) + cc - CHAR_0 ;
1187    #if defined COMPILE_PCRE8
1188            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1189    #elif defined COMPILE_PCRE16
1190            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1191    #elif defined COMPILE_PCRE32
1192            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1193    #endif
1194            }
1195          if (overflow)
1196            {
1197            while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1198            *errorcodeptr = ERR34;
1199            }
1200          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1201            {
1202            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1203            }
1204          else *errorcodeptr = ERR80;
1205          }
1206        break;
1207    
1208      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1209      than 0xff in utf or non-8bit mode, but only if the ddd are hex digits.      numbers. Otherwise it is a lowercase x letter. */
     If not, { is treated as a data character. */  
1210    
1211      case CHAR_x:      case CHAR_x:
1212      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1213        {        {
       /* In JavaScript, \x must be followed by two hexadecimal numbers.  
       Otherwise it is a lowercase x letter. */  
1214        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1215          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1216          {          {
1217          c = 0;          c = 0;
1218          for (i = 0; i < 2; ++i)          for (i = 0; i < 2; ++i)
1219            {            {
1220            register int cc = *(++ptr);            register pcre_uint32 cc = *(++ptr);
1221  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1222            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1223            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 1028  else Line 1227  else
1227  #endif  #endif
1228            }            }
1229          }          }
1230        break;        }    /* End JavaScript handling */
1231        }  
1232        /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1233      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1234        {      digits. If not, { used to be treated as a data character. However, Perl
1235        const pcre_uchar *pt = ptr + 2;      seems to read hex digits up to the first non-such, and ignore the rest, so
1236        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1237        c = 0;      now gives an error. */
1238        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)  
1239        else
1240          {
1241          if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1242          {          {
1243          register int cc = *pt++;          ptr += 2;
1244          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          c = 0;
1245            overflow = FALSE;
1246            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1247              {
1248              register pcre_uint32 cc = *ptr++;
1249              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1250    
1251    #ifdef COMPILE_PCRE32
1252              if (c >= 0x10000000l) { overflow = TRUE; break; }
1253    #endif
1254    
1255  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1256          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1257          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1258  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1259          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */            if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1260          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
 #endif  
   
 #ifdef COMPILE_PCRE8  
         if (c > (utf ? 0x10ffff : 0xff)) { c = -1; break; }  
 #else  
 #ifdef COMPILE_PCRE16  
         if (c > (utf ? 0x10ffff : 0xffff)) { c = -1; break; }  
 #endif  
1261  #endif  #endif
1262          }  
1263    #if defined COMPILE_PCRE8
1264        if (c < 0)            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1265          {  #elif defined COMPILE_PCRE16
1266          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1267          *errorcodeptr = ERR34;  #elif defined COMPILE_PCRE32
1268          }            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1269    #endif
1270        if (*pt == CHAR_RIGHT_CURLY_BRACKET)            }
1271          {  
1272          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;          if (overflow)
1273          ptr = pt;            {
1274          break;            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1275          }            *errorcodeptr = ERR34;
1276              }
1277        /* If the sequence of hex digits does not end with '}', then we don't  
1278        recognize this construct; fall through to the normal \x handling. */          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1279        }            {
1280              if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1281              }
1282    
1283            /* If the sequence of hex digits does not end with '}', give an error.
1284            We used just to recognize this construct and fall through to the normal
1285            \x handling, but nowadays Perl gives an error, which seems much more
1286            sensible, so we do too. */
1287    
1288            else *errorcodeptr = ERR79;
1289            }   /* End of \x{} processing */
1290    
1291      /* Read just a single-byte hex-defined char */        /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1292    
1293      c = 0;        else
1294      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)          {
1295        {          c = 0;
1296        int cc;                                  /* Some compilers don't like */          while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1297        cc = *(++ptr);                           /* ++ in initializers */            {
1298              pcre_uint32 cc;                          /* Some compilers don't like */
1299              cc = *(++ptr);                           /* ++ in initializers */
1300  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1301        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1302        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1303  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1304        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */            if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1305        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1306  #endif  #endif
1307        }            }
1308            }     /* End of \xdd handling */
1309          }       /* End of Perl-style \x handling */
1310      break;      break;
1311    
1312      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
# Line 1099  else Line 1316  else
1316    
1317      case CHAR_c:      case CHAR_c:
1318      c = *(++ptr);      c = *(++ptr);
1319      if (c == 0)      if (c == CHAR_NULL)
1320        {        {
1321        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1322        break;        break;
# Line 1139  else Line 1356  else
1356  newline". PCRE does not support \N{name}. However, it does support  newline". PCRE does not support \N{name}. However, it does support
1357  quantification such as \N{2,3}. */  quantification such as \N{2,3}. */
1358    
1359  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&  if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1360       !is_counted_repeat(ptr+2))       !is_counted_repeat(ptr+2))
1361    *errorcodeptr = ERR37;    *errorcodeptr = ERR37;
1362    
1363  /* If PCRE_UCP is set, we change the values for \d etc. */  /* If PCRE_UCP is set, we change the values for \d etc. */
1364    
1365  if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)  if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1366    c -= (ESC_DU - ESC_D);    escape += (ESC_DU - ESC_D);
1367    
1368  /* Set the pointer to the final character before returning. */  /* Set the pointer to the final character before returning. */
1369    
1370  *ptrptr = ptr;  *ptrptr = ptr;
1371  return c;  *chptr = c;
1372    return escape;
1373  }  }
1374    
1375    
# Line 1169  escape sequence. Line 1387  escape sequence.
1387  Argument:  Argument:
1388    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1389    negptr         points to a boolean that is set TRUE for negation else FALSE    negptr         points to a boolean that is set TRUE for negation else FALSE
1390    dptr           points to an int that is set to the detailed property value    ptypeptr       points to an unsigned int that is set to the type value
1391      pdataptr       points to an unsigned int that is set to the detailed property value
1392    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1393    
1394  Returns:         type value from ucp_type_table, or -1 for an invalid type  Returns:         TRUE if the type value was found, or FALSE for an invalid type
1395  */  */
1396    
1397  static int  static BOOL
1398  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1399      unsigned int *pdataptr, int *errorcodeptr)
1400  {  {
1401  int c, i, bot, top;  pcre_uchar c;
1402    int i, bot, top;
1403  const pcre_uchar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
1404  pcre_uchar name[32];  pcre_uchar name[32];
1405    
1406  c = *(++ptr);  c = *(++ptr);
1407  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1408    
1409  *negptr = FALSE;  *negptr = FALSE;
1410    
# Line 1200  if (c == CHAR_LEFT_CURLY_BRACKET) Line 1421  if (c == CHAR_LEFT_CURLY_BRACKET)
1421    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1422      {      {
1423      c = *(++ptr);      c = *(++ptr);
1424      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1425      if (c == CHAR_RIGHT_CURLY_BRACKET) break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1426      name[i] = c;      name[i] = c;
1427      }      }
# Line 1225  top = PRIV(utt_size); Line 1446  top = PRIV(utt_size);
1446    
1447  while (bot < top)  while (bot < top)
1448    {    {
1449      int r;
1450    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1451    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);
1452    if (c == 0)    if (r == 0)
1453      {      {
1454      *dptr = PRIV(utt)[i].value;      *ptypeptr = PRIV(utt)[i].type;
1455      return PRIV(utt)[i].type;      *pdataptr = PRIV(utt)[i].value;
1456        return TRUE;
1457      }      }
1458    if (c > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1459    }    }
1460    
1461  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1462  *ptrptr = ptr;  *ptrptr = ptr;
1463  return -1;  return FALSE;
1464    
1465  ERROR_RETURN:  ERROR_RETURN:
1466  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1467  *ptrptr = ptr;  *ptrptr = ptr;
1468  return -1;  return FALSE;
1469  }  }
1470  #endif  #endif
1471    
1472    
1473    
   
1474  /*************************************************  /*************************************************
1475  *         Read repeat counts                     *  *         Read repeat counts                     *
1476  *************************************************/  *************************************************/
# Line 1277  int max = -1; Line 1499  int max = -1;
1499  /* 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
1500  an integer overflow. */  an integer overflow. */
1501    
1502  while (IS_DIGIT(*p)) min = min * 10 + *p++ - CHAR_0;  while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);
1503  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1504    {    {
1505    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 1292  if (*p == CHAR_RIGHT_CURLY_BRACKET) max Line 1514  if (*p == CHAR_RIGHT_CURLY_BRACKET) max
1514    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1515      {      {
1516      max = 0;      max = 0;
1517      while(IS_DIGIT(*p)) max = max * 10 + *p++ - CHAR_0;      while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);
1518      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1519        {        {
1520        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 1317  return p; Line 1539  return p;
1539    
1540    
1541  /*************************************************  /*************************************************
 *  Subroutine for finding forward reference      *  
 *************************************************/  
   
 /* This recursive function is called only from find_parens() below. The  
 top-level call starts at the beginning of the pattern. All other calls must  
 start at a parenthesis. It scans along a pattern's text looking for capturing  
 subpatterns, and counting them. If it finds a named pattern that matches the  
 name it is given, it returns its number. Alternatively, if the name is NULL, it  
 returns when it reaches a given numbered subpattern. Recursion is used to keep  
 track of subpatterns that reset the capturing group numbers - the (?| feature.  
   
 This function was originally called only from the second pass, in which we know  
 that if (?< or (?' or (?P< is encountered, the name will be correctly  
 terminated because that is checked in the first pass. There is now one call to  
 this function in the first pass, to check for a recursive back reference by  
 name (so that we can make the whole group atomic). In this case, we need check  
 only up to the current position in the pattern, and that is still OK because  
 and previous occurrences will have been checked. To make this work, the test  
 for "end of pattern" is a check against cd->end_pattern in the main loop,  
 instead of looking for a binary zero. This means that the special first-pass  
 call can adjust cd->end_pattern temporarily. (Checks for binary zero while  
 processing items within the loop are OK, because afterwards the main loop will  
 terminate.)  
   
 Arguments:  
   ptrptr       address of the current character pointer (updated)  
   cd           compile background data  
   name         name to seek, or NULL if seeking a numbered subpattern  
   lorn         name length, or subpattern number if name is NULL  
   xmode        TRUE if we are in /x mode  
   utf          TRUE if we are in UTF-8 / UTF-16 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;  
 }  
   
   
   
   
 /*************************************************  
1542  *      Find first significant op code            *  *      Find first significant op code            *
1543  *************************************************/  *************************************************/
1544    
# Line 1650  for (;;) Line 1577  for (;;)
1577    
1578      case OP_CALLOUT:      case OP_CALLOUT:
1579      case OP_CREF:      case OP_CREF:
1580      case OP_NCREF:      case OP_DNCREF:
1581      case OP_RREF:      case OP_RREF:
1582      case OP_NRREF:      case OP_DNRREF:
1583      case OP_DEF:      case OP_DEF:
1584      code += PRIV(OP_lengths)[*code];      code += PRIV(OP_lengths)[*code];
1585      break;      break;
# Line 1666  for (;;) Line 1593  for (;;)
1593    
1594    
1595    
   
1596  /*************************************************  /*************************************************
1597  *        Find the fixed length of a branch       *  *        Find the fixed length of a branch       *
1598  *************************************************/  *************************************************/
# Line 1684  and doing the check at the end; a flag s Line 1610  and doing the check at the end; a flag s
1610    
1611  Arguments:  Arguments:
1612    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1613    utf      TRUE in UTF-8 / UTF-16 mode    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1614    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1615    cd       the "compile data" structure    cd       the "compile data" structure
1616    
# Line 1710  for (;;) Line 1636  for (;;)
1636    {    {
1637    int d;    int d;
1638    pcre_uchar *ce, *cs;    pcre_uchar *ce, *cs;
1639    register int op = *cc;    register pcre_uchar op = *cc;
1640    
1641    switch (op)    switch (op)
1642      {      {
# Line 1790  for (;;) Line 1716  for (;;)
1716      case OP_COMMIT:      case OP_COMMIT:
1717      case OP_CREF:      case OP_CREF:
1718      case OP_DEF:      case OP_DEF:
1719        case OP_DNCREF:
1720        case OP_DNRREF:
1721      case OP_DOLL:      case OP_DOLL:
1722      case OP_DOLLM:      case OP_DOLLM:
1723      case OP_EOD:      case OP_EOD:
1724      case OP_EODN:      case OP_EODN:
1725      case OP_FAIL:      case OP_FAIL:
     case OP_NCREF:  
     case OP_NRREF:  
1726      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1727      case OP_PRUNE:      case OP_PRUNE:
1728      case OP_REVERSE:      case OP_REVERSE:
# Line 1830  for (;;) Line 1756  for (;;)
1756      case OP_EXACTI:      case OP_EXACTI:
1757      case OP_NOTEXACT:      case OP_NOTEXACT:
1758      case OP_NOTEXACTI:      case OP_NOTEXACTI:
1759      branchlength += GET2(cc,1);      branchlength += (int)GET2(cc,1);
1760      cc += 2 + IMM2_SIZE;      cc += 2 + IMM2_SIZE;
1761  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
1762      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
# Line 1839  for (;;) Line 1765  for (;;)
1765    
1766      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1767      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1768      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP) cc += 2;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1769          cc += 2;
1770      cc += 1 + IMM2_SIZE + 1;      cc += 1 + IMM2_SIZE + 1;
1771      break;      break;
1772    
# Line 1874  for (;;) Line 1801  for (;;)
1801    
1802      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1803    
 #if defined SUPPORT_UTF || defined COMPILE_PCRE16  
     case OP_XCLASS:  
     cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];  
     /* Fall through */  
 #endif  
   
