/[pcre]/code/trunk/pcre_compile.c
ViewVC logotype

Diff of /code/trunk/pcre_compile.c

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 1045 by ph10, Sun Sep 23 16:50:00 2012 UTC revision 1375 by zherczeg, Sat Oct 12 17:56:40 2013 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2012 University of Cambridge             Copyright (c) 1997-2013 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When PCRE_DEBUG is defined, we need the pcre(16)_printint() function, which  /* When PCRE_DEBUG is defined, we need the pcre(16|32)_printint() function, which
57  is also used by pcretest. PCRE_DEBUG is not defined when building a production  is also used by pcretest. PCRE_DEBUG is not defined when building a production
58  library. We do not need to select pcre16_printint.c specially, because the  library. We do not need to select pcre16_printint.c specially, because the
59  COMPILE_PCREx macro will already be appropriately set. */  COMPILE_PCREx macro will already be appropriately set. */
# Line 80  to check them every time. */ Line 80  to check them every time. */
80  /* Definitions to allow mutual recursion */  /* Definitions to allow mutual recursion */
81    
82  static int  static int
83    add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,    add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84      const pcre_uint32 *, unsigned int);      const pcre_uint32 *, unsigned int);
85    
86  static BOOL  static BOOL
87    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL,    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
88      int, int, int *, int *, branch_chain *, compile_data *, int *);      pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
89        compile_data *, int *);
90    
91    
92    
# Line 114  kicks in at the same number of forward r Line 115  kicks in at the same number of forward r
115  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
116  #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)  #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
117    
118    /* This value determines the size of the initial vector that is used for
119    remembering named groups during the pre-compile. It is allocated on the stack,
120    but if it is too small, it is expanded using malloc(), in a similar way to the
121    workspace. The value is the number of slots in the list. */
122    
123    #define NAMED_GROUP_LIST_SIZE  20
124    
125  /* The overrun tests check for a slightly smaller size so that they detect the  /* The overrun tests check for a slightly smaller size so that they detect the
126  overrun before it actually does run off the end of the data block. */  overrun before it actually does run off the end of the data block. */
127    
# Line 121  overrun before it actually does run off Line 129  overrun before it actually does run off
129    
130  /* Private flags added to firstchar and reqchar. */  /* Private flags added to firstchar and reqchar. */
131    
132  #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */  #define REQ_CASELESS    (1 << 0)        /* Indicates caselessness */
133  #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */  #define REQ_VARY        (1 << 1)        /* Reqchar followed non-literal item */
134    /* Negative values for the firstchar and reqchar flags */
135    #define REQ_UNSET       (-2)
136    #define REQ_NONE        (-1)
137    
138  /* Repeated character flags. */  /* Repeated character flags. */
139    
# Line 451  static const char error_texts[] = Line 462  static const char error_texts[] =
462    "POSIX collating elements are not supported\0"    "POSIX collating elements are not supported\0"
463    "this version of PCRE is compiled without UTF support\0"    "this version of PCRE is compiled without UTF support\0"
464    "spare error\0"  /** DEAD **/    "spare error\0"  /** DEAD **/
465    "character value in \\x{...} sequence is too large\0"    "character value in \\x{} or \\o{} is too large\0"
466    /* 35 */    /* 35 */
467    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
468    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
# Line 483  static const char error_texts[] = Line 494  static const char error_texts[] =
494    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
495    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
496    /* 60 */    /* 60 */
497    "(*VERB) not recognized\0"    "(*VERB) not recognized or malformed\0"
498    "number is too big\0"    "number is too big\0"
499    "subpattern name expected\0"    "subpattern name expected\0"
500    "digit expected after (?+\0"    "digit expected after (?+\0"
# Line 503  static const char error_texts[] = Line 514  static const char error_texts[] =
514    /* 75 */    /* 75 */
515    "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"    "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
516    "character value in \\u.... sequence is too large\0"    "character value in \\u.... sequence is too large\0"
517      "invalid UTF-32 string\0"
518      "setting UTF is disabled by the application\0"
519      "non-hex character in \\x{} (closing brace missing?)\0"
520      /* 80 */
521      "non-octal character in \\o{} (closing brace missing?)\0"
522      "missing opening brace after \\o\0"
523    ;    ;
524    
525  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 642  static const pcre_uint8 ebcdic_chartab[] Line 659  static const pcre_uint8 ebcdic_chartab[]
659  #endif  #endif
660    
661    
662    /* This table is used to check whether auto-possessification is possible
663    between adjacent character-type opcodes. The left-hand (repeated) opcode is
664    used to select the row, and the right-hand opcode is use to select the column.
665    A value of 1 means that auto-possessification is OK. For example, the second
666    value in the first row means that \D+\d can be turned into \D++\d.
667    
668    The Unicode property types (\P and \p) have to be present to fill out the table
669    because of what their opcode values are, but the table values should always be
670    zero because property types are handled separately in the code. The last four
671    columns apply to items that cannot be repeated, so there is no need to have
672    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
673    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
674    
675    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
676    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
677    
678    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
679    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
680      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
681      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
682      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
683      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
684      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
685      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
686      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
687      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
688      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
689      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
690      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
691      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
692      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
693      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
694      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
695      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
696      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
697    };
698    
699    
700    /* This table is used to check whether auto-possessification is possible
701    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
702    left-hand (repeated) opcode is used to select the row, and the right-hand
703    opcode is used to select the column. The values are as follows:
704    
705      0   Always return FALSE (never auto-possessify)
706      1   Character groups are distinct (possessify if both are OP_PROP)
707      2   Check character categories in the same group (general or particular)
708      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
709    
710      4   Check left general category vs right particular category
711      5   Check right general category vs left particular category
712    
713      6   Left alphanum vs right general category
714      7   Left space vs right general category
715      8   Left word vs right general category
716    
717      9   Right alphanum vs left general category
718     10   Right space vs left general category
719     11   Right word vs left general category
720    
721     12   Left alphanum vs right particular category
722     13   Left space vs right particular category
723     14   Left word vs right particular category
724    
725     15   Right alphanum vs left particular category
726     16   Right space vs left particular category
727     17   Right word vs left particular category
728    */
729    
730    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
731    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
732      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
733      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
734      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
735      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
736      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
737      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
738      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
739      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
740      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
741      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
742      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
743    };
744    
745    /* This table is used to check whether auto-possessification is possible
746    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
747    specifies a general category and the other specifies a particular category. The
748    row is selected by the general category and the column by the particular
749    category. The value is 1 if the particular category is not part of the general
750    category. */
751    
752    static const pcre_uint8 catposstab[7][30] = {
753    /* Cc Cf Cn Co Cs Ll Lm Lo Lt Lu Mc Me Mn Nd Nl No Pc Pd Pe Pf Pi Po Ps Sc Sk Sm So Zl Zp Zs */
754      { 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* C */
755      { 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* L */
756      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* M */
757      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* N */
758      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1 },  /* P */
759      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1 },  /* S */
760      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0 }   /* Z */
761    };
762    
763    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
764    a general or particular category. The properties in each row are those
765    that apply to the character set in question. Duplication means that a little
766    unnecessary work is done when checking, but this keeps things much simpler
767    because they can all use the same code. For more details see the comment where
768    this table is used.
769    
770    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
771    "space", but from Perl 5.18 it's included, so both categories are treated the
772    same here. */
773    
774    static const pcre_uint8 posspropstab[3][4] = {
775      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
776      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
777      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
778    };
779    
780    
781    
782  /*************************************************  /*************************************************
# Line 663  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 747  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 781  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 790  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 < CHAR_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 803  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 828  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 838  else Line 978  else
978  #endif  #endif
979            }            }
980    
981  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
982          if (c > (utf ? 0x10ffff : 0xff))          if (c > (utf ? 0x10ffffU : 0xffU))
983  #else  #elif defined COMPILE_PCRE16
984  #ifdef COMPILE_PCRE16          if (c > (utf ? 0x10ffffU : 0xffffU))
985          if (c > (utf ? 0x10ffff : 0xffff))  #elif defined COMPILE_PCRE32
986  #endif          if (utf && c > 0x10ffffU)
987  #endif  #endif
988            {            {
989            *errorcodeptr = ERR76;            *errorcodeptr = ERR76;
# Line 875  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 890  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 910  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 934  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 942  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 972  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    
1151      if ((c = *ptr) >= CHAR_8)      if ((c = *ptr) >= CHAR_8) break;
1152        {  
1153        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1154    
1155      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
1156      larger first octal digit. The original code used just to take the least      larger first octal digit. The original code used just to take the least
# Line 1023  else Line 1167  else
1167  #endif  #endif
1168      break;      break;
1169    
1170      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \o is a relatively new Perl feature, supporting a more general way of
1171      than 0xff in utf or non-8bit mode, but only if the ddd are hex digits.      specifying character codes in octal. The only supported form is \o{ddd}. */
1172      If not, { is treated as a data character. */  
1173        case CHAR_o:
1174        if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1175          {
1176          ptr += 2;
1177          c = 0;
1178          overflow = FALSE;
1179          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1180            {
1181            register pcre_uint32 cc = *ptr++;
1182            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1183    #ifdef COMPILE_PCRE32
1184            if (c >= 0x20000000l) { overflow = TRUE; break; }
1185    #endif
1186            c = (c << 3) + cc - CHAR_0 ;
1187    #if defined COMPILE_PCRE8
1188            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1189    #elif defined COMPILE_PCRE16
1190            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1191    #elif defined COMPILE_PCRE32
1192            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1193    #endif
1194            }
1195          if (overflow)
1196            {
1197            while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1198            *errorcodeptr = ERR34;
1199            }
1200          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1201            {
1202            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1203            }
1204          else *errorcodeptr = ERR80;
1205          }
1206        break;
1207    
1208        /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1209        numbers. Otherwise it is a lowercase x letter. */
1210    
1211      case CHAR_x:      case CHAR_x:
1212      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1213        {        {
       /* In JavaScript, \x must be followed by two hexadecimal numbers.  
       Otherwise it is a lowercase x letter. */  
1214        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1215          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1216          {          {
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 1048  else Line 1227  else
1227  #endif  #endif
1228            }            }
1229          }          }
1230        break;        }    /* End JavaScript handling */
       }  
1231    
1232      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1233        {      greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1234        const pcre_uchar *pt = ptr + 2;      digits. If not, { used to be treated as a data character. However, Perl
1235        seems to read hex digits up to the first non-such, and ignore the rest, so
1236        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1237        now gives an error. */
1238    
1239        c = 0;      else
1240        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)        {
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));
1261  #endif  #endif
1262    
1263  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
1264          if (c > (utf ? 0x10ffff : 0xff)) { c = -1; break; }            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1265  #else  #elif defined COMPILE_PCRE16
1266  #ifdef COMPILE_PCRE16            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1267          if (c > (utf ? 0x10ffff : 0xffff)) { c = -1; break; }  #elif defined COMPILE_PCRE32
1268  #endif            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1269  #endif  #endif
1270          }            }
1271    
1272        if (c < 0)          if (overflow)
1273          {            {
1274          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1275          *errorcodeptr = ERR34;            *errorcodeptr = ERR34;
1276          }            }
1277    
1278        if (*pt == CHAR_RIGHT_CURLY_BRACKET)          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1279          {            {
1280          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1281          ptr = pt;            }
         break;  
         }  
1282    
1283        /* If the sequence of hex digits does not end with '}', then we don't          /* If the sequence of hex digits does not end with '}', give an error.
1284        recognize this construct; fall through to the normal \x handling. */          We used just to recognize this construct and fall through to the normal
1285        }          \x handling, but nowadays Perl gives an error, which seems much more
1286            sensible, so we do too. */
1287    
1288      /* Read just a single-byte hex-defined char */          else *errorcodeptr = ERR79;
1289            }   /* End of \x{} processing */
1290    
1291      c = 0;        /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1292      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)  
1293        {        else
1294        int cc;                                  /* Some compilers don't like */          {
1295        cc = *(++ptr);                           /* ++ in initializers */          c = 0;
1296            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1297              {
1298              pcre_uint32 cc;                          /* Some compilers don't like */
1299              cc = *(++ptr);                           /* ++ in initializers */
1300  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1301        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1302        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1303  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1304        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */            if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1305        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1306  #endif  #endif
1307        }            }
1308            }     /* End of \xdd handling */
1309          }       /* End of Perl-style \x handling */
1310      break;      break;
1311    
1312      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
# Line 1119  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 1159  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 1189  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 1220  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 1245  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 1297  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 1312  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 1337  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 1670  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 1686  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 1704  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 1730  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 1810  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 1850  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 1859  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 1894  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 1917  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 1984  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 2020  for (;;) Line 1933  for (;;)
1933    
1934    
1935    
   
1936  /*************************************************  /*************************************************
1937  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
1938  *************************************************/  *************************************************/
# Line 2033  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 2044  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 2067  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 2097  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 2120  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 2172  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 2182  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 2217  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 2240  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:
# Line 2326  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 2364  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      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)      else
2303        if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;        {
2304          recurse_check *r = recurses;
2305          const pcre_uchar *endgroup = scode;
2306    
2307      /* Not a forward reference, test for completed backward reference */        do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2308          if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2309    
2310      empty_branch = FALSE;        for (r = recurses; r != NULL; r = r->prev)
2311      scode = cd->start_code + GET(code, 1);          if (r->group == scode) break;
2312      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */        if (r != NULL) continue;   /* Mutual recursion */
2313          }
2314    
2315      /* Completed backwards reference */      /* Completed reference; scan the referenced group, remembering it on the
2316        stack chain to detect mutual recursions. */
2317    
2318        empty_branch = FALSE;
2319        this_recurse.prev = recurses;
2320        this_recurse.group = scode;
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 2438  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 2496  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_EXTUNI:      case OP_ANYNL:
2445      case OP_NOT_DIGIT:  
2446        case OP_NOT_HSPACE:
2447        case OP_HSPACE:
2448        case OP_NOT_VSPACE:
2449        case OP_VSPACE:
2450        case OP_EXTUNI:
2451    
2452        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 2543  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 2557  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_NOTUPTO:
2562        case OP_NOTUPTOI:
2563    
2564      case OP_MINUPTO:      case OP_MINUPTO:
2565      case OP_MINUPTOI:      case OP_MINUPTOI:
2566        case OP_NOTMINUPTO:
2567        case OP_NOTMINUPTOI:
2568    
2569      case OP_POSUPTO:      case OP_POSUPTO:
2570      case OP_POSUPTOI:      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]);      if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2575      break;      break;
2576  #endif  #endif
# Line 2591  for (code = first_significant_code(code Line 2581  for (code = first_significant_code(code
2581      case OP_MARK:      case OP_MARK:
2582      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
2583      case OP_SKIP_ARG:      case OP_SKIP_ARG:
     code += code[1];  
     break;  
   
2584      case OP_THEN_ARG:      case OP_THEN_ARG:
2585      code += code[1];      code += code[1];
2586      break;      break;
# Line 2625  Arguments: Line 2612  Arguments:
2612    code        points to start of the recursion    code        points to start of the recursion
2613    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2614    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2615    utf         TRUE if in UTF-8 / UTF-16 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2616    cd          pointers to tables etc    cd          pointers to tables etc
2617    
2618  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
# Line 2637  could_be_empty(const pcre_uchar *code, c Line 2624  could_be_empty(const pcre_uchar *code, c
2624  {  {
2625  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2626    {    {
2627    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2628      return FALSE;      return FALSE;
2629    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2630    }    }
# Line 2647  return TRUE; Line 2634  return TRUE;
2634    
2635    
2636  /*************************************************  /*************************************************
2637  *           Check for POSIX class syntax         *  *        Base opcode of repeated opcodes         *
2638  *************************************************/  *************************************************/
2639    
2640  /* This function is called when the sequence "[:" or "[." or "[=" is  /* Returns the base opcode for repeated single character type opcodes. If the
2641  encountered in a character class. It checks whether this is followed by a  opcode is not a repeated character type, it returns with the original value.
 sequence of characters terminated by a matching ":]" or ".]" or "=]". If we  
 reach an unescaped ']' without the special preceding character, return FALSE.  
   
 Originally, this function only recognized a sequence of letters between the  
 terminators, but it seems that Perl recognizes any sequence of characters,  
 though of course unknown POSIX names are subsequently rejected. Perl gives an  
 "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE  
 didn't consider this to be a POSIX class. Likewise for [:1234:].  
   
 The problem in trying to be exactly like Perl is in the handling of escapes. We  
 have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX  
 class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code  
 below handles the special case of \], but does not try to do any other escape  
 processing. This makes it different from Perl for cases such as [:l\ower:]  
 where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize  
 "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  
 I think.  
   
 A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.  
 It seems that the appearance of a nested POSIX class supersedes an apparent  
 external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or  
 a digit.  
   
 In Perl, unescaped square brackets may also appear as part of class names. For  
 example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for  
 [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not  
 seem right at all. PCRE does not allow closing square brackets in POSIX class  
 names.  
   
 Arguments:  
   ptr      pointer to the initial [  
   endptr   where to return the end pointer  
2642    
2643  Returns:   TRUE or FALSE  Arguments:  c opcode
2644    Returns:    base opcode for the type
2645  */  */
2646    
2647  static BOOL  static pcre_uchar
2648  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)  get_repeat_base(pcre_uchar c)
2649  {  {
2650  int terminator;          /* Don't combine these lines; the Solaris cc */  return (c > OP_TYPEPOSUPTO)? c :
2651  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */         (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2652  for (++ptr; *ptr != 0; ptr++)         (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2653    {         (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2654    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)         (c >= OP_STARI)?      OP_STARI :
2655      ptr++;                               OP_STAR;
   else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
   else  
     {  
     if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)  
       {  
       *endptr = ptr;  
       return TRUE;  
       }  
     if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&  
          (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||  
           ptr[1] == CHAR_EQUALS_SIGN) &&  
         check_posix_syntax(ptr, endptr))  
       return FALSE;  
     }  
   }  
 return FALSE;  
2656  }  }
2657    
2658    
2659    
2660    #ifdef SUPPORT_UCP
2661  /*************************************************  /*************************************************
2662  *          Check POSIX class name                *  *        Check a character and a property        *
2663  *************************************************/  *************************************************/
2664    
2665  /* This function is called to check the name given in a POSIX-style class entry  /* This function is called by check_auto_possessive() when a property item
2666  such as [:alnum:].  is adjacent to a fixed character.
2667    
2668  Arguments:  Arguments:
2669    ptr        points to the first letter    c            the character
2670    len        the length of the name    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:     a value representing the name, or -1 if unknown  Returns:       TRUE if auto-possessifying is OK
2675  */  */
2676    
2677  static int  static BOOL
2678  check_posix_name(const pcre_uchar *ptr, int len)  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2679      BOOL negated)
2680  {  {
2681  const char *pn = posix_names;  const pcre_uint32 *p;
2682  register int yield = 0;  const ucd_record *prop = GET_UCD(c);
 while (posix_name_lengths[yield] != 0)  
   {  
   if (len == posix_name_lengths[yield] &&  
     STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;  
   pn += posix_name_lengths[yield] + 1;  
   yield++;  
   }  
 return -1;  
 }  
   