1804      case OP_CLASS:      case OP_CLASS:
1805      case OP_NCLASS:      case OP_NCLASS:
1806    #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1807        case OP_XCLASS:
1808        /* The original code caused an unsigned overflow in 64 bit systems,
1809        so now we use a conditional statement. */
1810        if (op == OP_XCLASS)
1811          cc += GET(cc, 1);
1812        else
1813          cc += PRIV(OP_lengths)[OP_CLASS];
1814    #else
1815      cc += PRIV(OP_lengths)[OP_CLASS];      cc += PRIV(OP_lengths)[OP_CLASS];
1816    #endif
1817    
1818      switch (*cc)      switch (*cc)
1819        {        {
# Line 1897  for (;;) Line 1828  for (;;)
1828        case OP_CRRANGE:        case OP_CRRANGE:
1829        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1830        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1831        branchlength += GET2(cc,1);        branchlength += (int)GET2(cc,1);
1832        cc += 1 + 2 * IMM2_SIZE;        cc += 1 + 2 * IMM2_SIZE;
1833        break;        break;
1834    
# Line 1964  for (;;) Line 1895  for (;;)
1895      case OP_QUERYI:      case OP_QUERYI:
1896      case OP_REF:      case OP_REF:
1897      case OP_REFI:      case OP_REFI:
1898        case OP_DNREF:
1899        case OP_DNREFI:
1900      case OP_SBRA:      case OP_SBRA:
1901      case OP_SBRAPOS:      case OP_SBRAPOS:
1902      case OP_SCBRA:      case OP_SCBRA:
# Line 2000  for (;;) Line 1933  for (;;)
1933    
1934    
1935    
   
1936  /*************************************************  /*************************************************
1937  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
1938  *************************************************/  *************************************************/
# Line 2013  length. Line 1945  length.
1945    
1946  Arguments:  Arguments:
1947    code        points to start of expression    code        points to start of expression
1948    utf         TRUE in UTF-8 / UTF-16 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
1949    number      the required bracket number or negative to find a lookbehind    number      the required bracket number or negative to find a lookbehind
1950    
1951  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 2024  PRIV(find_bracket)(const pcre_uchar *cod Line 1956  PRIV(find_bracket)(const pcre_uchar *cod
1956  {  {
1957  for (;;)  for (;;)
1958    {    {
1959    register int c = *code;    register pcre_uchar c = *code;
1960    
1961    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1962    
# Line 2047  for (;;) Line 1979  for (;;)
1979    else if (c == OP_CBRA || c == OP_SCBRA ||    else if (c == OP_CBRA || c == OP_SCBRA ||
1980             c == OP_CBRAPOS || c == OP_SCBRAPOS)             c == OP_CBRAPOS || c == OP_SCBRAPOS)
1981      {      {
1982      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
1983      if (n == number) return (pcre_uchar *)code;      if (n == number) return (pcre_uchar *)code;
1984      code += PRIV(OP_lengths)[c];      code += PRIV(OP_lengths)[c];
1985      }      }
# Line 2077  for (;;) Line 2009  for (;;)
2009        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2010        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2011        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2012        if (code[1 + IMM2_SIZE] == OP_PROP        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2013          || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;          code += 2;
2014        break;        break;
2015    
2016        case OP_MARK:        case OP_MARK:
2017        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2018        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2019        case OP_THEN_ARG:        case OP_THEN_ARG:
2020        code += code[1];        code += code[1];
2021        break;        break;
# Line 2100  for (;;) Line 2029  for (;;)
2029    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
2030    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2031    
2032  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2033      if (utf) switch(c)      if (utf) switch(c)
2034        {        {
2035        case OP_CHAR:        case OP_CHAR:
# Line 2152  instance of OP_RECURSE. Line 2081  instance of OP_RECURSE.
2081    
2082  Arguments:  Arguments:
2083    code        points to start of expression    code        points to start of expression
2084    utf         TRUE in UTF-8 / UTF-16 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2085    
2086  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
2087  */  */
# Line 2162  find_recurse(const pcre_uchar *code, BOO Line 2091  find_recurse(const pcre_uchar *code, BOO
2091  {  {
2092  for (;;)  for (;;)
2093    {    {
2094    register int c = *code;    register pcre_uchar c = *code;
2095    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2096    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2097    
# Line 2197  for (;;) Line 2126  for (;;)
2126        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2127        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2128        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2129        if (code[1 + IMM2_SIZE] == OP_PROP        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2130          || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;          code += 2;
2131        break;        break;
2132    
2133        case OP_MARK:        case OP_MARK:
2134        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2135        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2136        case OP_THEN_ARG:        case OP_THEN_ARG:
2137        code += code[1];        code += code[1];
2138        break;        break;
# Line 2220  for (;;) Line 2146  for (;;)
2146      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
2147      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2148    
2149  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2150      if (utf) switch(c)      if (utf) switch(c)
2151        {        {
2152        case OP_CHAR:        case OP_CHAR:
2153        case OP_CHARI:        case OP_CHARI:
2154          case OP_NOT:
2155          case OP_NOTI:
2156        case OP_EXACT:        case OP_EXACT:
2157        case OP_EXACTI:        case OP_EXACTI:
2158          case OP_NOTEXACT:
2159          case OP_NOTEXACTI:
2160        case OP_UPTO:        case OP_UPTO:
2161        case OP_UPTOI:        case OP_UPTOI:
2162          case OP_NOTUPTO:
2163          case OP_NOTUPTOI:
2164        case OP_MINUPTO:        case OP_MINUPTO:
2165        case OP_MINUPTOI:        case OP_MINUPTOI:
2166          case OP_NOTMINUPTO:
2167          case OP_NOTMINUPTOI:
2168        case OP_POSUPTO:        case OP_POSUPTO:
2169        case OP_POSUPTOI:        case OP_POSUPTOI:
2170          case OP_NOTPOSUPTO:
2171          case OP_NOTPOSUPTOI:
2172        case OP_STAR:        case OP_STAR:
2173        case OP_STARI:        case OP_STARI:
2174          case OP_NOTSTAR:
2175          case OP_NOTSTARI:
2176        case OP_MINSTAR:        case OP_MINSTAR:
2177        case OP_MINSTARI:        case OP_MINSTARI:
2178          case OP_NOTMINSTAR:
2179          case OP_NOTMINSTARI:
2180        case OP_POSSTAR:        case OP_POSSTAR:
2181        case OP_POSSTARI:        case OP_POSSTARI:
2182          case OP_NOTPOSSTAR:
2183          case OP_NOTPOSSTARI:
2184        case OP_PLUS:        case OP_PLUS:
2185        case OP_PLUSI:        case OP_PLUSI:
2186          case OP_NOTPLUS:
2187          case OP_NOTPLUSI:
2188        case OP_MINPLUS:        case OP_MINPLUS:
2189        case OP_MINPLUSI:        case OP_MINPLUSI:
2190          case OP_NOTMINPLUS:
2191          case OP_NOTMINPLUSI:
2192        case OP_POSPLUS:        case OP_POSPLUS:
2193        case OP_POSPLUSI:        case OP_POSPLUSI:
2194          case OP_NOTPOSPLUS:
2195          case OP_NOTPOSPLUSI:
2196        case OP_QUERY:        case OP_QUERY:
2197        case OP_QUERYI:        case OP_QUERYI:
2198          case OP_NOTQUERY:
2199          case OP_NOTQUERYI:
2200        case OP_MINQUERY:        case OP_MINQUERY:
2201        case OP_MINQUERYI:        case OP_MINQUERYI:
2202          case OP_NOTMINQUERY:
2203          case OP_NOTMINQUERYI:
2204        case OP_POSQUERY:        case OP_POSQUERY:
2205        case OP_POSQUERYI:        case OP_POSQUERYI:
2206          case OP_NOTPOSQUERY:
2207          case OP_NOTPOSQUERYI:
2208        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2209        break;        break;
2210        }        }
# Line 2278  bracket whose current branch will alread Line 2232  bracket whose current branch will alread
2232  Arguments:  Arguments:
2233    code        points to start of search    code        points to start of search
2234    endcode     points to where to stop    endcode     points to where to stop
2235    utf         TRUE if in UTF-8 / UTF-16 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2236    cd          contains pointers to tables etc.    cd          contains pointers to tables etc.
2237      recurses    chain of recurse_check to catch mutual recursion
2238    
2239  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2240  */  */
2241    
2242    typedef struct recurse_check {
2243      struct recurse_check *prev;
2244      const pcre_uchar *group;
2245    } recurse_check;
2246    
2247  static BOOL  static BOOL
2248  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2249    BOOL utf, compile_data *cd)    BOOL utf, compile_data *cd, recurse_check *recurses)
2250  {  {
2251  register int c;  register pcre_uchar c;
2252    recurse_check this_recurse;
2253    
2254  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2255       code < endcode;       code < endcode;
2256       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
# Line 2316  for (code = first_significant_code(code Line 2278  for (code = first_significant_code(code
2278    
2279    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2280      {      {
2281      const pcre_uchar *scode;      const pcre_uchar *scode = cd->start_code + GET(code, 1);
2282      BOOL empty_branch;      BOOL empty_branch;
2283    
2284      /* Test for forward reference */      /* Test for forward reference or uncompleted reference. This is disabled
2285        when called to scan a completed pattern by setting cd->start_workspace to
2286        NULL. */
2287    
2288        if (cd->start_workspace != NULL)
2289          {
2290          const pcre_uchar *tcode;
2291          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2292            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2293          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2294          }
2295    
2296        /* If we are scanning a completed pattern, there are no forward references
2297        and all groups are complete. We need to detect whether this is a recursive
2298        call, as otherwise there will be an infinite loop. If it is a recursion,
2299        just skip over it. Simple recursions are easily detected. For mutual
2300        recursions we keep a chain on the stack. */
2301    
2302        else
2303          {
2304          recurse_check *r = recurses;
2305          const pcre_uchar *endgroup = scode;
2306    
2307          do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2308          if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2309    
2310      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)        for (r = recurses; r != NULL; r = r->prev)
2311        if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;          if (r->group == scode) break;
2312          if (r != NULL) continue;   /* Mutual recursion */
2313          }
2314    
2315      /* Not a forward reference, test for completed backward reference */      /* Completed reference; scan the referenced group, remembering it on the
2316        stack chain to detect mutual recursions. */
2317    
2318      empty_branch = FALSE;      empty_branch = FALSE;
2319      scode = cd->start_code + GET(code, 1);      this_recurse.prev = recurses;
2320      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      this_recurse.group = scode;
   
     /* Completed backwards reference */  
2321    
2322      do      do
2323        {        {
2324        if (could_be_empty_branch(scode, endcode, utf, cd))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2325          {          {
2326          empty_branch = TRUE;          empty_branch = TRUE;
2327          break;          break;
# Line 2390  for (code = first_significant_code(code Line 2377  for (code = first_significant_code(code
2377        empty_branch = FALSE;        empty_branch = FALSE;
2378        do        do
2379          {          {
2380          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2381            empty_branch = TRUE;            empty_branch = TRUE;
2382          code += GET(code, 1);          code += GET(code, 1);
2383          }          }
# Line 2448  for (code = first_significant_code(code Line 2435  for (code = first_significant_code(code
2435    
2436      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2437    
2438        case OP_ANY:
2439        case OP_ALLANY:
2440        case OP_ANYBYTE:
2441    
2442      case OP_PROP:      case OP_PROP:
2443      case OP_NOTPROP:      case OP_NOTPROP:
2444        case OP_ANYNL:
2445    
2446        case OP_NOT_HSPACE:
2447        case OP_HSPACE:
2448        case OP_NOT_VSPACE:
2449        case OP_VSPACE:
2450      case OP_EXTUNI:      case OP_EXTUNI:
2451    
2452      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2453      case OP_DIGIT:      case OP_DIGIT:
2454      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2455      case OP_WHITESPACE:      case OP_WHITESPACE:
2456      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2457      case OP_WORDCHAR:      case OP_WORDCHAR:
2458      case OP_ANY:  
     case OP_ALLANY:  
     case OP_ANYBYTE:  
2459      case OP_CHAR:      case OP_CHAR:
2460      case OP_CHARI:      case OP_CHARI:
2461      case OP_NOT:      case OP_NOT:
2462      case OP_NOTI:      case OP_NOTI:
2463    
2464      case OP_PLUS:      case OP_PLUS:
2465        case OP_PLUSI:
2466      case OP_MINPLUS:      case OP_MINPLUS:
2467      case OP_POSPLUS:      case OP_MINPLUSI:
2468      case OP_EXACT:  
2469      case OP_NOTPLUS:      case OP_NOTPLUS:
2470        case OP_NOTPLUSI:
2471      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2472        case OP_NOTMINPLUSI:
2473    
2474        case OP_POSPLUS:
2475        case OP_POSPLUSI:
2476      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2477        case OP_NOTPOSPLUSI:
2478    
2479        case OP_EXACT:
2480        case OP_EXACTI:
2481      case OP_NOTEXACT:      case OP_NOTEXACT:
2482        case OP_NOTEXACTI:
2483    
2484      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2485      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2486      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2487      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2488    
2489      return FALSE;      return FALSE;
2490    
2491      /* 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 2495  for (code = first_significant_code(code Line 2505  for (code = first_significant_code(code
2505      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2506      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2507      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2508      if (code[1 + IMM2_SIZE] == OP_PROP      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2509        || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;        code += 2;
2510      break;      break;
2511    
2512      /* End of branch */      /* End of branch */
# Line 2509  for (code = first_significant_code(code Line 2519  for (code = first_significant_code(code
2519      return TRUE;      return TRUE;
2520    
2521      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2522      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2523        followed by a multibyte character. */
2524    
2525  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2526      case OP_STAR:      case OP_STAR:
2527      case OP_STARI:      case OP_STARI:
2528        case OP_NOTSTAR:
2529        case OP_NOTSTARI:
2530    
2531      case OP_MINSTAR:      case OP_MINSTAR:
2532      case OP_MINSTARI:      case OP_MINSTARI:
2533        case OP_NOTMINSTAR:
2534        case OP_NOTMINSTARI:
2535    
2536      case OP_POSSTAR:      case OP_POSSTAR:
2537      case OP_POSSTARI:      case OP_POSSTARI:
2538        case OP_NOTPOSSTAR:
2539        case OP_NOTPOSSTARI:
2540    
2541      case OP_QUERY:      case OP_QUERY:
2542      case OP_QUERYI:      case OP_QUERYI:
2543        case OP_NOTQUERY:
2544        case OP_NOTQUERYI:
2545    
2546      case OP_MINQUERY:      case OP_MINQUERY:
2547      case OP_MINQUERYI:      case OP_MINQUERYI:
2548        case OP_NOTMINQUERY:
2549        case OP_NOTMINQUERYI:
2550    
2551      case OP_POSQUERY:      case OP_POSQUERY:
2552      case OP_POSQUERYI:      case OP_POSQUERYI:
2553        case OP_NOTPOSQUERY:
2554        case OP_NOTPOSQUERYI:
2555    
2556      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2557      break;      break;
2558    
2559      case OP_UPTO:      case OP_UPTO:
2560      case OP_UPTOI:      case OP_UPTOI:
2561      case OP_MINUPTO:      case OP_NOTUPTO:
2562        case OP_NOTUPTOI:
2563    
2564        case OP_MINUPTO:
2565        case OP_MINUPTOI:
2566        case OP_NOTMINUPTO:
2567        case OP_NOTMINUPTOI:
2568    
2569        case OP_POSUPTO:
2570        case OP_POSUPTOI:
2571        case OP_NOTPOSUPTO:
2572        case OP_NOTPOSUPTOI:
2573    
2574        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2575        break;
2576    #endif
2577    
2578        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2579        string. */
2580    
2581        case OP_MARK:
2582        case OP_PRUNE_ARG:
2583        case OP_SKIP_ARG:
2584        case OP_THEN_ARG:
2585        code += code[1];
2586        break;
2587    
2588        /* None of the remaining opcodes are required to match a character. */
2589    
2590        default:
2591        break;
2592        }
2593      }
2594    
2595    return TRUE;
2596    }
2597    
2598    
2599    
2600    /*************************************************
2601    *    Scan compiled regex for non-emptiness       *
2602    *************************************************/
2603    
2604    /* This function is called to check for left recursive calls. We want to check
2605    the current branch of the current pattern to see if it could match the empty
2606    string. If it could, we must look outwards for branches at other levels,
2607    stopping when we pass beyond the bracket which is the subject of the recursion.
2608    This function is called only during the real compile, not during the
2609    pre-compile.
2610    
2611    Arguments:
2612      code        points to start of the recursion
2613      endcode     points to where to stop (current RECURSE item)
2614      bcptr       points to the chain of current (unclosed) branch starts
2615      utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2616      cd          pointers to tables etc
2617    
2618    Returns:      TRUE if what is matched could be empty
2619    */
2620    
2621    static BOOL
2622    could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2623      branch_chain *bcptr, BOOL utf, compile_data *cd)
2624    {
2625    while (bcptr != NULL && bcptr->current_branch >= code)
2626      {
2627      if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2628        return FALSE;
2629      bcptr = bcptr->outer;
2630      }
2631    return TRUE;
2632    }
2633    
2634    
2635    
2636    /*************************************************
2637    *        Base opcode of repeated opcodes         *
2638    *************************************************/
2639    
2640    /* Returns the base opcode for repeated single character type opcodes. If the
2641    opcode is not a repeated character type, it returns with the original value.
2642    
2643    Arguments:  c opcode
2644    Returns:    base opcode for the type
2645    */
2646    
2647    static pcre_uchar
2648    get_repeat_base(pcre_uchar c)
2649    {
2650    return (c > OP_TYPEPOSUPTO)? c :
2651           (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2652           (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2653           (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2654           (c >= OP_STARI)?      OP_STARI :
2655                                 OP_STAR;
2656    }
2657    
2658    
2659    
2660    #ifdef SUPPORT_UCP
2661    /*************************************************
2662    *        Check a character and a property        *
2663    *************************************************/
2664    
2665    /* This function is called by check_auto_possessive() when a property item
2666    is adjacent to a fixed character.
2667    
2668    Arguments:
2669      c            the character
2670      ptype        the property type
2671      pdata        the data for the type
2672      negated      TRUE if it's a negated property (\P or \p{^)
2673    
2674    Returns:       TRUE if auto-possessifying is OK
2675    */
2676    
2677    static BOOL
2678    check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2679      BOOL negated)
2680    {
2681    const pcre_uint32 *p;
2682    const ucd_record *prop = GET_UCD(c);
2683    
2684    switch(ptype)
2685      {
2686      case PT_LAMP:
2687      return (prop->chartype == ucp_Lu ||
2688              prop->chartype == ucp_Ll ||
2689              prop->chartype == ucp_Lt) == negated;
2690    
2691      case PT_GC:
2692      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2693    
2694      case PT_PC:
2695      return (pdata == prop->chartype) == negated;
2696    
2697      case PT_SC:
2698      return (pdata == prop->script) == negated;
2699    
2700      /* These are specials */
2701    
2702      case PT_ALNUM:
2703      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2704              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2705    
2706      /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2707      means that Perl space and POSIX space are now identical. PCRE was changed
2708      at release 8.34. */
2709    
2710      case PT_SPACE:    /* Perl space */
2711      case PT_PXSPACE:  /* POSIX space */
2712      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2713              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2714              c == CHAR_FF || c == CHAR_CR)
2715              == negated;
2716    
2717      case PT_WORD:
2718      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2719              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2720              c == CHAR_UNDERSCORE) == negated;
2721    
2722      case PT_CLIST:
2723      p = PRIV(ucd_caseless_sets) + prop->caseset;
2724      for (;;)
2725        {
2726        if (c < *p) return !negated;
2727        if (c == *p++) return negated;
2728        }
2729      break;  /* Control never reaches here */
2730      }
2731    
2732    return FALSE;
2733    }
2734    #endif  /* SUPPORT_UCP */
2735    
2736    
2737    
2738    /*************************************************
2739    *        Fill the character property list        *
2740    *************************************************/
2741    
2742    /* Checks whether the code points to an opcode that can take part in auto-
2743    possessification, and if so, fills a list with its properties.
2744    
2745    Arguments:
2746      code        points to start of expression
2747      utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2748      fcc         points to case-flipping table
2749      list        points to output list
2750                  list[0] will be filled with the opcode
2751                  list[1] will be non-zero if this opcode
2752                    can match an empty character string
2753                  list[2..7] depends on the opcode
2754    
2755    Returns:      points to the start of the next opcode if *code is accepted
2756                  NULL if *code is not accepted
2757    */
2758    
2759    static const pcre_uchar *
2760    get_chr_property_list(const pcre_uchar *code, BOOL utf,
2761      const pcre_uint8 *fcc, pcre_uint32 *list)
2762    {
2763    pcre_uchar c = *code;
2764    const pcre_uchar *end;
2765    const pcre_uint32 *clist_src;
2766    pcre_uint32 *clist_dest;
2767    pcre_uint32 chr;
2768    pcre_uchar base;
2769    
2770    list[0] = c;
2771    list[1] = FALSE;
2772    code++;
2773    
2774    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2775      {
2776      base = get_repeat_base(c);
2777      c -= (base - OP_STAR);
2778    
2779      if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2780        code += IMM2_SIZE;
2781    
2782      list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2783    
2784      switch(base)
2785        {
2786        case OP_STAR:
2787        list[0] = OP_CHAR;
2788        break;
2789    
2790        case OP_STARI:
2791        list[0] = OP_CHARI;
2792        break;
2793    
2794        case OP_NOTSTAR:
2795        list[0] = OP_NOT;
2796        break;
2797    
2798        case OP_NOTSTARI:
2799        list[0] = OP_NOTI;
2800        break;
2801    
2802        case OP_TYPESTAR:
2803        list[0] = *code;
2804        code++;
2805        break;
2806        }
2807      c = list[0];
2808      }
2809    
2810    switch(c)
2811      {
2812      case OP_NOT_DIGIT:
2813      case OP_DIGIT:
2814      case OP_NOT_WHITESPACE:
2815      case OP_WHITESPACE:
2816      case OP_NOT_WORDCHAR:
2817      case OP_WORDCHAR:
2818      case OP_ANY:
2819      case OP_ALLANY:
2820      case OP_ANYNL:
2821      case OP_NOT_HSPACE:
2822      case OP_HSPACE:
2823      case OP_NOT_VSPACE:
2824      case OP_VSPACE:
2825      case OP_EXTUNI:
2826      case OP_EODN:
2827      case OP_EOD:
2828      case OP_DOLL:
2829      case OP_DOLLM:
2830      return code;
2831    
2832      case OP_CHAR:
2833      case OP_NOT:
2834      GETCHARINCTEST(chr, code);
2835      list[2] = chr;
2836      list[3] = NOTACHAR;
2837      return code;
2838    
2839      case OP_CHARI:
2840      case OP_NOTI:
2841      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2842      GETCHARINCTEST(chr, code);
2843      list[2] = chr;
2844    
2845    #ifdef SUPPORT_UCP
2846      if (chr < 128 || (chr < 256 && !utf))
2847        list[3] = fcc[chr];
2848      else
2849        list[3] = UCD_OTHERCASE(chr);
2850    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2851      list[3] = (chr < 256) ? fcc[chr] : chr;
2852    #else
2853      list[3] = fcc[chr];
2854    #endif
2855    
2856      /* The othercase might be the same value. */
2857    
2858      if (chr == list[3])
2859        list[3] = NOTACHAR;
2860      else
2861        list[4] = NOTACHAR;
2862      return code;
2863    
2864    #ifdef SUPPORT_UCP
2865      case OP_PROP:
2866      case OP_NOTPROP:
2867      if (code[0] != PT_CLIST)
2868        {
2869        list[2] = code[0];
2870        list[3] = code[1];
2871        return code + 2;
2872        }
2873    
2874      /* Convert only if we have anough space. */
2875    
2876      clist_src = PRIV(ucd_caseless_sets) + code[1];
2877      clist_dest = list + 2;
2878      code += 2;
2879    
2880      do {
2881         /* Early return if there is not enough space. */
2882         if (clist_dest >= list + 8)
2883           {
2884           list[2] = code[0];
2885           list[3] = code[1];
2886           return code;
2887           }
2888         *clist_dest++ = *clist_src;
2889         }
2890       while(*clist_src++ != NOTACHAR);
2891    
2892      /* Enough space to store all characters. */
2893    
2894      list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
2895      return code;
2896    #endif
2897    
2898      case OP_NCLASS:
2899      case OP_CLASS:
2900    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2901      case OP_XCLASS:
2902    
2903      if (c == OP_XCLASS)
2904        end = code + GET(code, 0);
2905      else
2906    #endif
2907        end = code + 32 / sizeof(pcre_uchar);
2908    
2909      switch(*end)
2910        {
2911        case OP_CRSTAR:
2912        case OP_CRMINSTAR:
2913        case OP_CRQUERY:
2914        case OP_CRMINQUERY:
2915        list[1] = TRUE;
2916        end++;
2917        break;
2918    
2919        case OP_CRRANGE:
2920        case OP_CRMINRANGE:
2921        list[1] = (GET2(end, 1) == 0);
2922        end += 1 + 2 * IMM2_SIZE;
2923        break;
2924        }
2925      list[2] = end - code;
2926      return end;
2927      }
2928    return NULL;    /* Opcode not accepted */
2929    }
2930    
2931    
2932    
2933    /*************************************************
2934    *    Scan further character sets for match       *
2935    *************************************************/
2936    
2937    /* Checks whether the base and the current opcode have a common character, in
2938    which case the base cannot be possessified.
2939    
2940    Arguments:
2941      code        points to the byte code
2942      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2943      cd          static compile data
2944      base_list   the data list of the base opcode
2945    
2946    Returns:      TRUE if the auto-possessification is possible
2947    */
2948    
2949    static BOOL
2950    compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
2951      const pcre_uint32* base_list)
2952    {
2953    pcre_uchar c;
2954    pcre_uint32 list[8];
2955    const pcre_uint32* chr_ptr;
2956    const pcre_uint32* ochr_ptr;
2957    const pcre_uint32* list_ptr;
2958    pcre_uint32 chr;
2959    
2960    for(;;)
2961      {
2962      c = *code;
2963    
2964      /* Skip over callouts */
2965    
2966      if (c == OP_CALLOUT)
2967        {
2968        code += PRIV(OP_lengths)[c];
2969        continue;
2970        }
2971    
2972      if (c == OP_ALT)
2973        {
2974        do code += GET(code, 1); while (*code == OP_ALT);
2975        c = *code;
2976        }
2977    
2978      switch(c)