   
 /*************************************************  
 *    Adjust OP_RECURSE items in repeated group   *  
 *************************************************/  
2683    
2684  /* OP_RECURSE items contain an offset from the start of the regex to the group  switch(ptype)
2685  that is referenced. This means that groups can be replicated for fixed    {
2686  repetition simply by copying (because the recursion is allowed to refer to    case PT_LAMP:
2687  earlier groups that are outside the current group). However, when a group is    return (prop->chartype == ucp_Lu ||
2688  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is            prop->chartype == ucp_Ll ||
2689  inserted before it, after it has been compiled. This means that any OP_RECURSE            prop->chartype == ucp_Lt) == negated;
 items within it that refer to the group itself or any contained groups have to  
 have their offsets adjusted. That one of the jobs of this function. Before it  
 is called, the partially compiled regex must be temporarily terminated with  
 OP_END.  
2690    
2691  This function has been extended with the possibility of forward references for    case PT_GC:
2692  recursions and subroutine calls. It must also check the list of such references    return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
 for the group we are dealing with. If it finds that one of the recursions in  
 the current group is on this list, it adjusts the offset in the list, not the  
 value in the reference (which is a group number).  
2693    
2694  Arguments:    case PT_PC:
2695    group      points to the start of the group    return (pdata == prop->chartype) == negated;
   adjust     the amount by which the group is to be moved  
   utf        TRUE in UTF-8 / UTF-16 mode  
   cd         contains pointers to tables etc.  
   save_hwm   the hwm forward reference pointer at the start of the group  
2696    
2697  Returns:     nothing    case PT_SC:
2698  */    return (pdata == prop->script) == negated;
2699    
2700  static void    /* These are specials */
 adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,  
   pcre_uchar *save_hwm)  
 {  
 pcre_uchar *ptr = group;  
2701    
2702  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)    case PT_ALNUM:
2703    {    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2704    int offset;            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
   pcre_uchar *hc;  
2705    
2706    /* See if this recursion is on the forward reference list. If so, adjust the    /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2707    reference. */    means that Perl space and POSIX space are now identical. PCRE was changed
2708      at release 8.34. */
2709    
2710    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    case PT_SPACE:    /* Perl space */
2711      {    case PT_PXSPACE:  /* POSIX space */
2712      offset = GET(hc, 0);    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2713      if (cd->start_code + offset == ptr + 1)            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2714        {            c == CHAR_FF || c == CHAR_CR)
2715        PUT(hc, 0, offset + adjust);            == negated;
       break;  
       }  
     }  
2716    
2717    /* Otherwise, adjust the recursion offset if it's after the start of this    case PT_WORD:
2718    group. */    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2719              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2720              c == CHAR_UNDERSCORE) == negated;
2721    
2722    if (hc >= cd->hwm)    case PT_CLIST:
2723      p = PRIV(ucd_caseless_sets) + prop->caseset;
2724      for (;;)
2725      {      {
2726      offset = GET(ptr, 1);      if (c < *p) return !negated;
2727      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      if (c == *p++) return negated;
2728      }      }
2729      break;  /* Control never reaches here */
   ptr += 1 + LINK_SIZE;  
2730    }    }
2731    
2732    return FALSE;
2733  }  }
2734    #endif  /* SUPPORT_UCP */
2735    
2736    
2737    
2738  /*************************************************  /*************************************************
2739  *        Insert an automatic callout point       *  *        Fill the character property list        *
2740  *************************************************/  *************************************************/
2741    
2742  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert  /* Checks whether the code points to an opcode that can take part in auto-
2743  callout points before each pattern item.  possessification, and if so, fills a list with its properties.
2744    
2745  Arguments:  Arguments:
2746    code           current code pointer    code        points to start of expression
2747    ptr            current pattern pointer    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2748    cd             pointers to tables etc    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:         new code pointer  Returns:      points to the start of the next opcode if *code is accepted
2756                  NULL if *code is not accepted
2757  */  */
2758    
2759  static pcre_uchar *  static const pcre_uchar *
2760  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)  get_chr_property_list(const pcre_uchar *code, BOOL utf,
2761      const pcre_uint8 *fcc, pcre_uint32 *list)
2762  {  {
2763  *code++ = OP_CALLOUT;  pcre_uchar c = *code;
2764  *code++ = 255;  const pcre_uchar *end;
2765  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */  const pcre_uint32 *clist_src;
2766  PUT(code, LINK_SIZE, 0);                       /* Default length */  pcre_uint32 *clist_dest;
2767  return code + 2 * LINK_SIZE;  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  *         Complete a callout item                *      code += IMM2_SIZE;
 *************************************************/  
2781    
2782  /* A callout item contains the length of the next item in the pattern, which    list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
 we can't fill in till after we have reached the relevant point. This is used  
 for both automatic and manual callouts.  
2783    
2784  Arguments:    switch(base)
2785    previous_callout   points to previous callout item      {
2786    ptr                current pattern pointer      case OP_STAR:
2787    cd                 pointers to tables etc      list[0] = OP_CHAR;
2788        break;
2789    
2790  Returns:             nothing      case OP_STARI:
2791  */      list[0] = OP_CHARI;
2792        break;
2793    
2794  static void      case OP_NOTSTAR:
2795  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)      list[0] = OP_NOT;
2796  {      break;
 int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));  
 PUT(previous_callout, 2 + LINK_SIZE, length);  
 }  
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  #ifdef SUPPORT_UCP  switch(c)
2811  /*************************************************    {
2812  *           Get othercase range                  *    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  /* This function is passed the start and end of a class range, in UTF-8 mode    case OP_CHAR:
2833  with UCP support. It searches up the characters, looking for ranges of    case OP_NOT:
2834  characters in the "other" case. Each call returns the next one, updating the    GETCHARINCTEST(chr, code);
2835  start address. A character with multiple other cases is returned on its own    list[2] = chr;
2836  with a special return value.    list[3] = NOTACHAR;
2837      return code;
2838    
2839  Arguments:    case OP_CHARI:
2840    cptr        points to starting character value; updated    case OP_NOTI:
2841    d           end value    list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2842    ocptr       where to put start of othercase range    GETCHARINCTEST(chr, code);
2843    odptr       where to put end of othercase range    list[2] = chr;
2844    
2845  Yield:        -1 when no more  #ifdef SUPPORT_UCP
2846                 0 when a range is returned    if (chr < 128 || (chr < 256 && !utf))
2847                >0 the CASESET offset for char with multiple other cases      list[3] = fcc[chr];
2848                  in this case, ocptr contains the original    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  static int    /* The othercase might be the same value. */
 get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,  
   unsigned int *odptr)  
 {  
 unsigned int c, othercase, next;  
 int co;  
2857    
2858  /* Find the first character that has an other case. If it has multiple other    if (chr == list[3])
2859  cases, return its case offset value. */      list[3] = NOTACHAR;
2860      else
2861        list[4] = NOTACHAR;
2862      return code;
2863    
2864  for (c = *cptr; c <= d; c++)  #ifdef SUPPORT_UCP
2865    {    case OP_PROP:
2866    if ((co = UCD_CASESET(c)) != 0)    case OP_NOTPROP:
2867      if (code[0] != PT_CLIST)
2868      {      {
2869      *ocptr = c++;   /* Character that has the set */      list[2] = code[0];
2870      *cptr = c;      /* Rest of input range */      list[3] = code[1];
2871      return co;      return code + 2;
2872      }      }
   if ((othercase = UCD_OTHERCASE(c)) != c) break;  
   }  
2873    
2874  if (c > d) return -1;  /* Reached end of range */    /* Convert only if we have enough space. */
2875    
2876  *ocptr = othercase;    clist_src = PRIV(ucd_caseless_sets) + code[1];
2877  next = othercase + 1;    clist_dest = list + 2;
2878      code += 2;
2879    
2880  for (++c; c <= d; c++)    do {
2881    {       if (clist_dest >= list + 8)
2882    if (UCD_OTHERCASE(c) != next) break;         {
2883    next++;         /* Early return if there is not enough space. This should never
2884    }         happen, since all clists are shorter than 5 character now. */
2885           list[2] = code[0];
2886           list[3] = code[1];
2887           return code;
2888           }
2889         *clist_dest++ = *clist_src;
2890         }
2891      while(*clist_src++ != NOTACHAR);
2892    
2893  *odptr = next - 1;     /* End of othercase range */    /* All characters are stored. The terminating NOTACHAR
2894  *cptr = c;             /* Rest of input range */    is copied form the clist itself. */
2895  return 0;  
2896      list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
2897      return code;
2898    #endif
2899    
2900      case OP_NCLASS:
2901      case OP_CLASS:
2902    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2903      case OP_XCLASS:
2904    
2905      if (c == OP_XCLASS)
2906        end = code + GET(code, 0);
2907      else
2908    #endif
2909        end = code + 32 / sizeof(pcre_uchar);
2910    
2911      switch(*end)
2912        {
2913        case OP_CRSTAR:
2914        case OP_CRMINSTAR:
2915        case OP_CRQUERY:
2916        case OP_CRMINQUERY:
2917        list[1] = TRUE;
2918        end++;
2919        break;
2920    
2921        case OP_CRRANGE:
2922        case OP_CRMINRANGE:
2923        list[1] = (GET2(end, 1) == 0);
2924        end += 1 + 2 * IMM2_SIZE;
2925        break;
2926        }
2927      list[2] = end - code;
2928      return end;
2929      }
2930    return NULL;    /* Opcode not accepted */
2931  }  }
2932    
2933    
2934    
2935  /*************************************************  /*************************************************
2936  *        Check a character and a property        *  *    Scan further character sets for match       *
2937  *************************************************/  *************************************************/
2938    
2939  /* This function is called by check_auto_possessive() when a property item  /* Checks whether the base and the current opcode have a common character, in
2940  is adjacent to a fixed character.  which case the base cannot be possessified.
2941    
2942  Arguments:  Arguments:
2943    c            the character    code        points to the byte code
2944    ptype        the property type    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2945    pdata        the data for the type    cd          static compile data
2946    negated      TRUE if it's a negated property (\P or \p{^)    base_list   the data list of the base opcode
2947    
2948  Returns:       TRUE if auto-possessifying is OK  Returns:      TRUE if the auto-possessification is possible
2949  */  */
2950    
2951  static BOOL  static BOOL
2952  check_char_prop(int c, int ptype, int pdata, BOOL negated)  compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
2953      const pcre_uint32* base_list)
2954  {  {
2955  const ucd_record *prop = GET_UCD(c);  pcre_uchar c;
2956  switch(ptype)  pcre_uint32 list[8];
2957    const pcre_uint32* chr_ptr;
2958    const pcre_uint32* ochr_ptr;
2959    const pcre_uint32* list_ptr;
2960    const pcre_uchar *next_code;
2961    pcre_uint32 chr;
2962    
2963    /* Note: the base_list[1] contains whether the current opcode has greedy
2964    (represented by a non-zero value) quantifier. This is a different from
2965    other character type lists, which stores here that the character iterator
2966    matches to an empty string (also represented by a non-zero value). */
2967    
2968    for(;;)
2969    {    {
2970    case PT_LAMP:    c = *code;
   return (prop->chartype == ucp_Lu ||  
           prop->chartype == ucp_Ll ||  
           prop->chartype == ucp_Lt) == negated;  
2971    
2972    case PT_GC:    /* Skip over callouts */
   return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;  
2973    
2974    case PT_PC:    if (c == OP_CALLOUT)
2975    return (pdata == prop->chartype) == negated;      {
2976        code += PRIV(OP_lengths)[c];
2977        continue;
2978        }
2979    
2980    case PT_SC:    if (c == OP_ALT)
2981    return (pdata == prop->script) == negated;      {
2982        do code += GET(code, 1); while (*code == OP_ALT);
2983        c = *code;
2984        }
2985    
2986    /* These are specials */    switch(c)
2987        {
2988        case OP_END:
2989        case OP_KETRPOS:
2990        /* TRUE only in greedy case. The non-greedy case could be replaced by
2991        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
2992        uses more memory, which we cannot get at this stage.) */
2993    
2994    case PT_ALNUM:      return base_list[1] != 0;
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||  
           PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;  
2995    
2996    case PT_SPACE:    /* Perl space */      case OP_KET:
2997    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||      /* If the bracket is capturing, and referenced by an OP_RECURSE, or
2998            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)      it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
2999            == negated;      cannot be converted to a possessive form. */
3000    
3001        if (base_list[1] == 0) return FALSE;
3002    
3003        switch(*(code - GET(code, 1)))
3004          {
3005          case OP_ASSERT:
3006          case OP_ASSERT_NOT:
3007          case OP_ASSERTBACK:
3008          case OP_ASSERTBACK_NOT:
3009          case OP_ONCE:
3010          case OP_ONCE_NC:
3011          /* Atomic sub-patterns and assertions can always auto-possessify their
3012          last iterator. */
3013          return TRUE;
3014          }
3015    
3016    case PT_PXSPACE:  /* POSIX space */      code += PRIV(OP_lengths)[c];
3017    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||      continue;
           c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||  
           c == CHAR_FF || c == CHAR_CR)  
           == negated;  
3018    
3019    case PT_WORD:      case OP_ONCE:
3020    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||      case OP_ONCE_NC:
3021            PRIV(ucp_gentype)[prop->chartype] == ucp_N ||      case OP_BRA:
3022            c == CHAR_UNDERSCORE) == negated;      case OP_CBRA:
3023    }      next_code = code;
3024  return FALSE;      do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
 }  
 #endif  /* SUPPORT_UCP */  
3025    
3026        /* We do not support repeated brackets, because they can lead to
3027        infinite recursion. */
3028    
3029        if (*next_code != OP_KET) return FALSE;
3030    
3031  /*************************************************      next_code = code + GET(code, 1);
3032  *     Check if auto-possessifying is possible    *      code += PRIV(OP_lengths)[c];
 *************************************************/  
3033    
3034  /* This function is called for unlimited repeats of certain items, to see      while (*next_code == OP_ALT)
3035  whether the next thing could possibly match the repeated item. If not, it makes        {
3036  sense to automatically possessify the repeated item.        if (!compare_opcodes(code, utf, cd, base_list)) return FALSE;
3037          code = next_code + 1 + LINK_SIZE;
3038          next_code += GET(next_code, 1);
3039          }
3040        continue;
3041    
3042  Arguments:      case OP_BRAZERO:
3043    previous      pointer to the repeated opcode      case OP_BRAMINZERO:
   utf           TRUE in UTF-8 / UTF-16 mode  
   ptr           next character in pattern  
   options       options bits  
   cd            contains pointers to tables etc.  
3044    
3045  Returns:        TRUE if possessifying is wanted      next_code = code + 1;
3046  */      if (*next_code != OP_BRA && *next_code != OP_CBRA)
3047          return FALSE;
3048    
3049  static BOOL      do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
 check_auto_possessive(const pcre_uchar *previous, BOOL utf,  
   const pcre_uchar *ptr, int options, compile_data *cd)  
 {  
 pcre_int32 c, next;  
 int op_code = *previous++;  
3050    
3051  /* Skip whitespace and comments in extended mode */      /* We do not support repeated brackets, because they can lead to
3052        infinite recursion. */
3053        if (*next_code != OP_KET) return FALSE;
3054    
3055        /* The bracket content will be checked by the
3056        OP_BRA/OP_CBRA case above. */
3057        next_code += 1 + LINK_SIZE;
3058        if (!compare_opcodes(next_code, utf, cd, base_list)) return FALSE;
3059    
3060  if ((options & PCRE_EXTENDED) != 0)      code += PRIV(OP_lengths)[c];
3061    {      continue;
3062    for (;;)      }
3063    
3064      /* Check for a supported opcode, and load its properties. */
3065    
3066      code = get_chr_property_list(code, utf, cd->fcc, list);
3067      if (code == NULL) return FALSE;    /* Unsupported */
3068    
3069      /* If either opcode is a small character list, set pointers for comparing
3070      characters from that list with another list, or with a property. */
3071    
3072      if (base_list[0] == OP_CHAR)
3073        {
3074        chr_ptr = base_list + 2;
3075        list_ptr = list;
3076        }
3077      else if (list[0] == OP_CHAR)
3078        {
3079        chr_ptr = list + 2;
3080        list_ptr = base_list;
3081        }
3082    
3083      /* Some property combinations also acceptable. Unicode property opcodes are
3084      processed specially; the rest can be handled with a lookup table. */
3085    
3086      else
3087      {      {
3088      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      pcre_uint32 leftop, rightop;
3089      if (*ptr == CHAR_NUMBER_SIGN)  
3090        if (list[1] != 0) return FALSE;   /* Must match at least one character */
3091        leftop = base_list[0];
3092        rightop = list[0];
3093    
3094    #ifdef SUPPORT_UCP
3095        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3096        {        {
3097        ptr++;        if (rightop == OP_EOD) return TRUE;
3098        while (*ptr != 0)        if (rightop == OP_PROP || rightop == OP_NOTPROP)
3099          {          {
3100          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          int n;
3101          ptr++;          const pcre_uint8 *p;
3102  #ifdef SUPPORT_UTF          BOOL same = leftop == rightop;
3103          if (utf) FORWARDCHAR(ptr);          BOOL lisprop = leftop == OP_PROP;
3104  #endif          BOOL risprop = rightop == OP_PROP;
3105            BOOL bothprop = lisprop && risprop;
3106    
3107            /* There's a table that specifies how each combination is to be
3108            processed:
3109              0   Always return FALSE (never auto-possessify)
3110              1   Character groups are distinct (possessify if both are OP_PROP)
3111              2   Check character categories in the same group (general or particular)
3112              3   Return TRUE if the two opcodes are not the same
3113              ... see comments below
3114            */
3115    
3116            n = propposstab[base_list[2]][list[2]];
3117            switch(n)
3118              {
3119              case 0: return FALSE;
3120              case 1: return bothprop;
3121              case 2: return (base_list[3] == list[3]) != same;
3122              case 3: return !same;
3123    
3124              case 4:  /* Left general category, right particular category */
3125              return risprop && catposstab[base_list[3]][list[3]] == same;
3126    
3127              case 5:  /* Right general category, left particular category */
3128              return lisprop && catposstab[list[3]][base_list[3]] == same;
3129    
3130              /* This code is logically tricky. Think hard before fiddling with it.
3131              The posspropstab table has four entries per row. Each row relates to
3132              one of PCRE's special properties such as ALNUM or SPACE or WORD.
3133              Only WORD actually needs all four entries, but using repeats for the
3134              others means they can all use the same code below.
3135    
3136              The first two entries in each row are Unicode general categories, and
3137              apply always, because all the characters they include are part of the
3138              PCRE character set. The third and fourth entries are a general and a
3139              particular category, respectively, that include one or more relevant
3140              characters. One or the other is used, depending on whether the check
3141              is for a general or a particular category. However, in both cases the
3142              category contains more characters than the specials that are defined
3143              for the property being tested against. Therefore, it cannot be used
3144              in a NOTPROP case.
3145    
3146              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3147              Underscore is covered by ucp_P or ucp_Po. */
3148    
3149              case 6:  /* Left alphanum vs right general category */
3150              case 7:  /* Left space vs right general category */
3151              case 8:  /* Left word vs right general category */
3152              p = posspropstab[n-6];
3153              return risprop && lisprop ==
3154                (list[3] != p[0] &&
3155                 list[3] != p[1] &&
3156                (list[3] != p[2] || !lisprop));
3157    
3158              case 9:   /* Right alphanum vs left general category */
3159              case 10:  /* Right space vs left general category */
3160              case 11:  /* Right word vs left general category */
3161              p = posspropstab[n-9];
3162              return lisprop && risprop ==
3163                (base_list[3] != p[0] &&
3164                 base_list[3] != p[1] &&
3165                (base_list[3] != p[2] || !risprop));
3166    
3167              case 12:  /* Left alphanum vs right particular category */
3168              case 13:  /* Left space vs right particular category */
3169              case 14:  /* Left word vs right particular category */
3170              p = posspropstab[n-12];
3171              return risprop && lisprop ==
3172                (catposstab[p[0]][list[3]] &&
3173                 catposstab[p[1]][list[3]] &&
3174                (list[3] != p[3] || !lisprop));
3175    
3176              case 15:  /* Right alphanum vs left particular category */
3177              case 16:  /* Right space vs left particular category */
3178              case 17:  /* Right word vs left particular category */
3179              p = posspropstab[n-15];
3180              return lisprop && risprop ==
3181                (catposstab[p[0]][base_list[3]] &&
3182                 catposstab[p[1]][base_list[3]] &&
3183                (base_list[3] != p[3] || !risprop));
3184              }
3185          }          }
3186          return FALSE;
3187        }        }
     else break;  
     }  
   }  
3188    
3189  /* If the next item is one that we can handle, get its value. A non-negative      else
3190  value is a character, a negative value is an escape value. */  #endif  /* SUPPORT_UCP */
3191    
3192  if (*ptr == CHAR_BACKSLASH)      return leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3193    {             rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3194    int temperrorcode = 0;             autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3195    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;  
3196    
3197  /* Skip whitespace and comments in extended mode */    /* Control reaches here only if one of the items is a small character list.
3198      All characters are checked against the other side. */
3199    
3200  if ((options & PCRE_EXTENDED) != 0)    do
   {  
   for (;;)  
3201      {      {
3202      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      chr = *chr_ptr;
3203      if (*ptr == CHAR_NUMBER_SIGN)  
3204        switch(list_ptr[0])
3205        {        {
3206        ptr++;        case OP_CHAR:
3207        while (*ptr != 0)        ochr_ptr = list_ptr + 2;
3208          do
3209          {          {
3210          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (chr == *ochr_ptr) return FALSE;
3211          ptr++;          ochr_ptr++;
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
3212          }          }
3213        }        while(*ochr_ptr != NOTACHAR);
3214      else break;        break;
     }  
   }  
3215    
3216  /* If the next thing is itself optional, we have to give up. */        case OP_NOT:
3217          ochr_ptr = list_ptr + 2;
3218          do
3219            {
3220            if (chr == *ochr_ptr)
3221              break;
3222            ochr_ptr++;
3223            }
3224          while(*ochr_ptr != NOTACHAR);
3225          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3226          break;
3227    
3228  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||        /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3229    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)        set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
     return FALSE;  
3230    
3231  /* Now compare the next item with the previous opcode. First, handle cases when        case OP_DIGIT:
3232  the next item is a character. */        if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3233          break;
3234    
3235  if (next >= 0) switch(op_code)        case OP_NOT_DIGIT:
3236    {        if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3237    case OP_CHAR:        break;
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   return c != next;  
3238    
3239    /* For CHARI (caseless character) we must check the other case. If we have        case OP_WHITESPACE:
3240    Unicode property support, we can use it to test the other case of        if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3241    high-valued characters. */        break;
3242    
3243    case OP_CHARI:        case OP_NOT_WHITESPACE:
3244  #ifdef SUPPORT_UTF        if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3245    GETCHARTEST(c, previous);        break;
 #else  
   c = *previous;  
 #endif  
   if (c == next) return FALSE;  
 #ifdef SUPPORT_UTF  
   if (utf)  
     {  
     unsigned int othercase;  
     if (next < 128) othercase = cd->fcc[next]; else  
 #ifdef SUPPORT_UCP  
     othercase = UCD_OTHERCASE((unsigned int)next);  
 #else  
     othercase = NOTACHAR;  
 #endif  
     return (unsigned int)c != othercase;  
     }  
   else  
 #endif  /* SUPPORT_UTF */  
   return (c != TABLE_GET((unsigned int)next, cd->fcc, next));  /* Non-UTF-8 mode */  
3246    
3247    case OP_NOT:        case OP_WORDCHAR:
3248  #ifdef SUPPORT_UTF        if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3249    GETCHARTEST(c, previous);        break;
 #else  
   c = *previous;  
 #endif  
   return c == next;  
3250    
3251    case OP_NOTI:        case OP_NOT_WORDCHAR:
3252  #ifdef SUPPORT_UTF        if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3253    GETCHARTEST(c, previous);        break;
3254  #else  
3255    c = *previous;        case OP_HSPACE:
3256          switch(chr)
3257            {
3258            HSPACE_CASES: return FALSE;
3259            default: break;
3260            }
3261          break;
3262    
3263          case OP_NOT_HSPACE:
3264          switch(chr)
3265            {
3266            HSPACE_CASES: break;
3267            default: return FALSE;
3268            }
3269          break;
3270    
3271          case OP_ANYNL:
3272          case OP_VSPACE:
3273          switch(chr)
3274            {
3275            VSPACE_CASES: return FALSE;
3276            default: break;
3277            }
3278          break;
3279    
3280          case OP_NOT_VSPACE:
3281          switch(chr)
3282            {
3283            VSPACE_CASES: break;
3284            default: return FALSE;
3285            }
3286          break;
3287    
3288          case OP_DOLL:
3289          case OP_EODN:
3290          switch (chr)
3291            {
3292            case CHAR_CR:
3293            case CHAR_LF:
3294            case CHAR_VT:
3295            case CHAR_FF:
3296            case CHAR_NEL:
3297    #ifndef EBCDIC
3298            case 0x2028:
3299            case 0x2029:
3300    #endif  /* Not EBCDIC */
3301            return FALSE;
3302            }
3303          break;
3304    
3305          case OP_EOD:    /* Can always possessify before \z */
3306          break;
3307    
3308          case OP_PROP:
3309          case OP_NOTPROP:
3310          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3311                list_ptr[0] == OP_NOTPROP))
3312            return FALSE;
3313          break;
3314    
3315          /* The class comparisons work only when the class is the second item
3316          of the pair, because there are at present no possessive forms of the
3317          class opcodes. Note also that the "code" variable that is used below
3318          points after the second item, and that the pointer for the first item
3319          is not available, so even if there were possessive forms of the class
3320          opcodes, the correct comparison could not be done. */
3321    
3322          case OP_NCLASS:
3323          if (chr > 255) return FALSE;
3324          /* Fall through */
3325    
3326          case OP_CLASS:
3327          if (list_ptr != list) return FALSE;   /* Class is first opcode */
3328          if (chr > 255) break;
3329          if ((((pcre_uint8 *)(code - list_ptr[2]))[chr >> 3] & (1 << (chr & 7))) != 0)
3330            return FALSE;
3331          break;
3332    
3333    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3334          case OP_XCLASS:
3335          if (list_ptr != list) return FALSE;   /* Class is first opcode */
3336          if (PRIV(xclass)(chr, code - list_ptr[2] + LINK_SIZE, utf))
3337            return FALSE;
3338          break;
3339  #endif  #endif
3340    if (c == next) return TRUE;  
3341  #ifdef SUPPORT_UTF        default:
3342    if (utf)        return FALSE;
3343          }
3344    
3345        chr_ptr++;
3346        }
3347      while(*chr_ptr != NOTACHAR);
3348    
3349      /* At least one character must be matched from this opcode. */
3350    
3351      if (list[1] == 0) return TRUE;
3352      }
3353    
3354    return FALSE;
3355    }
3356    
3357    
3358    
3359    /*************************************************
3360    *    Scan compiled regex for auto-possession     *
3361    *************************************************/
3362    
3363    /* Replaces single character iterations with their possessive alternatives
3364    if appropriate. This function modifies the compiled opcode!
3365    
3366    Arguments:
3367      code        points to start of the byte code
3368      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3369      cd          static compile data
3370    
3371    Returns:      nothing
3372    */
3373    
3374    static void
3375    auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3376    {
3377    register pcre_uchar c;
3378    const pcre_uchar *end;
3379    pcre_uint32 list[8];
3380    
3381    for (;;)
3382      {
3383      c = *code;
3384    
3385      if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3386      {      {
3387      unsigned int othercase;      c -= get_repeat_base(c) - OP_STAR;
3388      if (next < 128) othercase = cd->fcc[next]; else      end = (c <= OP_MINUPTO) ?
3389  #ifdef SUPPORT_UCP        get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3390      othercase = UCD_OTHERCASE((unsigned int)next);      list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3391    
3392        if (end != NULL && compare_opcodes(end, utf, cd, list))
3393          {
3394          switch(c)
3395            {
3396            case OP_STAR:
3397            *code += OP_POSSTAR - OP_STAR;
3398            break;
3399    
3400            case OP_MINSTAR:
3401            *code += OP_POSSTAR - OP_MINSTAR;
3402            break;
3403    
3404            case OP_PLUS:
3405            *code += OP_POSPLUS - OP_PLUS;
3406            break;
3407    
3408            case OP_MINPLUS:
3409            *code += OP_POSPLUS - OP_MINPLUS;
3410            break;
3411    
3412            case OP_QUERY:
3413            *code += OP_POSQUERY - OP_QUERY;
3414            break;
3415    
3416            case OP_MINQUERY:
3417            *code += OP_POSQUERY - OP_MINQUERY;
3418            break;
3419    
3420            case OP_UPTO:
3421            *code += OP_POSUPTO - OP_UPTO;
3422            break;
3423    
3424            case OP_MINUPTO:
3425            *code += OP_MINUPTO - OP_UPTO;
3426            break;
3427            }
3428          }
3429        c = *code;
3430        }
3431    
3432      switch(c)
3433        {
3434        case OP_END:
3435        return;
3436    
3437        case OP_TYPESTAR:
3438        case OP_TYPEMINSTAR:
3439        case OP_TYPEPLUS:
3440        case OP_TYPEMINPLUS:
3441        case OP_TYPEQUERY:
3442        case OP_TYPEMINQUERY:
3443        case OP_TYPEPOSSTAR:
3444        case OP_TYPEPOSPLUS:
3445        case OP_TYPEPOSQUERY:
3446        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3447        break;
3448    
3449        case OP_TYPEUPTO:
3450        case OP_TYPEMINUPTO:
3451        case OP_TYPEEXACT:
3452        case OP_TYPEPOSUPTO:
3453        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3454          code += 2;
3455        break;
3456    
3457        case OP_XCLASS:
3458        code += GET(code, 1);
3459        break;
3460    
3461        case OP_MARK:
3462        case OP_PRUNE_ARG:
3463        case OP_SKIP_ARG:
3464        case OP_THEN_ARG:
3465        code += code[1];
3466        break;
3467        }
3468    
3469      /* Add in the fixed length from the table */
3470    
3471      code += PRIV(OP_lengths)[c];
3472    
3473      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3474      a multi-byte character. The length in the table is a minimum, so we have to
3475      arrange to skip the extra bytes. */
3476    
3477    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3478      if (utf) switch(c)
3479        {
3480        case OP_CHAR:
3481        case OP_CHARI:
3482        case OP_NOT:
3483        case OP_NOTI:
3484        case OP_STAR:
3485        case OP_MINSTAR:
3486        case OP_PLUS:
3487        case OP_MINPLUS:
3488        case OP_QUERY:
3489        case OP_MINQUERY:
3490        case OP_UPTO:
3491        case OP_MINUPTO:
3492        case OP_EXACT:
3493        case OP_POSSTAR:
3494        case OP_POSPLUS:
3495        case OP_POSQUERY:
3496        case OP_POSUPTO:
3497        case OP_STARI:
3498        case OP_MINSTARI:
3499        case OP_PLUSI:
3500        case OP_MINPLUSI:
3501        case OP_QUERYI:
3502        case OP_MINQUERYI:
3503        case OP_UPTOI:
3504        case OP_MINUPTOI:
3505        case OP_EXACTI:
3506        case OP_POSSTARI:
3507        case OP_POSPLUSI:
3508        case OP_POSQUERYI:
3509        case OP_POSUPTOI:
3510        case OP_NOTSTAR:
3511        case OP_NOTMINSTAR:
3512        case OP_NOTPLUS:
3513        case OP_NOTMINPLUS:
3514        case OP_NOTQUERY:
3515        case OP_NOTMINQUERY:
3516        case OP_NOTUPTO:
3517        case OP_NOTMINUPTO:
3518        case OP_NOTEXACT:
3519        case OP_NOTPOSSTAR:
3520        case OP_NOTPOSPLUS:
3521        case OP_NOTPOSQUERY:
3522        case OP_NOTPOSUPTO:
3523        case OP_NOTSTARI:
3524        case OP_NOTMINSTARI:
3525        case OP_NOTPLUSI:
3526        case OP_NOTMINPLUSI:
3527        case OP_NOTQUERYI:
3528        case OP_NOTMINQUERYI:
3529        case OP_NOTUPTOI:
3530        case OP_NOTMINUPTOI:
3531        case OP_NOTEXACTI:
3532        case OP_NOTPOSSTARI:
3533        case OP_NOTPOSPLUSI:
3534        case OP_NOTPOSQUERYI:
3535        case OP_NOTPOSUPTOI:
3536        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3537        break;
3538        }
3539  #else  #else
3540      othercase = NOTACHAR;    (void)(utf);  /* Keep compiler happy by referencing function argument */
3541  #endif  #endif
3542      return (unsigned int)c == othercase;    }
3543      }  }
3544    
3545    
3546    
3547    /*************************************************
3548    *           Check for POSIX class syntax         *
3549    *************************************************/
3550    
3551    /* This function is called when the sequence "[:" or "[." or "[=" is
3552    encountered in a character class. It checks whether this is followed by a
3553    sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3554    reach an unescaped ']' without the special preceding character, return FALSE.
3555    
3556    Originally, this function only recognized a sequence of letters between the
3557    terminators, but it seems that Perl recognizes any sequence of characters,
3558    though of course unknown POSIX names are subsequently rejected. Perl gives an
3559    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3560    didn't consider this to be a POSIX class. Likewise for [:1234:].
3561    
3562    The problem in trying to be exactly like Perl is in the handling of escapes. We
3563    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3564    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3565    below handles the special case of \], but does not try to do any other escape
3566    processing. This makes it different from Perl for cases such as [:l\ower:]
3567    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3568    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
3569    I think.
3570    
3571    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3572    It seems that the appearance of a nested POSIX class supersedes an apparent
3573    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3574    a digit.
3575    
3576    In Perl, unescaped square brackets may also appear as part of class names. For
3577    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3578    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3579    seem right at all. PCRE does not allow closing square brackets in POSIX class
3580    names.
3581    
3582    Arguments:
3583      ptr      pointer to the initial [
3584      endptr   where to return the end pointer
3585    
3586    Returns:   TRUE or FALSE
3587    */
3588    
3589    static BOOL
3590    check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3591    {
3592    pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
3593    terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
3594    for (++ptr; *ptr != CHAR_NULL; ptr++)
3595      {
3596      if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3597        ptr++;
3598      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3599    else    else
3600  #endif  /* SUPPORT_UTF */      {
3601    return (c == TABLE_GET((unsigned int)next, cd->fcc, next));  /* Non-UTF-8 mode */      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3602          {
3603          *endptr = ptr;
3604          return TRUE;
3605          }
3606        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3607             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3608              ptr[1] == CHAR_EQUALS_SIGN) &&
3609            check_posix_syntax(ptr, endptr))
3610          return FALSE;
3611        }
3612      }
3613    return FALSE;
3614    }
3615    
3616    
3617    
3618    
3619    /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.  /*************************************************
3620    When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */  *          Check POSIX class name                *
3621    *************************************************/
3622    
3623    /* This function is called to check the name given in a POSIX-style class entry
3624    such as [:alnum:].
3625    
3626    Arguments:
3627      ptr        points to the first letter
3628      len        the length of the name
3629    
3630    Returns:     a value representing the name, or -1 if unknown
3631    */
3632    
3633    static int
3634    check_posix_name(const pcre_uchar *ptr, int len)
3635    {
3636    const char *pn = posix_names;
3637    register int yield = 0;
3638    while (posix_name_lengths[yield] != 0)
3639      {
3640      if (len == posix_name_lengths[yield] &&
3641        STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3642      pn += posix_name_lengths[yield] + 1;
3643      yield++;
3644      }
3645    return -1;
3646    }
3647    
3648    
3649    /*************************************************
3650    *    Adjust OP_RECURSE items in repeated group   *
3651    *************************************************/
3652    
3653    /* OP_RECURSE items contain an offset from the start of the regex to the group
3654    that is referenced. This means that groups can be replicated for fixed
3655    repetition simply by copying (because the recursion is allowed to refer to
3656    earlier groups that are outside the current group). However, when a group is
3657    optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3658    inserted before it, after it has been compiled. This means that any OP_RECURSE
3659    items within it that refer to the group itself or any contained groups have to
3660    have their offsets adjusted. That one of the jobs of this function. Before it
3661    is called, the partially compiled regex must be temporarily terminated with
3662    OP_END.
3663    
3664    case OP_DIGIT:  This function has been extended with the possibility of forward references for
3665    return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;  recursions and subroutine calls. It must also check the list of such references
3666    for the group we are dealing with. If it finds that one of the recursions in
3667    the current group is on this list, it adjusts the offset in the list, not the
3668    value in the reference (which is a group number).
3669    
3670    case OP_NOT_DIGIT:  Arguments:
3671    return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;    group      points to the start of the group
3672      adjust     the amount by which the group is to be moved
3673      utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
3674      cd         contains pointers to tables etc.
3675      save_hwm   the hwm forward reference pointer at the start of the group
3676    
3677    case OP_WHITESPACE:  Returns:     nothing
3678    return next > 255 || (cd->ctypes[next] & ctype_space) == 0;  */
3679    
3680    case OP_NOT_WHITESPACE:  static void
3681    return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3682      pcre_uchar *save_hwm)
3683    {
3684    pcre_uchar *ptr = group;
3685    
3686    case OP_WORDCHAR:  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3687    return next > 255 || (cd->ctypes[next] & ctype_word) == 0;    {
3688      int offset;
3689      pcre_uchar *hc;
3690    
3691    case OP_NOT_WORDCHAR:    /* See if this recursion is on the forward reference list. If so, adjust the
3692    return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;    reference. */
3693    
3694    case OP_HSPACE:    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
   case OP_NOT_HSPACE:  
   switch(next)  
3695      {      {
3696      HSPACE_CASES:      offset = (int)GET(hc, 0);
3697      return op_code == OP_NOT_HSPACE;      if (cd->start_code + offset == ptr + 1)
3698          {
3699      default:        PUT(hc, 0, offset + adjust);
3700      return op_code != OP_NOT_HSPACE;        break;
3701          }
3702      }      }
3703    
3704    case OP_ANYNL:    /* Otherwise, adjust the recursion offset if it's after the start of this
3705    case OP_VSPACE:    group. */
3706    case OP_NOT_VSPACE:  
3707    switch(next)    if (hc >= cd->hwm)
3708      {      {
3709      VSPACE_CASES:      offset = (int)GET(ptr, 1);
3710      return op_code == OP_NOT_VSPACE;      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
   