2979        {
2980        case OP_END:
2981        /* TRUE only in greedy case. The non-greedy case could be replaced by an
2982        OP_EXACT, but it is probably not worth it. (And note that OP_EXACT uses
2983        more memory, which we cannot get at this stage.) */
2984    
2985        return base_list[1] != 0;
2986    
2987        case OP_KET:
2988        /* If the bracket is capturing, and referenced by an OP_RECURSE, the
2989        non-greedy case cannot be converted to a possessive form. We do not test
2990        the bracket type at the moment, but we might do it in the future to improve
2991        this condition. (But note that recursive calls are always atomic.) */
2992    
2993        if (base_list[1] == 0) return FALSE;
2994        code += PRIV(OP_lengths)[c];
2995        continue;
2996        }
2997    
2998      /* Check for a supported opcode, and load its properties. */
2999    
3000      code = get_chr_property_list(code, utf, cd->fcc, list);
3001      if (code == NULL) return FALSE;    /* Unsupported */
3002    
3003      /* If either opcode is a small character list, set pointers for comparing
3004      characters from that list with another list, or with a property. */
3005    
3006      if (base_list[0] == OP_CHAR)
3007        {
3008        chr_ptr = base_list + 2;
3009        list_ptr = list;
3010        }
3011      else if (list[0] == OP_CHAR)
3012        {
3013        chr_ptr = list + 2;
3014        list_ptr = base_list;
3015        }
3016    
3017      /* Some property combinations also acceptable. Unicode property opcodes are
3018      processed specially; the rest can be handled with a lookup table. */
3019    
3020      else
3021        {
3022        pcre_uint32 leftop, rightop;
3023    
3024        if (list[1] != 0) return FALSE;   /* Must match at least one character */
3025        leftop = base_list[0];
3026        rightop = list[0];
3027    
3028    #ifdef SUPPORT_UCP
3029        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3030          {
3031          if (rightop == OP_EOD) return TRUE;
3032          if (rightop == OP_PROP || rightop == OP_NOTPROP)
3033            {
3034            int n;
3035            const pcre_uint8 *p;
3036            BOOL same = leftop == rightop;
3037            BOOL lisprop = leftop == OP_PROP;
3038            BOOL risprop = rightop == OP_PROP;
3039            BOOL bothprop = lisprop && risprop;
3040    
3041            /* There's a table that specifies how each combination is to be
3042            processed:
3043              0   Always return FALSE (never auto-possessify)
3044              1   Character groups are distinct (possessify if both are OP_PROP)
3045              2   Check character categories in the same group (general or particular)
3046              3   Return TRUE if the two opcodes are not the same
3047              ... see comments below
3048            */
3049    
3050            n = propposstab[base_list[2]][list[2]];
3051            switch(n)
3052              {
3053              case 0: return FALSE;
3054              case 1: return bothprop;
3055              case 2: return (base_list[3] == list[3]) != same;
3056              case 3: return !same;
3057    
3058              case 4:  /* Left general category, right particular category */
3059              return risprop && catposstab[base_list[3]][list[3]] == same;
3060    
3061              case 5:  /* Right general category, left particular category */
3062              return lisprop && catposstab[list[3]][base_list[3]] == same;
3063    
3064              /* This code is logically tricky. Think hard before fiddling with it.
3065              The posspropstab table has four entries per row. Each row relates to
3066              one of PCRE's special properties such as ALNUM or SPACE or WORD.
3067              Only WORD actually needs all four entries, but using repeats for the
3068              others means they can all use the same code below.
3069    
3070              The first two entries in each row are Unicode general categories, and
3071              apply always, because all the characters they include are part of the
3072              PCRE character set. The third and fourth entries are a general and a
3073              particular category, respectively, that include one or more relevant
3074              characters. One or the other is used, depending on whether the check
3075              is for a general or a particular category. However, in both cases the
3076              category contains more characters than the specials that are defined
3077              for the property being tested against. Therefore, it cannot be used
3078              in a NOTPROP case.
3079    
3080              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3081              Underscore is covered by ucp_P or ucp_Po. */
3082    
3083              case 6:  /* Left alphanum vs right general category */
3084              case 7:  /* Left space vs right general category */
3085              case 8:  /* Left word vs right general category */
3086              p = posspropstab[n-6];
3087              return risprop && lisprop ==
3088                (list[3] != p[0] &&
3089                 list[3] != p[1] &&
3090                (list[3] != p[2] || !lisprop));
3091    
3092              case 9:   /* Right alphanum vs left general category */
3093              case 10:  /* Right space vs left general category */
3094              case 11:  /* Right word vs left general category */
3095              p = posspropstab[n-9];
3096              return lisprop && risprop ==
3097                (base_list[3] != p[0] &&
3098                 base_list[3] != p[1] &&
3099                (base_list[3] != p[2] || !risprop));
3100    
3101              case 12:  /* Left alphanum vs right particular category */
3102              case 13:  /* Left space vs right particular category */
3103              case 14:  /* Left word vs right particular category */
3104              p = posspropstab[n-12];
3105              return risprop && lisprop ==
3106                (catposstab[p[0]][list[3]] &&
3107                 catposstab[p[1]][list[3]] &&
3108                (list[3] != p[3] || !lisprop));
3109    
3110              case 15:  /* Right alphanum vs left particular category */
3111              case 16:  /* Right space vs left particular category */
3112              case 17:  /* Right word vs left particular category */
3113              p = posspropstab[n-15];
3114              return lisprop && risprop ==
3115                (catposstab[p[0]][base_list[3]] &&
3116                 catposstab[p[1]][base_list[3]] &&
3117                (base_list[3] != p[3] || !risprop));
3118              }
3119            }
3120          return FALSE;
3121          }
3122    
3123        else
3124    #endif  /* SUPPORT_UCP */
3125    
3126        return leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3127               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3128               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3129        }
3130    
3131      /* Control reaches here only if one of the items is a small character list.
3132      All characters are checked against the other side. */
3133    
3134      do
3135        {
3136        chr = *chr_ptr;
3137    
3138        switch(list_ptr[0])
3139          {
3140          case OP_CHAR:
3141          ochr_ptr = list_ptr + 2;
3142          do
3143            {
3144            if (chr == *ochr_ptr) return FALSE;
3145            ochr_ptr++;
3146            }
3147          while(*ochr_ptr != NOTACHAR);
3148          break;
3149    
3150          case OP_NOT:
3151          ochr_ptr = list_ptr + 2;
3152          do
3153            {
3154            if (chr == *ochr_ptr)
3155              break;
3156            ochr_ptr++;
3157            }
3158          while(*ochr_ptr != NOTACHAR);
3159          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3160          break;
3161    
3162          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3163          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3164    
3165          case OP_DIGIT:
3166          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3167          break;
3168    
3169          case OP_NOT_DIGIT:
3170          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3171          break;
3172    
3173          case OP_WHITESPACE:
3174          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3175          break;
3176    
3177          case OP_NOT_WHITESPACE:
3178          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3179          break;
3180    
3181          case OP_WORDCHAR:
3182          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3183          break;
3184    
3185          case OP_NOT_WORDCHAR:
3186          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3187          break;
3188    
3189          case OP_HSPACE:
3190          switch(chr)
3191            {
3192            HSPACE_CASES: return FALSE;
3193            default: break;
3194            }
3195          break;
3196    
3197          case OP_NOT_HSPACE:
3198          switch(chr)
3199            {
3200            HSPACE_CASES: break;
3201            default: return FALSE;
3202            }
3203          break;
3204    
3205          case OP_ANYNL:
3206          case OP_VSPACE:
3207          switch(chr)
3208            {
3209            VSPACE_CASES: return FALSE;
3210            default: break;
3211            }
3212          break;
3213    
3214          case OP_NOT_VSPACE:
3215          switch(chr)
3216            {
3217            VSPACE_CASES: break;
3218            default: return FALSE;
3219            }
3220          break;
3221    
3222          case OP_DOLL:
3223          case OP_EODN:
3224          switch (chr)
3225            {
3226            case CHAR_CR:
3227            case CHAR_LF:
3228            case CHAR_VT:
3229            case CHAR_FF:
3230            case CHAR_NEL:
3231    #ifndef EBCDIC
3232            case 0x2028:
3233            case 0x2029:
3234    #endif  /* Not EBCDIC */
3235            return FALSE;
3236            }
3237          break;
3238    
3239          case OP_EOD:    /* Can always possessify before \z */
3240          break;
3241    
3242          case OP_PROP:
3243          case OP_NOTPROP:
3244          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3245                list_ptr[0] == OP_NOTPROP))
3246            return FALSE;
3247          break;
3248    
3249          /* The class comparisons work only when the class is the second item
3250          of the pair, because there are at present no possessive forms of the
3251          class opcodes. Note also that the "code" variable that is used below
3252          points after the second item, and that the pointer for the first item
3253          is not available, so even if there were possessive forms of the class
3254          opcodes, the correct comparison could not be done. */
3255    
3256          case OP_NCLASS:
3257          if (chr > 255) return FALSE;
3258          /* Fall through */
3259    
3260          case OP_CLASS:
3261          if (list_ptr != list) return FALSE;   /* Class is first opcode */
3262          if (chr > 255) break;
3263          if ((((pcre_uint8 *)(code - list_ptr[2] + 1))[chr >> 3] & (1 << (chr & 7))) != 0)
3264            return FALSE;
3265          break;
3266    
3267    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3268          case OP_XCLASS:
3269          if (list_ptr != list) return FALSE;   /* Class is first opcode */
3270          if (PRIV(xclass)(chr, code - list_ptr[2] + 1 + LINK_SIZE, utf))
3271            return FALSE;
3272          break;
3273    #endif
3274    
3275          default:
3276          return FALSE;
3277          }
3278    
3279        chr_ptr++;
3280        }
3281      while(*chr_ptr != NOTACHAR);
3282    
3283      /* At least one character must be matched from this opcode. */
3284    
3285      if (list[1] == 0) return TRUE;
3286      }
3287    
3288    return FALSE;
3289    }
3290    
3291    
3292    
3293    /*************************************************
3294    *    Scan compiled regex for auto-possession     *
3295    *************************************************/
3296    
3297    /* Replaces single character iterations with their possessive alternatives
3298    if appropriate. This function modifies the compiled opcode!
3299    
3300    Arguments:
3301      code        points to start of the byte code
3302      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3303      cd          static compile data
3304    
3305    Returns:      nothing
3306    */
3307    
3308    static void
3309    auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3310    {
3311    register pcre_uchar c;
3312    const pcre_uchar *end;
3313    pcre_uint32 list[8];
3314    
3315    for (;;)
3316      {
3317      c = *code;
3318    
3319      if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3320        {
3321        c -= get_repeat_base(c) - OP_STAR;
3322        end = (c <= OP_MINUPTO) ?
3323          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3324        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3325    
3326        if (end != NULL && compare_opcodes(end, utf, cd, list))
3327          {
3328          switch(c)
3329            {
3330            case OP_STAR:
3331            *code += OP_POSSTAR - OP_STAR;
3332            break;
3333    
3334            case OP_MINSTAR:
3335            *code += OP_POSSTAR - OP_MINSTAR;
3336            break;
3337    
3338            case OP_PLUS:
3339            *code += OP_POSPLUS - OP_PLUS;
3340            break;
3341    
3342            case OP_MINPLUS:
3343            *code += OP_POSPLUS - OP_MINPLUS;
3344            break;
3345    
3346            case OP_QUERY:
3347            *code += OP_POSQUERY - OP_QUERY;
3348            break;
3349    
3350            case OP_MINQUERY:
3351            *code += OP_POSQUERY - OP_MINQUERY;
3352            break;
3353    
3354            case OP_UPTO:
3355            *code += OP_POSUPTO - OP_UPTO;
3356            break;
3357    
3358            case OP_MINUPTO:
3359            *code += OP_MINUPTO - OP_UPTO;
3360            break;
3361            }
3362          }
3363        c = *code;
3364        }
3365    
3366      switch(c)
3367        {
3368        case OP_END:
3369        return;
3370    
3371        case OP_TYPESTAR:
3372        case OP_TYPEMINSTAR:
3373        case OP_TYPEPLUS:
3374        case OP_TYPEMINPLUS:
3375        case OP_TYPEQUERY:
3376        case OP_TYPEMINQUERY:
3377        case OP_TYPEPOSSTAR:
3378        case OP_TYPEPOSPLUS:
3379        case OP_TYPEPOSQUERY:
3380        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3381        break;
3382    
3383        case OP_TYPEUPTO:
3384        case OP_TYPEMINUPTO:
3385        case OP_TYPEEXACT:
3386        case OP_TYPEPOSUPTO:
3387        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3388          code += 2;
3389        break;
3390    
3391        case OP_XCLASS:
3392        code += GET(code, 1);
3393        break;
3394    
3395        case OP_MARK:
3396        case OP_PRUNE_ARG:
3397        case OP_SKIP_ARG:
3398        case OP_THEN_ARG:
3399        code += code[1];
3400        break;
3401        }
3402    
3403      /* Add in the fixed length from the table */
3404    
3405      code += PRIV(OP_lengths)[c];
3406    
3407      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3408      a multi-byte character. The length in the table is a minimum, so we have to
3409      arrange to skip the extra bytes. */
3410    
3411    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3412      if (utf) switch(c)
3413        {
3414        case OP_CHAR:
3415        case OP_CHARI:
3416        case OP_NOT:
3417        case OP_NOTI:
3418        case OP_STAR:
3419        case OP_MINSTAR:
3420        case OP_PLUS:
3421        case OP_MINPLUS:
3422        case OP_QUERY:
3423        case OP_MINQUERY:
3424        case OP_UPTO:
3425        case OP_MINUPTO:
3426        case OP_EXACT:
3427        case OP_POSSTAR:
3428        case OP_POSPLUS:
3429        case OP_POSQUERY:
3430        case OP_POSUPTO:
3431        case OP_STARI:
3432        case OP_MINSTARI:
3433        case OP_PLUSI:
3434        case OP_MINPLUSI:
3435        case OP_QUERYI:
3436        case OP_MINQUERYI:
3437        case OP_UPTOI:
3438      case OP_MINUPTOI:      case OP_MINUPTOI:
3439      case OP_POSUPTO:      case OP_EXACTI:
3440        case OP_POSSTARI:
3441        case OP_POSPLUSI:
3442        case OP_POSQUERYI:
3443      case OP_POSUPTOI:      case OP_POSUPTOI:
3444      if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);      case OP_NOTSTAR:
3445      break;      case OP_NOTMINSTAR:
3446  #endif      case OP_NOTPLUS:
3447        case OP_NOTMINPLUS:
3448      /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument      case OP_NOTQUERY:
3449      string. */      case OP_NOTMINQUERY:
3450        case OP_NOTUPTO:
3451      case OP_MARK:      case OP_NOTMINUPTO:
3452      case OP_PRUNE_ARG:      case OP_NOTEXACT:
3453      case OP_SKIP_ARG:      case OP_NOTPOSSTAR:
3454      code += code[1];      case OP_NOTPOSPLUS:
3455      break;      case OP_NOTPOSQUERY:
3456        case OP_NOTPOSUPTO:
3457      case OP_THEN_ARG:      case OP_NOTSTARI:
3458      code += code[1];      case OP_NOTMINSTARI:
3459      break;      case OP_NOTPLUSI:
3460        case OP_NOTMINPLUSI:
3461      /* None of the remaining opcodes are required to match a character. */      case OP_NOTQUERYI:
3462        case OP_NOTMINQUERYI:
3463      default:      case OP_NOTUPTOI:
3464        case OP_NOTMINUPTOI:
3465        case OP_NOTEXACTI:
3466        case OP_NOTPOSSTARI:
3467        case OP_NOTPOSPLUSI:
3468        case OP_NOTPOSQUERYI:
3469        case OP_NOTPOSUPTOI:
3470        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3471      break;      break;
3472      }      }
3473    #else
3474      (void)(utf);  /* Keep compiler happy by referencing function argument */
3475    #endif
3476    }    }
   