     default:  
     return op_code != OP_NOT_VSPACE;  
3711      }      }
3712    
3713  #ifdef SUPPORT_UCP    ptr += 1 + LINK_SIZE;
   case OP_PROP:  
   return check_char_prop(next, previous[0], previous[1], FALSE);  
   
   case OP_NOTPROP:  
   return check_char_prop(next, previous[0], previous[1], TRUE);  
 #endif  
   
   default:  
   return FALSE;  
3714    }    }
3715    }
3716    
3717    
 /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP  
 is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are  
 generated only when PCRE_UCP is *not* set, that is, when only ASCII  
 characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are  
 replaced by OP_PROP codes when PCRE_UCP is set. */  
   
 switch(op_code)  
   {  
   case OP_CHAR:  
   case OP_CHARI:  
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   switch(-next)  
     {  
     case ESC_d:  
     return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;  
   
     case ESC_D:  
     return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;  
   
     case ESC_s:  
     return c > 255 || (cd->ctypes[c] & ctype_space) == 0;  
   
     case ESC_S:  
     return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;  
   
     case ESC_w:  
     return c > 255 || (cd->ctypes[c] & ctype_word) == 0;  
3718    
3719      case ESC_W:  /*************************************************
3720      return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;  *        Insert an automatic callout point       *
3721    *************************************************/
3722    
3723      case ESC_h:  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
3724      case ESC_H:  callout points before each pattern item.
     switch(c)  
       {  
       HSPACE_CASES:  
       return -next != ESC_h;  
   
       default:  
       return -next == ESC_h;  
       }  
3725    
3726      case ESC_v:  Arguments:
3727      case ESC_V:    code           current code pointer
3728      switch(c)    ptr            current pattern pointer
3729        {    cd             pointers to tables etc
       VSPACE_CASES:  
       return -next != ESC_v;  
   