 return TRUE;  
 }  
   
   
   
 /*************************************************  
 *    Scan compiled regex for non-emptiness       *  
 *************************************************/  
   
 /* This function is called to check for left recursive calls. We want to check  
 the current branch of the current pattern to see if it could match the empty  
 string. If it could, we must look outwards for branches at other levels,  
 stopping when we pass beyond the bracket which is the subject of the recursion.  
 This function is called only during the real compile, not during the  
 pre-compile.  
   
 Arguments:  
   code        points to start of the recursion  
   endcode     points to where to stop (current RECURSE item)  
   bcptr       points to the chain of current (unclosed) branch starts  
   utf         TRUE if in UTF-8 / UTF-16 mode  
   cd          pointers to tables etc  
   
 Returns:      TRUE if what is matched could be empty  
 */  
   
 static BOOL  
 could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,  
   branch_chain *bcptr, BOOL utf, compile_data *cd)  
 {  
 while (bcptr != NULL && bcptr->current_branch >= code)  
   {  
   if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))  
     return FALSE;  
   bcptr = bcptr->outer;  
   }  
 return TRUE;  
3477  }  }
3478    
3479    
# Line 2643  Returns:   TRUE or FALSE Line 3523  Returns:   TRUE or FALSE
3523  static BOOL  static BOOL
3524  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3525  {  {
3526  int terminator;          /* Don't combine these lines; the Solaris cc */  pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
3527  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
3528  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != CHAR_NULL; ptr++)
3529    {    {
3530    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3531      ptr++;      ptr++;
# Line 2692  register int yield = 0; Line 3572  register int yield = 0;
3572  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
3573    {    {
3574    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
3575      STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3576    pn += posix_name_lengths[yield] + 1;    pn += posix_name_lengths[yield] + 1;
3577    yield++;    yield++;
3578    }    }
# Line 2724  value in the reference (which is a group Line 3604  value in the reference (which is a group
3604  Arguments:  Arguments:
3605    group      points to the start of the group    group      points to the start of the group
3606    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
3607    utf        TRUE in UTF-8 / UTF-16 mode    utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
3608    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
3609    save_hwm   the hwm forward reference pointer at the start of the group    save_hwm   the hwm forward reference pointer at the start of the group
3610    
# Line 2747  while ((ptr = (pcre_uchar *)find_recurse Line 3627  while ((ptr = (pcre_uchar *)find_recurse
3627    
3628    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3629      {      {
3630      offset = GET(hc, 0);      offset = (int)GET(hc, 0);
3631      if (cd->start_code + offset == ptr + 1)      if (cd->start_code + offset == ptr + 1)
3632        {        {
3633        PUT(hc, 0, offset + adjust);        PUT(hc, 0, offset + adjust);
# Line 2760  while ((ptr = (pcre_uchar *)find_recurse Line 3640  while ((ptr = (pcre_uchar *)find_recurse
3640    
3641    if (hc >= cd->hwm)    if (hc >= cd->hwm)
3642      {      {
3643      offset = GET(ptr, 1);      offset = (int)GET(ptr, 1);
3644      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3645      }      }
3646    
# Line 2828  PUT(previous_callout, 2 + LINK_SIZE, len Line 3708  PUT(previous_callout, 2 + LINK_SIZE, len
3708  *************************************************/  *************************************************/
3709    
3710  /* This function is passed the start and end of a class range, in UTF-8 mode  /* This function is passed the start and end of a class range, in UTF-8 mode
3711  with UCP support. It searches up the characters, looking for internal ranges of  with UCP support. It searches up the characters, looking for ranges of
3712  characters in the "other" case. Each call returns the next one, updating the  characters in the "other" case. Each call returns the next one, updating the
3713  start address.  start address. A character with multiple other cases is returned on its own
3714    with a special return value.
3715    
3716  Arguments:  Arguments:
3717    cptr        points to starting character value; updated    cptr        points to starting character value; updated
# Line 2838  Arguments: Line 3719  Arguments:
3719    ocptr       where to put start of othercase range    ocptr       where to put start of othercase range
3720    odptr       where to put end of othercase range    odptr       where to put end of othercase range
3721    
3722  Yield:        TRUE when range returned; FALSE when no more  Yield:        -1 when no more
3723                   0 when a range is returned
3724                  >0 the CASESET offset for char with multiple other cases
3725                    in this case, ocptr contains the original
3726  */  */
3727    
3728  static BOOL  static int
3729  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
3730    unsigned int *odptr)    pcre_uint32 *odptr)
3731  {  {
3732  unsigned int c, othercase, next;  pcre_uint32 c, othercase, next;
3733    unsigned int co;
3734    
3735    /* Find the first character that has an other case. If it has multiple other
3736    cases, return its case offset value. */
3737    
3738  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
3739    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }    {
3740      if ((co = UCD_CASESET(c)) != 0)
3741        {
3742        *ocptr = c++;   /* Character that has the set */
3743        *cptr = c;      /* Rest of input range */
3744        return (int)co;
3745        }
3746      if ((othercase = UCD_OTHERCASE(c)) != c) break;
3747      }
3748    
3749  if (c > d) return FALSE;  if (c > d) return -1;  /* Reached end of range */
3750    
3751  *ocptr = othercase;  *ocptr = othercase;
3752  next = othercase + 1;  next = othercase + 1;
# Line 2861  for (++c; c <= d; c++) Line 3757  for (++c; c <= d; c++)
3757    next++;    next++;
3758    }    }
3759    
3760  *odptr = next - 1;  *odptr = next - 1;     /* End of othercase range */
3761  *cptr = c;  *cptr = c;             /* Rest of input range */
3762    return 0;
 return TRUE;  
 }  
   
   
   
 /*************************************************  
 *        Check a character and a property        *  
 *************************************************/  
   
 /* This function is called by check_auto_possessive() when a property item  
 is adjacent to a fixed character.  
   
 Arguments:  
   c            the character  
   ptype        the property type  
   pdata        the data for the type  
   negated      TRUE if it's a negated property (\P or \p{^)  
   
 Returns:       TRUE if auto-possessifying is OK  
 */  
   
 static BOOL  
 check_char_prop(int c, int ptype, int pdata, BOOL negated)  
 {  
 const ucd_record *prop = GET_UCD(c);  
 switch(ptype)  
   {  
   case PT_LAMP:  
   return (prop->chartype == ucp_Lu ||  
           prop->chartype == ucp_Ll ||  
           prop->chartype == ucp_Lt) == negated;  
   
   case PT_GC:  
   return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;  
   
   case PT_PC:  
   return (pdata == prop->chartype) == negated;  
   
   case PT_SC:  
   return (pdata == prop->script) == negated;  
   
   /* These are specials */  
   
   case PT_ALNUM:  
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||  
           PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;  
   
   case PT_SPACE:    /* Perl space */  
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||  
           c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)  
           == negated;  
   
   case PT_PXSPACE:  /* POSIX space */  
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||  
           c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||  
           c == CHAR_FF || c == CHAR_CR)  
           == negated;  
   
   case PT_WORD:  
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||  
           PRIV(ucp_gentype)[prop->chartype] == ucp_N ||  
           c == CHAR_UNDERSCORE) == negated;  
   }  
 return FALSE;  
3763  }  }
3764  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3765    
3766    
3767    
3768  /*************************************************  /*************************************************
3769  *     Check if auto-possessifying is possible    *  *        Add a character or range to a class     *
3770  *************************************************/  *************************************************/
3771    
3772  /* This function is called for unlimited repeats of certain items, to see  /* This function packages up the logic of adding a character or range of
3773  whether the next thing could possibly match the repeated item. If not, it makes  characters to a class. The character values in the arguments will be within the
3774  sense to automatically possessify the repeated item.  valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
3775    mutually recursive with the function immediately below.
3776    
3777  Arguments:  Arguments:
3778    previous      pointer to the repeated opcode    classbits     the bit map for characters < 256
3779    utf           TRUE in UTF-8 / UTF-16 mode    uchardptr     points to the pointer for extra data
3780    ptr           next character in pattern    options       the options word
   options       options bits  
3781    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
3782      start         start of range character
3783      end           end of range character
3784    
3785  Returns:        TRUE if possessifying is wanted  Returns:        the number of < 256 characters added
3786                    the pointer to extra data is updated
3787  */  */
3788    
3789  static BOOL  static int
3790  check_auto_possessive(const pcre_uchar *previous, BOOL utf,  add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
3791    const pcre_uchar *ptr, int options, compile_data *cd)    compile_data *cd, pcre_uint32 start, pcre_uint32 end)
3792  {  {
3793  pcre_int32 c, next;  pcre_uint32 c;
3794  int op_code = *previous++;  int n8 = 0;
   