       default:  
       return -next == ESC_v;  
       }  
3730    
3731      /* When PCRE_UCP is set, these values get generated for \d etc. Find  Returns:         new code pointer
3732      their substitutions and process them. The result will always be either  */
     -ESC_p or -ESC_P. Then fall through to process those values. */  
3733    
3734  #ifdef SUPPORT_UCP  static pcre_uchar *
3735      case ESC_du:  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
3736      case ESC_DU:  {
3737      case ESC_wu:  *code++ = OP_CALLOUT;
3738      case ESC_WU:  *code++ = 255;
3739      case ESC_su:  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
3740      case ESC_SU:  PUT(code, LINK_SIZE, 0);                       /* Default length */
3741        {  return code + 2 * LINK_SIZE;
3742        int temperrorcode = 0;  }
       ptr = substitutes[-next - ESC_DU];  
       next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);  
       if (temperrorcode != 0) return FALSE;  
       ptr++;    /* For compatibility */  
       }  
     /* Fall through */  
3743    
     case ESC_p:  
     case ESC_P:  
       {  
       int ptype, pdata, errorcodeptr;  
       BOOL negated;  
3744    
       ptr--;      /* Make ptr point at the p or P */  
       ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);  
       if (ptype < 0) return FALSE;  
       ptr++;      /* Point past the final curly ket */  
3745    
3746        /* If the property item is optional, we have to give up. (When generated  /*************************************************
3747        from \d etc by PCRE_UCP, this test will have been applied much earlier,  *         Complete a callout item                *
3748        to the original \d etc. At this point, ptr will point to a zero byte. */  *************************************************/
3749    
3750        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  /* A callout item contains the length of the next item in the pattern, which
3751          STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)  we can't fill in till after we have reached the relevant point. This is used
3752            return FALSE;  for both automatic and manual callouts.
3753    
3754        /* Do the property check. */  Arguments:
3755      previous_callout   points to previous callout item
3756      ptr                current pattern pointer
3757      cd                 pointers to tables etc
3758    
3759        return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);  Returns:             nothing
3760        }  */
 #endif  
3761    
3762      default:  static void
3763      return FALSE;  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
3764      }  {
3765    int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
3766    PUT(previous_callout, 2 + LINK_SIZE, length);
3767    }
3768    
   /* In principle, support for Unicode properties should be integrated here as  
   well. It means re-organizing the above code so as to get hold of the property  
   values before switching on the op-code. However, I wonder how many patterns  
   combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,  
   these op-codes are never generated.) */  
3769    
   case OP_DIGIT:  
   return next == -ESC_D || next == -ESC_s || next == -ESC_W ||  
          next == -ESC_h || next == -ESC_v || next == -ESC_R;  
3770    
3771    case OP_NOT_DIGIT:  #ifdef SUPPORT_UCP
3772    return next == -ESC_d;  /*************************************************
3773    *           Get othercase range                  *
3774    *************************************************/
3775    
3776    case OP_WHITESPACE:  /* This function is passed the start and end of a class range, in UTF-8 mode
3777    return next == -ESC_S || next == -ESC_d || next == -ESC_w;  with UCP support. It searches up the characters, looking for ranges of
3778    characters in the "other" case. Each call returns the next one, updating the
3779    start address. A character with multiple other cases is returned on its own
3780    with a special return value.
3781    
3782    case OP_NOT_WHITESPACE:  Arguments:
3783    return next == -ESC_s || next == -ESC_h || next == -ESC_v || next == -ESC_R;    cptr        points to starting character value; updated
3784      d           end value
3785      ocptr       where to put start of othercase range
3786      odptr       where to put end of othercase range
3787    
3788    case OP_HSPACE:  Yield:        -1 when no more
3789    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||                 0 when a range is returned
3790           next == -ESC_w || next == -ESC_v || next == -ESC_R;                >0 the CASESET offset for char with multiple other cases
3791                    in this case, ocptr contains the original
3792    */
3793    
3794    case OP_NOT_HSPACE:  static int
3795    return next == -ESC_h;  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
3796      pcre_uint32 *odptr)
3797    {
3798    pcre_uint32 c, othercase, next;
3799    unsigned int co;
3800    
3801    /* Can't have \S in here because VT matches \S (Perl anomaly) */  /* Find the first character that has an other case. If it has multiple other
3802    case OP_ANYNL:  cases, return its case offset value. */
   case OP_VSPACE:  
   return next == -ESC_V || next == -ESC_d || next == -ESC_w;  
3803    
3804    case OP_NOT_VSPACE:  for (c = *cptr; c <= d; c++)
3805    return next == -ESC_v || next == -ESC_R;    {
3806      if ((co = UCD_CASESET(c)) != 0)
3807        {
3808        *ocptr = c++;   /* Character that has the set */
3809        *cptr = c;      /* Rest of input range */
3810        return (int)co;
3811        }
3812      if ((othercase = UCD_OTHERCASE(c)) != c) break;
3813      }
3814    
3815    case OP_WORDCHAR:  if (c > d) return -1;  /* Reached end of range */
   return next == -ESC_W || next == -ESC_s || next == -ESC_h ||  
          next == -ESC_v || next == -ESC_R;  
3816    
3817    case OP_NOT_WORDCHAR:  *ocptr = othercase;
3818    return next == -ESC_w || next == -ESC_d;  next = othercase + 1;
3819    
3820    default:  for (++c; c <= d; c++)
3821    return FALSE;    {
3822      if (UCD_OTHERCASE(c) != next) break;
3823      next++;
3824    }    }
3825    
3826  /* Control does not reach here */  *odptr = next - 1;     /* End of othercase range */
3827    *cptr = c;             /* Rest of input range */
3828    return 0;
3829  }  }
3830    #endif  /* SUPPORT_UCP */
3831    
3832    
3833    
# Line 3381  switch(op_code) Line 3836  switch(op_code)
3836  *************************************************/  *************************************************/
3837    
3838  /* This function packages up the logic of adding a character or range of  /* This function packages up the logic of adding a character or range of
3839  characters to a class. The character values in the arguments will be within the  characters to a class. The character values in the arguments will be within the
3840  valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is  valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
3841  mutually recursive with the function immediately below.  mutually recursive with the function immediately below.
3842    
3843  Arguments:  Arguments:
3844    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
3845    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
3846    options       the options word    options       the options word
3847    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
3848    start         start of range character    start         start of range character
3849    end           end of range character    end           end of range character
3850    
3851  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
3852                  the pointer to extra data is updated                  the pointer to extra data is updated
3853  */  */
3854    
3855  static int  static int
3856  add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,  add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
3857    compile_data *cd, unsigned int start, unsigned int end)    compile_data *cd, pcre_uint32 start, pcre_uint32 end)
3858  {  {
3859  unsigned int c;  pcre_uint32 c;
3860  int n8 = 0;  int n8 = 0;
3861    
3862  /* If caseless matching is required, scan the range and process alternate  /* If caseless matching is required, scan the range and process alternate
3863  cases. In Unicode, there are 8-bit characters that have alternate cases that  cases. In Unicode, there are 8-bit characters that have alternate cases that
3864  are greater than 255 and vice-versa. Sometimes we can just extend the original  are greater than 255 and vice-versa. Sometimes we can just extend the original
3865  range. */  range. */
3866    
3867  if ((options & PCRE_CASELESS) != 0)  if ((options & PCRE_CASELESS) != 0)
3868    {    {
3869  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3870    if ((options & PCRE_UTF8) != 0)    if ((options & PCRE_UTF8) != 0)
3871      {      {
3872      int rc;      int rc;
3873      unsigned int oc, od;      pcre_uint32 oc, od;
3874    
3875      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
3876      c = start;      c = start;
3877    
3878      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
3879        {        {
3880        /* Handle a single character that has more than one other case. */        /* Handle a single character that has more than one other case. */
3881    
3882        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
3883          PRIV(ucd_caseless_sets) + rc, oc);          PRIV(ucd_caseless_sets) + rc, oc);
3884    
3885        /* Do nothing if the other case range is within the original range. */        /* Do nothing if the other case range is within the original range. */
3886    
3887        else if (oc >= start && od <= end) continue;        else if (oc >= start && od <= end) continue;
3888    
3889        /* Extend the original range if there is overlap, noting that if oc < c, we        /* Extend the original range if there is overlap, noting that if oc < c, we
3890        can't have od > end because a subrange is always shorter than the basic        can't have od > end because a subrange is always shorter than the basic
3891        range. Otherwise, use a recursive call to add the additional range. */        range. Otherwise, use a recursive call to add the additional range. */
3892    
3893        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
3894        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
3895        else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);        else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
# Line 3444  if ((options & PCRE_CASELESS) != 0) Line 3899  if ((options & PCRE_CASELESS) != 0)
3899  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3900    
3901    /* Not UTF-mode, or no UCP */    /* Not UTF-mode, or no UCP */
3902    
3903    for (c = start; c <= end && c < 256; c++)    for (c = start; c <= end && c < 256; c++)
3904      {      {
3905      SETBIT(classbits, cd->fcc[c]);      SETBIT(classbits, cd->fcc[c]);
3906      n8++;      n8++;
3907      }      }
3908    }    }
3909    
3910  /* Now handle the original range. Adjust the final value according to the bit  /* Now handle the original range. Adjust the final value according to the bit
3911  length - this means that the same lists of (e.g.) horizontal spaces can be used  length - this means that the same lists of (e.g.) horizontal spaces can be used
3912  in all cases. */  in all cases. */
3913    
3914  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
3915  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
3916    if ((options & PCRE_UTF8) == 0)    if ((options & PCRE_UTF8) == 0)
3917  #endif  #endif
3918    if (end > 0xff) end = 0xff;    if (end > 0xff) end = 0xff;
 #endif  
3919    
3920  #ifdef COMPILE_PCRE16  #elif defined COMPILE_PCRE16
3921  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
3922    if ((options & PCRE_UTF16) == 0)    if ((options & PCRE_UTF16) == 0)
3923  #endif  #endif
3924    if (end > 0xffff) end = 0xffff;    if (end > 0xffff) end = 0xffff;
3925  #endif  
3926    #endif /* COMPILE_PCRE[8|16] */
3927    
3928  /* If all characters are less than 256, use the bit map. Otherwise use extra  /* If all characters are less than 256, use the bit map. Otherwise use extra
3929  data. */  data. */
# Line 3477  if (end < 0x100) Line 3932  if (end < 0x100)
3932    {    {
3933    for (c = start; c <= end; c++)    for (c = start; c <= end; c++)
3934      {      {
3935      n8++;      n8++;
3936      SETBIT(classbits, c);      SETBIT(classbits, c);
3937      }      }
3938    }    }
3939    
3940  else  else
3941    {    {
3942    pcre_uchar *uchardata = *uchardptr;    pcre_uchar *uchardata = *uchardptr;
3943    
3944  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
3945    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
3946      {      {
3947      if (start < end)      if (start < end)
3948        {        {
3949        *uchardata++ = XCL_RANGE;        *uchardata++ = XCL_RANGE;
3950        uchardata += PRIV(ord2utf)(start, uchardata);        uchardata += PRIV(ord2utf)(start, uchardata);
3951        uchardata += PRIV(ord2utf)(end, uchardata);        uchardata += PRIV(ord2utf)(end, uchardata);
3952        }        }
3953      else if (start == end)      else if (start == end)
3954        {        {
3955        *uchardata++ = XCL_SINGLE;        *uchardata++ = XCL_SINGLE;
3956        uchardata += PRIV(ord2utf)(start, uchardata);        uchardata += PRIV(ord2utf)(start, uchardata);
3957        }        }
3958      }      }
3959    else    else
3960  #endif  /* SUPPORT_UTF */  #endif  /* SUPPORT_UTF */
3961    
3962    /* Without UTF support, character values are constrained by the bit length,    /* Without UTF support, character values are constrained by the bit length,
3963    and can only be > 256 for 16-bit and 32-bit libraries. */    and can only be > 256 for 16-bit and 32-bit libraries. */
3964    
3965  #ifdef COMPILE_PCRE8  #ifdef COMPILE_PCRE8
3966      {}      {}
3967  #else  #else
3968    if (start < end)    if (start < end)
3969      {      {
3970      *uchardata++ = XCL_RANGE;      *uchardata++ = XCL_RANGE;
# Line 3520  else Line 3975  else
3975      {      {
3976      *uchardata++ = XCL_SINGLE;      *uchardata++ = XCL_SINGLE;
3977      *uchardata++ = start;      *uchardata++ = start;
3978      }      }
3979  #endif  #endif
3980    
3981    *uchardptr = uchardata;   /* Updata extra data pointer */    *uchardptr = uchardata;   /* Updata extra data pointer */
3982    }    }
3983    
3984  return n8;    /* Number of 8-bit characters */  return n8;    /* Number of 8-bit characters */
3985  }  }
3986    
3987    
3988    
3989    
3990  /*************************************************  /*************************************************
3991  *        Add a list of characters to a class     *  *        Add a list of characters to a class     *
3992  *************************************************/  *************************************************/
3993    
3994  /* This function is used for adding a list of case-equivalent characters to a  /* This function is used for adding a list of case-equivalent characters to a
3995  class, and also for adding a list of horizontal or vertical whitespace. If the  class, and also for adding a list of horizontal or vertical whitespace. If the
3996  list is in order (which it should be), ranges of characters are detected and  list is in order (which it should be), ranges of characters are detected and
3997  handled appropriately. This function is mutually recursive with the function  handled appropriately. This function is mutually recursive with the function
# Line 3546  Arguments: Line 4001  Arguments:
4001    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4002    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4003    options       the options word    options       the options word
4004    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4005    p             points to row of 32-bit values, terminated by NOTACHAR    p             points to row of 32-bit values, terminated by NOTACHAR
4006    except        character to omit; this is used when adding lists of    except        character to omit; this is used when adding lists of
4007                    case-equivalent characters to avoid including the one we                    case-equivalent characters to avoid including the one we
4008                    already know about                    already know about
4009    
4010  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4011                  the pointer to extra data is updated                  the pointer to extra data is updated
4012  */  */
# Line 3565  while (p[0] < NOTACHAR) Line 4020  while (p[0] < NOTACHAR)
4020    {    {
4021    int n = 0;    int n = 0;
4022    if (p[0] != except)    if (p[0] != except)
4023      {      {
4024      while(p[n+1] == p[0] + n + 1) n++;      while(p[n+1] == p[0] + n + 1) n++;
4025      n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);      n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4026      }      }
4027    p += n + 1;    p += n + 1;
4028    }    }
4029  return n8;  return n8;
4030  }  }
4031    
4032    
4033    
# Line 3587  Arguments: Line 4042  Arguments:
4042    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4043    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4044    options       the options word    options       the options word
4045    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4046    p             points to row of 32-bit values, terminated by NOTACHAR    p             points to row of 32-bit values, terminated by NOTACHAR
4047    
4048  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4049                  the pointer to extra data is updated                  the pointer to extra data is updated
4050  */  */
4051    
4052  static int  static int
4053  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
4054    int options, compile_data *cd, const pcre_uint32 *p)    int options, compile_data *cd, const pcre_uint32 *p)
4055  {  {
4056    BOOL utf = (options & PCRE_UTF8) != 0;
4057  int n8 = 0;  int n8 = 0;
4058  if (p[0] > 0)  if (p[0] > 0)
4059    n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);    n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
4060  while (p[0] < NOTACHAR)  while (p[0] < NOTACHAR)
4061    {    {
4062    while (p[1] == p[0] + 1) p++;    while (p[1] == p[0] + 1) p++;
4063    n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,    n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
4064      (p[1] == NOTACHAR)? 0x10ffff : p[1] - 1);      (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
4065    p++;    p++;
4066    }    }
4067  return n8;  return n8;
4068  }  }
4069    
4070    
4071    
# Line 3624  to find out the amount of memory needed, Line 4080  to find out the amount of memory needed,
4080  phase. The value of lengthptr distinguishes the two phases.  phase. The value of lengthptr distinguishes the two phases.
4081    
4082  Arguments:  Arguments:
4083    optionsptr     pointer to the option bits    optionsptr        pointer to the option bits
4084    codeptr        points to the pointer to the current code point    codeptr           points to the pointer to the current code point
4085    ptrptr         points to the current pattern pointer    ptrptr            points to the current pattern pointer
4086    errorcodeptr   points to error code variable    errorcodeptr      points to error code variable
4087    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr      place to put the first required character
4088    reqcharptr     set to the last literal character required, else < 0    firstcharflagsptr place to put the first character flags, or a negative number
4089    bcptr          points to current branch chain    reqcharptr        place to put the last required character
4090    cond_depth     conditional nesting depth    reqcharflagsptr   place to put the last required character flags, or a negative number
4091    cd             contains pointers to tables etc.    bcptr             points to current branch chain
4092    lengthptr      NULL during the real compile phase    cond_depth        conditional nesting depth
4093                   points to length accumulator during pre-compile phase    cd                contains pointers to tables etc.
4094      lengthptr         NULL during the real compile phase
4095                        points to length accumulator during pre-compile phase
4096    
4097  Returns:         TRUE on success  Returns:            TRUE on success
4098                   FALSE, with *errorcodeptr set non-zero on error                      FALSE, with *errorcodeptr set non-zero on error
4099  */  */
4100    
4101  static BOOL  static BOOL
4102  compile_branch(int *optionsptr, pcre_uchar **codeptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
4103    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,    const pcre_uchar **ptrptr, int *errorcodeptr,
4104    pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,    pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
4105      pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
4106      branch_chain *bcptr, int cond_depth,
4107    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
4108  {  {
4109  int repeat_type, op_type;  int repeat_type, op_type;
4110  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
4111  int bravalue = 0;  int bravalue = 0;
4112  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
4113  pcre_int32 firstchar, reqchar;  pcre_uint32 firstchar, reqchar;
4114  pcre_int32 zeroreqchar, zerofirstchar;  pcre_int32 firstcharflags, reqcharflags;
4115    pcre_uint32 zeroreqchar, zerofirstchar;
4116    pcre_int32 zeroreqcharflags, zerofirstcharflags;
4117  pcre_int32 req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
4118  int options = *optionsptr;               /* May change dynamically */  int options = *optionsptr;               /* May change dynamically */
4119  int after_manual_callout = 0;  int after_manual_callout = 0;
4120  int length_prevgroup = 0;  int length_prevgroup = 0;
4121  register int c;  register pcre_uint32 c;
4122    int escape;
4123  register pcre_uchar *code = *codeptr;  register pcre_uchar *code = *codeptr;
4124  pcre_uchar *last_code = code;  pcre_uchar *last_code = code;
4125  pcre_uchar *orig_code = code;  pcre_uchar *orig_code = code;
# Line 3676  must not do this for other options (e.g. Line 4139  must not do this for other options (e.g.
4139  dynamically as we process the pattern. */  dynamically as we process the pattern. */
4140    
4141  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4142  /* PCRE_UTF16 has the same value as PCRE_UTF8. */  /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
4143  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
4144    #ifndef COMPILE_PCRE32
4145  pcre_uchar utf_chars[6];  pcre_uchar utf_chars[6];
4146    #endif
4147  #else  #else
4148  BOOL utf = FALSE;  BOOL utf = FALSE;
4149  #endif  #endif
4150    
4151  /* Helper variables for OP_XCLASS opcode (for characters > 255). */  /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
4152    class_uchardata always so that it can be passed to add_to_class() always,
4153    though it will not be used in non-UTF 8-bit cases. This avoids having to supply
4154    alternative calls for the different cases. */
4155    
4156    pcre_uchar *class_uchardata;
4157  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4158  BOOL xclass;  BOOL xclass;
 pcre_uchar *class_uchardata;  
4159  pcre_uchar *class_uchardata_base;  pcre_uchar *class_uchardata_base;
4160  #endif  #endif
4161    
# Line 3710  to take the zero repeat into account. Th Line 4178  to take the zero repeat into account. Th
4178  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
4179  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
4180    
4181  firstchar = reqchar = zerofirstchar = zeroreqchar = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
4182    firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
4183    
4184  /* The variable req_caseopt contains either the REQ_CASELESS value  /* The variable req_caseopt contains either the REQ_CASELESS value
4185  or zero, according to the current setting of the caseless flag. The  or zero, according to the current setting of the caseless flag. The
# Line 3736  for (;; ptr++) Line 4205  for (;; ptr++)
4205    int recno;    int recno;
4206    int refsign;    int refsign;
4207    int skipbytes;    int skipbytes;
4208    int subreqchar;    pcre_uint32 subreqchar, subfirstchar;
4209    int subfirstchar;    pcre_int32 subreqcharflags, subfirstcharflags;
4210    int terminator;    int terminator;
4211    int mclength;    unsigned int mclength;
4212    int tempbracount;    unsigned int tempbracount;
4213      pcre_uint32 ec;
4214    pcre_uchar mcbuffer[8];    pcre_uchar mcbuffer[8];
4215    
4216    /* Get next character in the pattern */    /* Get next character in the pattern */
# Line 3750  for (;; ptr++) Line 4220  for (;; ptr++)
4220    /* 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
4221    string. Nesting only happens one level deep. */    string. Nesting only happens one level deep. */
4222    
4223    if (c == 0 && nestptr != NULL)    if (c == CHAR_NULL && nestptr != NULL)
4224      {      {
4225      ptr = nestptr;      ptr = nestptr;
4226      nestptr = NULL;      nestptr = NULL;
# Line 3825  for (;; ptr++) Line 4295  for (;; ptr++)
4295    
4296    /* 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 */
4297    
4298    if (inescq && c != 0)    if (inescq && c != CHAR_NULL)
4299      {      {
4300      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4301        {        {
# Line 3850  for (;; ptr++) Line 4320  for (;; ptr++)
4320        }        }
4321      }      }
4322    
   /* Fill in length of a previous callout, except when the next thing is  
   a quantifier. */  
   