 /* Skip whitespace and comments in extended mode */  
   
 if ((options & PCRE_EXTENDED) != 0)  
   {  
   for (;;)  
     {  
     while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;  
     if (*ptr == CHAR_NUMBER_SIGN)  
       {  
       ptr++;  
       while (*ptr != 0)  
         {  
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
       }  
     else break;  
     }  
   }  
   
 /* If the next item is one that we can handle, get its value. A non-negative  
 value is a character, a negative value is an escape value. */  
   
 if (*ptr == CHAR_BACKSLASH)  
   {  
   int temperrorcode = 0;  
   next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);  
   if (temperrorcode != 0) return FALSE;  
   ptr++;    /* Point after the escape sequence */  
   }  
 else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)  
   {  
 #ifdef SUPPORT_UTF  
   if (utf) { GETCHARINC(next, ptr); } else  
 #endif  
   next = *ptr++;  
   }  
 else return FALSE;  
3795    
3796  /* Skip whitespace and comments in extended mode */  /* If caseless matching is required, scan the range and process alternate
3797    cases. In Unicode, there are 8-bit characters that have alternate cases that
3798    are greater than 255 and vice-versa. Sometimes we can just extend the original
3799    range. */
3800    
3801  if ((options & PCRE_EXTENDED) != 0)  if ((options & PCRE_CASELESS) != 0)
3802    {    {
   for (;;)  
     {  
     while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;  
     if (*ptr == CHAR_NUMBER_SIGN)  
       {  
       ptr++;  
       while (*ptr != 0)  
         {  
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
       }  
     else break;  
     }  
   }  
   
 /* If the next thing is itself optional, we have to give up. */  
   
 if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  
   STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)  
     return FALSE;  
   
 /* Now compare the next item with the previous opcode. First, handle cases when  
 the next item is a character. */  
   
 if (next >= 0) switch(op_code)  
   {  
   case OP_CHAR:  
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   return c != next;  
   
   /* For CHARI (caseless character) we must check the other case. If we have  
   Unicode property support, we can use it to test the other case of  
   high-valued characters. */  
   
   case OP_CHARI:  
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   if (c == next) return FALSE;  
 #ifdef SUPPORT_UTF  
   if (utf)  
     {  
     unsigned int othercase;  
     if (next < 128) othercase = cd->fcc[next]; else  
3803  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3804      othercase = UCD_OTHERCASE((unsigned int)next);    if ((options & PCRE_UTF8) != 0)
 #else  
     othercase = NOTACHAR;  
 #endif  
     return (unsigned int)c != othercase;  
     }  
   else  
 #endif  /* SUPPORT_UTF */  
   return (c != TABLE_GET(next, cd->fcc, next));  /* Non-UTF-8 mode */  
   
   /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These  
   opcodes are not used for multi-byte characters, because they are coded using  
   an XCLASS instead. */  
   
   case OP_NOT:  
   return (c = *previous) == next;  
   
   case OP_NOTI:  
   if ((c = *previous) == next) return TRUE;  
 #ifdef SUPPORT_UTF  
   if (utf)  
3805      {      {
3806      unsigned int othercase;      int rc;
3807      if (next < 128) othercase = cd->fcc[next]; else      pcre_uint32 oc, od;
 #ifdef SUPPORT_UCP  
     othercase = UCD_OTHERCASE(next);  
 #else  
     othercase = NOTACHAR;  
 #endif  
     return (unsigned int)c == othercase;  
     }  
   else  
 #endif  /* SUPPORT_UTF */  
   return (c == TABLE_GET(next, cd->fcc, next));  /* Non-UTF-8 mode */  
3808    
3809    /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
3810    When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */      c = start;
3811    
3812    case OP_DIGIT:      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
3813    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;        {
3814          /* Handle a single character that has more than one other case. */
   case OP_NOT_DIGIT:  
   return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;  
   
   case OP_WHITESPACE:  
   return next > 127 || (cd->ctypes[next] & ctype_space) == 0;  
   
   case OP_NOT_WHITESPACE:  
   return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;  
   
   case OP_WORDCHAR:  
   return next > 127 || (cd->ctypes[next] & ctype_word) == 0;  
3815    
3816    case OP_NOT_WORDCHAR:        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
3817    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;          PRIV(ucd_caseless_sets) + rc, oc);
3818    
3819    case OP_HSPACE:        /* Do nothing if the other case range is within the original range. */
   case OP_NOT_HSPACE:  
   switch(next)  
     {  
     case 0x09:  
     case 0x20:  
     case 0xa0:  
     case 0x1680:  
     case 0x180e:  
     case 0x2000:  
     case 0x2001:  
     case 0x2002:  
     case 0x2003:  
     case 0x2004:  
     case 0x2005:  
     case 0x2006:  
     case 0x2007:  
     case 0x2008:  
     case 0x2009:  
     case 0x200A:  
     case 0x202f:  
     case 0x205f:  
     case 0x3000:  
     return op_code == OP_NOT_HSPACE;  
     default:  
     return op_code != OP_NOT_HSPACE;  
     }  
3820    
3821    case OP_ANYNL:        else if (oc >= start && od <= end) continue;
   case OP_VSPACE:  
   case OP_NOT_VSPACE:  
   switch(next)  
     {  
     case 0x0a:  
     case 0x0b:  
     case 0x0c:  
     case 0x0d:  
     case 0x85:  
     case 0x2028:  
     case 0x2029:  
     return op_code == OP_NOT_VSPACE;  
     default:  
     return op_code != OP_NOT_VSPACE;  
     }  
3822    
3823  #ifdef SUPPORT_UCP        /* Extend the original range if there is overlap, noting that if oc < c, we
3824    case OP_PROP:        can't have od > end because a subrange is always shorter than the basic
3825    return check_char_prop(next, previous[0], previous[1], FALSE);        range. Otherwise, use a recursive call to add the additional range. */
3826    
3827    case OP_NOTPROP:        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
3828    return check_char_prop(next, previous[0], previous[1], TRUE);        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
3829  #endif        else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
3830          }
3831        }
3832      else
3833    #endif  /* SUPPORT_UCP */
3834    
3835    default:    /* Not UTF-mode, or no UCP */
   return FALSE;  
   }  
3836    
3837      for (c = start; c <= end && c < 256; c++)
3838        {
3839        SETBIT(classbits, cd->fcc[c]);
3840        n8++;
3841        }
3842      }
3843    
3844  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP  /* Now handle the original range. Adjust the final value according to the bit
3845  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are  length - this means that the same lists of (e.g.) horizontal spaces can be used
3846  generated only when PCRE_UCP is *not* set, that is, when only ASCII  in all cases. */
 characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are  
 replaced by OP_PROP codes when PCRE_UCP is set. */  
3847    
3848  switch(op_code)  #if defined COMPILE_PCRE8
   {  
   case OP_CHAR:  
   case OP_CHARI:  
3849  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
3850    GETCHARTEST(c, previous);    if ((options & PCRE_UTF8) == 0)
 #else  
   c = *previous;  
3851  #endif  #endif
3852    switch(-next)    if (end > 0xff) end = 0xff;
     {  
     case ESC_d:  
     return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;  
3853    
3854      case ESC_D:  #elif defined COMPILE_PCRE16
3855      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;  #ifdef SUPPORT_UTF
3856      if ((options & PCRE_UTF16) == 0)
3857    #endif
3858      if (end > 0xffff) end = 0xffff;
3859    
3860      case ESC_s:  #endif /* COMPILE_PCRE[8|16] */
     return c > 127 || (cd->ctypes[c] & ctype_space) == 0;  
3861    
3862      case ESC_S:  /* If all characters are less than 256, use the bit map. Otherwise use extra
3863      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;  data. */
3864    
3865      case ESC_w:  if (end < 0x100)
3866      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;    {
3867      for (c = start; c <= end; c++)
3868        {
3869        n8++;
3870        SETBIT(classbits, c);
3871        }
3872      }
3873    
3874      case ESC_W:  else
3875      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;    {
3876      pcre_uchar *uchardata = *uchardptr;
3877    
3878      case ESC_h:  #ifdef SUPPORT_UTF
3879      case ESC_H:    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
3880      switch(c)      {
3881        if (start < end)
3882        {        {
3883        case 0x09:        *uchardata++ = XCL_RANGE;
3884        case 0x20:        uchardata += PRIV(ord2utf)(start, uchardata);
3885        case 0xa0:        uchardata += PRIV(ord2utf)(end, uchardata);
       case 0x1680:  
       case 0x180e:  
       case 0x2000:  
       case 0x2001:  
       case 0x2002:  
       case 0x2003:  
       case 0x2004:  
       case 0x2005:  
       case 0x2006:  
       case 0x2007:  
       case 0x2008:  
       case 0x2009:  
       case 0x200A:  
       case 0x202f:  
       case 0x205f:  
       case 0x3000:  
       return -next != ESC_h;  
       default:  
       return -next == ESC_h;  
3886        }        }
3887        else if (start == end)
     case ESC_v:  
     case ESC_V:  
     switch(c)  
3888        {        {
3889        case 0x0a:        *uchardata++ = XCL_SINGLE;
3890        case 0x0b:        uchardata += PRIV(ord2utf)(start, uchardata);
       case 0x0c:  
       case 0x0d:  
       case 0x85:  
       case 0x2028:  
       case 0x2029:  
       return -next != ESC_v;  
       default:  
       return -next == ESC_v;  
       }  
   
     /* When PCRE_UCP is set, these values get generated for \d etc. Find  
     their substitutions and process them. The result will always be either  
     -ESC_p or -ESC_P. Then fall through to process those values. */  
   
 #ifdef SUPPORT_UCP  
     case ESC_du:  
     case ESC_DU:  
     case ESC_wu:  
     case ESC_WU:  
     case ESC_su:  
     case ESC_SU:  
       {  
       int temperrorcode = 0;  
       ptr = substitutes[-next - ESC_DU];  
       next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);  
       if (temperrorcode != 0) return FALSE;  
       ptr++;    /* For compatibility */  
3891        }        }
3892      /* Fall through */      }
3893      else
3894    #endif  /* SUPPORT_UTF */
3895    
3896      case ESC_p:    /* Without UTF support, character values are constrained by the bit length,
3897      case ESC_P:    and can only be > 256 for 16-bit and 32-bit libraries. */
       {  
       int ptype, pdata, errorcodeptr;  
       BOOL negated;  
3898    
3899        ptr--;      /* Make ptr point at the p or P */  #ifdef COMPILE_PCRE8
3900        ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);      {}
3901        if (ptype < 0) return FALSE;  #else
3902        ptr++;      /* Point past the final curly ket */    if (start < end)
3903        {
3904        *uchardata++ = XCL_RANGE;
3905        *uchardata++ = start;
3906        *uchardata++ = end;
3907        }
3908      else if (start == end)
3909        {
3910        *uchardata++ = XCL_SINGLE;
3911        *uchardata++ = start;
3912        }
3913    #endif
3914    
3915        /* If the property item is optional, we have to give up. (When generated    *uchardptr = uchardata;   /* Updata extra data pointer */
3916        from \d etc by PCRE_UCP, this test will have been applied much earlier,    }
       to the original \d etc. At this point, ptr will point to a zero byte. */  
3917    
3918        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  return n8;    /* Number of 8-bit characters */
3919          STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)  }
           return FALSE;  
3920    
       /* Do the property check. */  
3921    
       return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);  
       }  
 #endif  
3922    
     default:  
     return FALSE;  
     }  
3923    
3924    /* In principle, support for Unicode properties should be integrated here as  /*************************************************
3925    well. It means re-organizing the above code so as to get hold of the property  *        Add a list of characters to a class     *
3926    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.) */  
3927    
3928    case OP_DIGIT:  /* This function is used for adding a list of case-equivalent characters to a
3929    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||  class, and also for adding a list of horizontal or vertical whitespace. If the
3930           next == -ESC_h || next == -ESC_v || next == -ESC_R;  list is in order (which it should be), ranges of characters are detected and
3931    handled appropriately. This function is mutually recursive with the function
3932    above.
3933    
3934    case OP_NOT_DIGIT:  Arguments:
3935    return next == -ESC_d;    classbits     the bit map for characters < 256
3936      uchardptr     points to the pointer for extra data
3937      options       the options word
3938      cd            contains pointers to tables etc.
3939      p             points to row of 32-bit values, terminated by NOTACHAR
3940      except        character to omit; this is used when adding lists of
3941                      case-equivalent characters to avoid including the one we
3942                      already know about
3943    
3944    case OP_WHITESPACE:  Returns:        the number of < 256 characters added
3945    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;                  the pointer to extra data is updated
3946    */
3947    
3948    case OP_NOT_WHITESPACE:  static int
3949    return next == -ESC_s || next == -ESC_h || next == -ESC_v;  add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
3950      compile_data *cd, const pcre_uint32 *p, unsigned int except)
3951    {
3952    int n8 = 0;
3953    while (p[0] < NOTACHAR)
3954      {
3955      int n = 0;
3956      if (p[0] != except)
3957        {
3958        while(p[n+1] == p[0] + n + 1) n++;
3959        n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
3960        }
3961      p += n + 1;
3962      }
3963    return n8;
3964    }
3965    
   case OP_HSPACE:  
   return next == -ESC_S || next == -ESC_H || next == -ESC_d ||  
          next == -ESC_w || next == -ESC_v || next == -ESC_R;  
3966    
   case OP_NOT_HSPACE:  
   return next == -ESC_h;  
3967    
3968    /* Can't have \S in here because VT matches \S (Perl anomaly) */  /*************************************************
3969    case OP_ANYNL:  *    Add characters not in a list to a class     *
3970    case OP_VSPACE:  *************************************************/
   return next == -ESC_V || next == -ESC_d || next == -ESC_w;  
3971    
3972    case OP_NOT_VSPACE:  /* This function is used for adding the complement of a list of horizontal or
3973    return next == -ESC_v || next == -ESC_R;  vertical whitespace to a class. The list must be in order.
3974    
3975    case OP_WORDCHAR:  Arguments:
3976    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||    classbits     the bit map for characters < 256
3977           next == -ESC_v || next == -ESC_R;    uchardptr     points to the pointer for extra data
3978      options       the options word
3979      cd            contains pointers to tables etc.
3980      p             points to row of 32-bit values, terminated by NOTACHAR
3981    
3982    case OP_NOT_WORDCHAR:  Returns:        the number of < 256 characters added
3983    return next == -ESC_w || next == -ESC_d;                  the pointer to extra data is updated
3984    */
3985    
3986    default:  static int
3987    return FALSE;  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
3988      int options, compile_data *cd, const pcre_uint32 *p)
3989    {
3990    BOOL utf = (options & PCRE_UTF8) != 0;
3991    int n8 = 0;
3992    if (p[0] > 0)
3993      n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
3994    while (p[0] < NOTACHAR)
3995      {
3996      while (p[1] == p[0] + 1) p++;
3997      n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
3998        (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
3999      p++;
4000    }    }
4001    return n8;
 /* Control does not reach here */  
4002  }  }
4003    
4004    
# Line 3362  to find out the amount of memory needed, Line 4014  to find out the amount of memory needed,
4014  phase. The value of lengthptr distinguishes the two phases.  phase. The value of lengthptr distinguishes the two phases.
4015    
4016  Arguments:  Arguments:
4017    optionsptr     pointer to the option bits    optionsptr        pointer to the option bits
4018    codeptr        points to the pointer to the current code point    codeptr           points to the pointer to the current code point
4019    ptrptr         points to the current pattern pointer    ptrptr            points to the current pattern pointer
4020    errorcodeptr   points to error code variable    errorcodeptr      points to error code variable
4021    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr      place to put the first required character
4022    reqcharptr     set to the last literal character required, else < 0    firstcharflagsptr place to put the first character flags, or a negative number
4023    bcptr          points to current branch chain    reqcharptr        place to put the last required character
4024    cond_depth     conditional nesting depth    reqcharflagsptr   place to put the last required character flags, or a negative number
4025    cd             contains pointers to tables etc.    bcptr             points to current branch chain
4026    lengthptr      NULL during the real compile phase    cond_depth        conditional nesting depth
4027                   points to length accumulator during pre-compile phase    cd                contains pointers to tables etc.
4028      lengthptr         NULL during the real compile phase
4029                        points to length accumulator during pre-compile phase
4030    
4031  Returns:         TRUE on success  Returns:            TRUE on success
4032                   FALSE, with *errorcodeptr set non-zero on error                      FALSE, with *errorcodeptr set non-zero on error
4033  */  */
4034    
4035  static BOOL  static BOOL
4036  compile_branch(int *optionsptr, pcre_uchar **codeptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
4037    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,    const pcre_uchar **ptrptr, int *errorcodeptr,
4038    pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,    pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
4039      pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
4040      branch_chain *bcptr, int cond_depth,
4041    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
4042  {  {
4043  int repeat_type, op_type;  int repeat_type, op_type;
4044  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
4045  int bravalue = 0;  int bravalue = 0;
4046  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
4047  pcre_int32 firstchar, reqchar;  pcre_uint32 firstchar, reqchar;
4048  pcre_int32 zeroreqchar, zerofirstchar;  pcre_int32 firstcharflags, reqcharflags;
4049    pcre_uint32 zeroreqchar, zerofirstchar;
4050    pcre_int32 zeroreqcharflags, zerofirstcharflags;
4051  pcre_int32 req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
4052  int options = *optionsptr;               /* May change dynamically */  int options = *optionsptr;               /* May change dynamically */
4053  int after_manual_callout = 0;  int after_manual_callout = 0;
4054  int length_prevgroup = 0;  int length_prevgroup = 0;
4055  register int c;  register pcre_uint32 c;
4056    int escape;
4057  register pcre_uchar *code = *codeptr;  register pcre_uchar *code = *codeptr;
4058  pcre_uchar *last_code = code;  pcre_uchar *last_code = code;
4059  pcre_uchar *orig_code = code;  pcre_uchar *orig_code = code;
# Line 3414  must not do this for other options (e.g. Line 4073  must not do this for other options (e.g.
4073  dynamically as we process the pattern. */  dynamically as we process the pattern. */
4074    
4075  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4076  /* PCRE_UTF16 has the same value as PCRE_UTF8. */  /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
4077  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
4078    #ifndef COMPILE_PCRE32
4079  pcre_uchar utf_chars[6];  pcre_uchar utf_chars[6];
4080    #endif
4081  #else  #else
4082  BOOL utf = FALSE;  BOOL utf = FALSE;
4083  #endif  #endif
4084    
4085  /* Helper variables for OP_XCLASS opcode (for characters > 255). */  /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
4086    class_uchardata always so that it can be passed to add_to_class() always,
4087    though it will not be used in non-UTF 8-bit cases. This avoids having to supply
4088    alternative calls for the different cases. */
4089    
4090    pcre_uchar *class_uchardata;
4091  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4092  BOOL xclass;  BOOL xclass;
 pcre_uchar *class_uchardata;  
4093  pcre_uchar *class_uchardata_base;  pcre_uchar *class_uchardata_base;
4094  #endif  #endif
4095    
# Line 3448  to take the zero repeat into account. Th Line 4112  to take the zero repeat into account. Th
4112  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
4113  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
4114    
4115  firstchar = reqchar = zerofirstchar = zeroreqchar = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
4116    firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
4117    
4118  /* The variable req_caseopt contains either the REQ_CASELESS value  /* The variable req_caseopt contains either the REQ_CASELESS value
4119  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 3469  for (;; ptr++) Line 4134  for (;; ptr++)
4134    BOOL is_recurse;    BOOL is_recurse;
4135    BOOL reset_bracount;    BOOL reset_bracount;
4136    int class_has_8bitchar;    int class_has_8bitchar;
4137    int class_single_char;    int class_one_char;
4138    int newoptions;    int newoptions;
4139    int recno;    int recno;
4140    int refsign;    int refsign;
4141    int skipbytes;    int skipbytes;
4142    int subreqchar;    pcre_uint32 subreqchar, subfirstchar;
4143    int subfirstchar;    pcre_int32 subreqcharflags, subfirstcharflags;
4144    int terminator;    int terminator;
4145    int mclength;    unsigned int mclength;
4146    int tempbracount;    unsigned int tempbracount;
4147      pcre_uint32 ec;
4148    pcre_uchar mcbuffer[8];    pcre_uchar mcbuffer[8];
4149    
4150    /* Get next character in the pattern */    /* Get next character in the pattern */
# Line 3488  for (;; ptr++) Line 4154  for (;; ptr++)
4154    /* 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
4155    string. Nesting only happens one level deep. */    string. Nesting only happens one level deep. */
4156    
4157    if (c == 0 && nestptr != NULL)    if (c == CHAR_NULL && nestptr != NULL)
4158      {      {
4159      ptr = nestptr;      ptr = nestptr;
4160      nestptr = NULL;      nestptr = NULL;
# Line 3563  for (;; ptr++) Line 4229  for (;; ptr++)
4229    
4230    /* 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 */
4231    
4232    if (inescq && c != 0)    if (inescq && c != CHAR_NULL)
4233      {      {
4234      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4235        {        {
# Line 3611  for (;; ptr++) Line 4277  for (;; ptr++)
4277      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
4278        {        {
4279        ptr++;        ptr++;
4280        while (*ptr != 0)        while (*ptr != CHAR_NULL)
4281          {          {
4282          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
4283          ptr++;          ptr++;
# Line 3619  for (;; ptr++) Line 4285  for (;; ptr++)
4285          if (utf) FORWARDCHAR(ptr);          if (utf) FORWARDCHAR(ptr);
4286  #endif  #endif
4287          }          }
4288        if (*ptr != 0) continue;        if (*ptr != CHAR_NULL) continue;
4289    
4290        /* Else fall through to handle end of string */        /* Else fall through to handle end of string */
4291        c = 0;        c = 0;
# Line 3641  for (;; ptr++) Line 4307  for (;; ptr++)
4307      case CHAR_VERTICAL_LINE:       /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
4308      case CHAR_RIGHT_PARENTHESIS:      case CHAR_RIGHT_PARENTHESIS:
4309      *firstcharptr = firstchar;      *firstcharptr = firstchar;
4310        *firstcharflagsptr = firstcharflags;
4311      *reqcharptr = reqchar;      *reqcharptr = reqchar;
4312        *reqcharflagsptr = reqcharflags;
4313      *codeptr = code;      *codeptr = code;
4314      *ptrptr = ptr;      *ptrptr = ptr;
4315      if (lengthptr != NULL)      if (lengthptr != NULL)
# Line 3665  for (;; ptr++) Line 4333  for (;; ptr++)
4333      previous = NULL;      previous = NULL;
4334      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
4335        {        {
4336        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;        if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4337        *code++ = OP_CIRCM;        *code++ = OP_CIRCM;
4338        }        }
4339      else *code++ = OP_CIRC;      else *code++ = OP_CIRC;
# Line 3680  for (;; ptr++) Line 4348  for (;; ptr++)
4348      repeats. The value of reqchar doesn't change either. */      repeats. The value of reqchar doesn't change either. */
4349    
4350      case CHAR_DOT:      case CHAR_DOT:
4351      if (firstchar == REQ_UNSET) firstchar = REQ_NONE;      if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4352      zerofirstchar = firstchar;      zerofirstchar = firstchar;
4353        zerofirstcharflags = firstcharflags;
4354      zeroreqchar = reqchar;      zeroreqchar = reqchar;
4355        zeroreqcharflags = reqcharflags;
4356      previous = code;      previous = code;
4357      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
4358      break;      break;
# Line 3756  for (;; ptr++) Line 4426  for (;; ptr++)
4426          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4427        {        {
4428        *code++ = negate_class? OP_ALLANY : OP_FAIL;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
4429        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;        if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4430        zerofirstchar = firstchar;        zerofirstchar = firstchar;
4431          zerofirstcharflags = firstcharflags;
4432        break;        break;
4433        }        }
4434    
# Line 3767  for (;; ptr++) Line 4438  for (;; ptr++)
4438    
4439      should_flip_negation = FALSE;      should_flip_negation = FALSE;
4440    
4441      /* For optimization purposes, we track some properties of the class.      /* For optimization purposes, we track some properties of the class:
4442      class_has_8bitchar will be non-zero, if the class contains at least one      class_has_8bitchar will be non-zero if the class contains at least one <
4443      < 256 character. class_single_char will be 1 if the class contains only      256 character; class_one_char will be 1 if the class contains just one
4444      a single character. */      character. */
4445    
4446      class_has_8bitchar = 0;      class_has_8bitchar = 0;
4447      class_single_char = 0;      class_one_char = 0;
4448    
4449      /* Initialize the 32-char bit map to all zeros. We build the map in a      /* Initialize the 32-char bit map to all zeros. We build the map in a
4450      temporary bit of memory, in case the class contains only 1 character (less      temporary bit of memory, in case the class contains fewer than two
4451      than 256), because in that case the compiled code doesn't use the bit map.      8-bit characters because in that case the compiled code doesn't use the bit
4452      */      map. */
4453    
4454      memset(classbits, 0, 32 * sizeof(pcre_uint8));      memset(classbits, 0, 32 * sizeof(pcre_uint8));
4455    
4456  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4457      xclass = FALSE;                           /* No chars >= 256 */      xclass = FALSE;
4458      class_uchardata = code + LINK_SIZE + 2;   /* For UTF-8 items */      class_uchardata = code + LINK_SIZE + 2;   /* For XCLASS items */
4459      class_uchardata_base = class_uchardata;   /* For resetting in pass 1 */      class_uchardata_base = class_uchardata;   /* Save the start */
4460  #endif  #endif
4461    
4462      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
4463      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
4464      loop, c contains the first byte of the character. */      loop, c contains the first byte of the character. */
4465    
4466      if (c != 0) do      if (c != CHAR_NULL) do
4467        {        {
4468        const pcre_uchar *oldptr;        const pcre_uchar *oldptr;
4469    
# Line 3807  for (;; ptr++) Line 4478  for (;; ptr++)
4478        /* In the pre-compile phase, accumulate the length of any extra        /* In the pre-compile phase, accumulate the length of any extra
4479        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
4480        contain a zillion > 255 characters no longer overwrite the work space        contain a zillion > 255 characters no longer overwrite the work space
4481        (which is on the stack). */        (which is on the stack). We have to remember that there was XCLASS data,
4482          however. */
4483    
4484        if (lengthptr != NULL)        if (lengthptr != NULL && class_uchardata > class_uchardata_base)
4485          {          {
4486            xclass = TRUE;
4487          *lengthptr += class_uchardata - class_uchardata_base;          *lengthptr += class_uchardata - class_uchardata_base;
4488          class_uchardata = class_uchardata_base;          class_uchardata = class_uchardata_base;
4489          }          }
# Line 3912  for (;; ptr++) Line 4585  for (;; ptr++)
4585              for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];              for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];
4586            }            }
4587    
4588          /* Not see if we need to remove any special characters. An option          /* Now see if we need to remove any special characters. An option
4589          value of 1 removes vertical space and 2 removes underscore. */          value of 1 removes vertical space and 2 removes underscore. */
4590    
4591          if (tabopt < 0) tabopt = -tabopt;          if (tabopt < 0) tabopt = -tabopt;
# Line 3928  for (;; ptr++) Line 4601  for (;; ptr++)
4601            for (c = 0; c < 32; c++) classbits[c] |= pbits[c];            for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
4602    
4603          ptr = tempptr + 1;          ptr = tempptr + 1;
4604          /* Every class contains at least one < 256 characters. */          /* Every class contains at least one < 256 character. */
4605          class_has_8bitchar = 1;          class_has_8bitchar = 1;
4606          /* Every class contains at least two characters. */          /* Every class contains at least two characters. */
4607          class_single_char = 2;          class_one_char = 2;
4608          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
4609          }          }
4610    
# Line 3939  for (;; ptr++) Line 4612  for (;; ptr++)
4612        of the specials, which just set a flag. The sequence \b is a special        of the specials, which just set a flag. The sequence \b is a special
4613        case. Inside a class (and only there) it is treated as backspace. We        case. Inside a class (and only there) it is treated as backspace. We
4614        assume that other escapes have more than one character in them, so        assume that other escapes have more than one character in them, so
4615        speculatively set both class_has_8bitchar and class_single_char bigger        speculatively set both class_has_8bitchar and class_one_char bigger
4616        than one. Unrecognized escapes fall through and are either treated        than one. Unrecognized escapes fall through and are either treated
4617        as literal characters (by default), or are faulted if        as literal characters (by default), or are faulted if
4618        PCRE_EXTRA is set. */        PCRE_EXTRA is set. */
4619    
4620        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
4621          {          {
4622          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options,
4623              TRUE);
4624          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
4625            if (escape == 0) c = ec;
4626          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 */
4627          else if (-c == ESC_N)            /* \N is not supported in a class */          else if (escape == ESC_N)          /* \N is not supported in a class */
4628            {            {
4629            *errorcodeptr = ERR71;            *errorcodeptr = ERR71;
4630            goto FAILED;            goto FAILED;
4631            }            }
4632          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (escape == ESC_Q)            /* Handle start of quoted string */
4633            {            {
4634            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
4635              {              {
# Line 3964  for (;; ptr++) Line 4638  for (;; ptr++)
4638            else inescq = TRUE;            else inescq = TRUE;
4639            continue;            continue;
4640            }            }
4641          else if (-c == ESC_E) continue;  /* Ignore orphan \E */          else if (escape == ESC_E) continue;  /* Ignore orphan \E */
4642    
4643          if (c < 0)          else
4644            {            {
4645            register const pcre_uint8 *cbits = cd->cbits;            register const pcre_uint8 *cbits = cd->cbits;
4646            /* Every class contains at least two < 256 characters. */            /* Every class contains at least two < 256 characters. */
4647            class_has_8bitchar++;            class_has_8bitchar++;
4648            /* Every class contains at least two characters. */            /* Every class contains at least two characters. */
4649            class_single_char += 2;            class_one_char += 2;
4650    
4651            switch (-c)            switch (escape)
4652              {              {
4653  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4654              case ESC_du:     /* These are the values given for \d etc */              case ESC_du:     /* These are the values given for \d etc */
# Line 3984  for (;; ptr++) Line 4658  for (;; ptr++)
4658              case ESC_su:     /* of the default ASCII testing. */              case ESC_su:     /* of the default ASCII testing. */
4659              case ESC_SU:              case ESC_SU:
4660              nestptr = ptr;              nestptr = ptr;
4661              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */              ptr = substitutes[escape - ESC_DU] - 1;  /* Just before substitute */
4662              class_has_8bitchar--;                /* Undo! */              class_has_8bitchar--;                /* Undo! */
4663              continue;              continue;
4664  #endif  #endif
# Line 4006  for (;; ptr++) Line 4680  for (;; ptr++)
4680              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
4681              continue;              continue;
4682    
4683              /* 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
4684              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
4685              class. */              previously set by something earlier in the character class.
4686                Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
4687                we could just adjust the appropriate bit. From PCRE 8.34 we no
4688                longer treat \s and \S specially. */
4689    
4690              case ESC_s:              case ESC_s:
4691              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];  
4692              continue;              continue;
4693    
4694              case ESC_S:              case ESC_S:
4695              should_flip_negation = TRUE;              should_flip_negation = TRUE;
4696              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 */  
4697              continue;              continue;
4698    
4699                /* The rest apply in both UCP and non-UCP cases. */
4700    
4701              case ESC_h:              case ESC_h:
4702              SETBIT(classbits, 0x09); /* VT */              (void)add_list_to_class(classbits, &class_uchardata, options, cd,
4703              SETBIT(classbits, 0x20); /* SPACE */                PRIV(hspace_list), NOTACHAR);
             SETBIT(classbits, 0xa0); /* NSBP */  
 #ifndef COMPILE_PCRE8  
             xclass = TRUE;  
             *class_uchardata++ = XCL_SINGLE;  
             *class_uchardata++ = 0x1680;  
             *class_uchardata++ = XCL_SINGLE;  
             *class_uchardata++ = 0x180e;  
             *class_uchardata++ = XCL_RANGE;  
             *class_uchardata++ = 0x2000;  
             *class_uchardata++ = 0x200a;  
             *class_uchardata++ = XCL_SINGLE;  
             *class_uchardata++ = 0x202f;  
             *class_uchardata++ = XCL_SINGLE;  
             *class_uchardata++ = 0x205f;  
             *class_uchardata++ = XCL_SINGLE;  
             *class_uchardata++ = 0x3000;  
 #elif defined SUPPORT_UTF  
             if (utf)  
               {  
               xclass = TRUE;  
               *class_uchardata++ = XCL_SINGLE;  
               class_uchardata += PRIV(ord2utf)(0x1680, class_uchardata);  
               *class_uchardata++ = XCL_SINGLE;  
               class_uchardata += PRIV(ord2utf)(0x180e, class_uchardata);  
               *class_uchardata++ = XCL_RANGE;  
               class_uchardata += PRIV(ord2utf)(0x2000, class_uchardata);  
               class_uchardata += PRIV(ord2utf)(0x200a, class_uchardata);  
               *class_uchardata++ = XCL_SINGLE;  
               class_uchardata += PRIV(ord2utf)(0x202f, class_uchardata);  
               *class_uchardata++ = XCL_SINGLE;  
               class_uchardata += PRIV(ord2utf)(0x205f, class_uchardata);  
               *class_uchardata++ = XCL_SINGLE;  
               class_uchardata += PRIV(ord2utf)(0x3000, class_uchardata);  
               }  
 #endif  
4704              continue;              continue;
4705    
4706              case ESC_H:              case ESC_H:
4707              for (c = 0; c < 32; c++)              (void)add_not_list_to_class(classbits, &class_uchardata, options,
4708                {                cd, PRIV(hspace_list));
               int x = 0xff;  
               switch (c)  
                 {  
                 case 0x09/8: x ^= 1 << (0x09%8); break;  
                 case 0x20/8: x ^= 1 << (0x20%8); break;  
                 case 0xa0/8: x ^= 1 << (0xa0%8); break;  
                 default: break;  
                 }  
               classbits[c] |= x;  
               }  
 #ifndef COMPILE_PCRE8  
             xclass = TRUE;  
             *class_uchardata++ = XCL_RANGE;  
             *class_uchardata++ = 0x0100;  
             *class_uchardata++ = 0x167f;  
             *class_uchardata++ = XCL_RANGE;  
             *class_uchardata++ = 0x1681;  
             *class_uchardata++ = 0x180d;  
             *class_uchardata++ = XCL_RANGE;  
             *class_uchardata++ = 0x180f;  
             *class_uchardata++ = 0x1fff;  
             *class_uchardata++ = XCL_RANGE;  
             *class_uchardata++ = 0x200b;  
             *class_uchardata++ = 0x202e;  
             *class_uchardata++ = XCL_RANGE;  
             *class_uchardata++ = 0x2030;  
             *class_uchardata++ = 0x205e;  
             *class_uchardata++ = XCL_RANGE;  
             *class_uchardata++ = 0x2060;  
             *class_uchardata++ = 0x2fff;  
             *class_uchardata++ = XCL_RANGE;  
             *class_uchardata++ = 0x3001;  
 #ifdef SUPPORT_UTF  
             if (utf)  
               class_uchardata += PRIV(ord2utf)(0x10ffff, class_uchardata);  
             else  
 #endif  
               *class_uchardata++ = 0xffff;  
 #elif defined SUPPORT_UTF  
             if (utf)  
               {  
               xclass = TRUE;  
               *class_uchardata++ = XCL_RANGE;  
               class_uchardata += PRIV(ord2utf)(0x0100, class_uchardata);  
               class_uchardata += PRIV(ord2utf)(0x167f, class_uchardata);  
               *class_uchardata++ = XCL_RANGE;  
               class_uchardata += PRIV(ord2utf)(0x1681, class_uchardata);  
               class_uchardata += PRIV(ord2utf)(0x180d, class_uchardata);  
               *class_uchardata++ = XCL_RANGE;  
               class_uchardata += PRIV(ord2utf)(0x180f, class_uchardata);  
               class_uchardata += PRIV(ord2utf)(0x1fff, class_uchardata);  
               *class_uchardata++ = XCL_RANGE;  
               class_uchardata += PRIV(ord2utf)(0x200b, class_uchardata);  
               class_uchardata += PRIV(ord2utf)(0x202e, class_uchardata);  
               *class_uchardata++ = XCL_RANGE;  
               class_uchardata += PRIV(ord2utf)(0x2030, class_uchardata);  
               class_uchardata += PRIV(ord2utf)(0x205e, class_uchardata);  
               *class_uchardata++ = XCL_RANGE;  
               class_uchardata += PRIV(ord2utf)(0x2060, class_uchardata);  
               class_uchardata += PRIV(ord2utf)(0x2fff, class_uchardata);  
               *class_uchardata++ = XCL_RANGE;  
               class_uchardata += PRIV(ord2utf)(0x3001, class_uchardata);  
               class_uchardata += PRIV(ord2utf)(0x10ffff, class_uchardata);  
               }  
 #endif  
4709              continue;              continue;
4710    
4711              case ESC_v:              case ESC_v:
4712              SETBIT(classbits, 0x0a); /* LF */              (void)add_list_to_class(classbits, &class_uchardata, options, cd,
4713              SETBIT(classbits, 0x0b); /* VT */                PRIV(vspace_list), NOTACHAR);
             SETBIT(classbits, 0x0c); /* FF */  
             SETBIT(classbits, 0x0d); /* CR */  
             SETBIT(classbits, 0x85); /* NEL */  
 #ifndef COMPILE_PCRE8  
             xclass = TRUE;  
             *class_uchardata++ = XCL_RANGE;  
             *class_uchardata++ = 0x2028;  
             *class_uchardata++ = 0x2029;