4323    is_quantifier =    is_quantifier =
4324      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4325      (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));      (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4326    
4327    if (!is_quantifier && previous_callout != NULL &&    /* Fill in length of a previous callout, except when the next thing is a
4328      quantifier or when processing a property substitution string in UCP mode. */
4329    
4330      if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4331         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
4332      {      {
4333      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
# Line 3873  for (;; ptr++) Line 4343  for (;; ptr++)
4343      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
4344        {        {
4345        ptr++;        ptr++;
4346        while (*ptr != 0)        while (*ptr != CHAR_NULL)
4347          {          {
4348          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
4349          ptr++;          ptr++;
# Line 3881  for (;; ptr++) Line 4351  for (;; ptr++)
4351          if (utf) FORWARDCHAR(ptr);          if (utf) FORWARDCHAR(ptr);
4352  #endif  #endif
4353          }          }
4354        if (*ptr != 0) continue;        if (*ptr != CHAR_NULL) continue;
4355    
4356        /* Else fall through to handle end of string */        /* Else fall through to handle end of string */
4357        c = 0;        c = 0;
4358        }        }
4359      }      }
4360    
4361    /* No auto callout for quantifiers. */    /* No auto callout for quantifiers, or while processing property strings that
4362      are substituted for \w etc in UCP mode. */
4363    
4364    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4365      {      {
4366      previous_callout = code;      previous_callout = code;
4367      code = auto_callout(code, ptr, cd);      code = auto_callout(code, ptr, cd);
# Line 3903  for (;; ptr++) Line 4374  for (;; ptr++)
4374      case CHAR_VERTICAL_LINE:       /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
4375      case CHAR_RIGHT_PARENTHESIS:      case CHAR_RIGHT_PARENTHESIS:
4376      *firstcharptr = firstchar;      *firstcharptr = firstchar;
4377        *firstcharflagsptr = firstcharflags;
4378      *reqcharptr = reqchar;      *reqcharptr = reqchar;
4379        *reqcharflagsptr = reqcharflags;
4380      *codeptr = code;      *codeptr = code;
4381      *ptrptr = ptr;      *ptrptr = ptr;
4382      if (lengthptr != NULL)      if (lengthptr != NULL)
# Line 3927  for (;; ptr++) Line 4400  for (;; ptr++)
4400      previous = NULL;      previous = NULL;
4401      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
4402        {        {
4403        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;        if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4404        *code++ = OP_CIRCM;        *code++ = OP_CIRCM;
4405        }        }
4406      else *code++ = OP_CIRC;      else *code++ = OP_CIRC;
# Line 3942  for (;; ptr++) Line 4415  for (;; ptr++)
4415      repeats. The value of reqchar doesn't change either. */      repeats. The value of reqchar doesn't change either. */
4416    
4417      case CHAR_DOT:      case CHAR_DOT:
4418      if (firstchar == REQ_UNSET) firstchar = REQ_NONE;      if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4419      zerofirstchar = firstchar;      zerofirstchar = firstchar;
4420        zerofirstcharflags = firstcharflags;
4421      zeroreqchar = reqchar;      zeroreqchar = reqchar;
4422        zeroreqcharflags = reqcharflags;
4423      previous = code;      previous = code;
4424      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
4425      break;      break;
# Line 4018  for (;; ptr++) Line 4493  for (;; ptr++)
4493          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4494        {        {
4495        *code++ = negate_class? OP_ALLANY : OP_FAIL;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
4496        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;        if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4497        zerofirstchar = firstchar;        zerofirstchar = firstchar;
4498          zerofirstcharflags = firstcharflags;
4499        break;        break;
4500        }        }
4501    
# Line 4054  for (;; ptr++) Line 4530  for (;; ptr++)
4530      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
4531      loop, c contains the first byte of the character. */      loop, c contains the first byte of the character. */
4532    
4533      if (c != 0) do      if (c != CHAR_NULL) do
4534        {        {
4535        const pcre_uchar *oldptr;        const pcre_uchar *oldptr;
4536    
# Line 4069  for (;; ptr++) Line 4545  for (;; ptr++)
4545        /* In the pre-compile phase, accumulate the length of any extra        /* In the pre-compile phase, accumulate the length of any extra
4546        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
4547        contain a zillion > 255 characters no longer overwrite the work space        contain a zillion > 255 characters no longer overwrite the work space
4548        (which is on the stack). We have to remember that there was XCLASS data,        (which is on the stack). We have to remember that there was XCLASS data,
4549        however. */        however. */
4550    
4551        if (lengthptr != NULL && class_uchardata > class_uchardata_base)        if (lengthptr != NULL && class_uchardata > class_uchardata_base)
# Line 4210  for (;; ptr++) Line 4686  for (;; ptr++)
4686    
4687        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
4688          {          {
4689          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options,
4690              TRUE);
4691          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
4692            if (escape == 0) c = ec;
4693          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 */
4694          else if (-c == ESC_N)            /* \N is not supported in a class */          else if (escape == ESC_N)          /* \N is not supported in a class */
4695            {            {
4696            *errorcodeptr = ERR71;            *errorcodeptr = ERR71;
4697            goto FAILED;            goto FAILED;
4698            }            }
4699          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (escape == ESC_Q)            /* Handle start of quoted string */
4700            {            {
4701            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
4702              {              {
# Line 4228  for (;; ptr++) Line 4705  for (;; ptr++)
4705            else inescq = TRUE;            else inescq = TRUE;
4706            continue;            continue;
4707            }            }
4708          else if (-c == ESC_E) continue;  /* Ignore orphan \E */          else if (escape == ESC_E) continue;  /* Ignore orphan \E */
4709    
4710          if (c < 0)          else
4711            {            {
4712            register const pcre_uint8 *cbits = cd->cbits;            register const pcre_uint8 *cbits = cd->cbits;
4713            /* Every class contains at least two < 256 characters. */            /* Every class contains at least two < 256 characters. */
# Line 4238  for (;; ptr++) Line 4715  for (;; ptr++)
4715            /* Every class contains at least two characters. */            /* Every class contains at least two characters. */
4716            class_one_char += 2;            class_one_char += 2;
4717    
4718            switch (-c)            switch (escape)
4719              {              {
4720  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4721              case ESC_du:     /* These are the values given for \d etc */              case ESC_du:     /* These are the values given for \d etc */
# Line 4248  for (;; ptr++) Line 4725  for (;; ptr++)
4725              case ESC_su:     /* of the default ASCII testing. */              case ESC_su:     /* of the default ASCII testing. */
4726              case ESC_SU:              case ESC_SU:
4727              nestptr = ptr;              nestptr = ptr;
4728              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */              ptr = substitutes[escape - ESC_DU] - 1;  /* Just before substitute */
4729              class_has_8bitchar--;                /* Undo! */              class_has_8bitchar--;                /* Undo! */
4730              continue;              continue;
4731  #endif  #endif
# Line 4270  for (;; ptr++) Line 4747  for (;; ptr++)
4747              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
4748              continue;              continue;
4749    
4750              /* 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
4751              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
4752              class. Luckily, the value of CHAR_VT is 0x0b in both ASCII and              previously set by something earlier in the character class.
4753              EBCDIC, so we lazily just adjust the appropriate bit. */              Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
4754                we could just adjust the appropriate bit. From PCRE 8.34 we no
4755                longer treat \s and \S specially. */
4756    
4757              case ESC_s:              case ESC_s:
4758              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];  
4759              continue;              continue;
4760    
4761              case ESC_S:              case ESC_S:
4762              should_flip_negation = TRUE;              should_flip_negation = TRUE;
4763              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 */  
4764              continue;              continue;
4765    
4766              /* The rest apply in both UCP and non-UCP cases. */              /* The rest apply in both UCP and non-UCP cases. */
4767    
4768              case ESC_h:              case ESC_h:
4769              (void)add_list_to_class(classbits, &class_uchardata, options, cd,              (void)add_list_to_class(classbits, &class_uchardata, options, cd,
4770                PRIV(hspace_list), NOTACHAR);                PRIV(hspace_list), NOTACHAR);
4771              continue;              continue;
4772    
4773              case ESC_H:              case ESC_H:
4774              (void)add_not_list_to_class(classbits, &class_uchardata, options,              (void)add_not_list_to_class(classbits, &class_uchardata, options,
4775                cd, PRIV(hspace_list));                cd, PRIV(hspace_list));
4776              continue;              continue;
4777    
4778              case ESC_v:              case ESC_v:
4779              (void)add_list_to_class(classbits, &class_uchardata, options, cd,              (void)add_list_to_class(classbits, &class_uchardata, options, cd,
4780                PRIV(vspace_list), NOTACHAR);                PRIV(vspace_list), NOTACHAR);
4781              continue;              continue;
4782    
4783              case ESC_V:              case ESC_V:
4784              (void)add_not_list_to_class(classbits, &class_uchardata, options,              (void)add_not_list_to_class(classbits, &class_uchardata, options,
4785                cd, PRIV(vspace_list));                cd, PRIV(vspace_list));
4786              continue;              continue;
4787    
4788  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
# Line 4314  for (;; ptr++) Line 4790  for (;; ptr++)
4790              case ESC_P:              case ESC_P:
4791                {                {
4792                BOOL negated;                BOOL negated;
4793                int pdata;                unsigned int ptype = 0, pdata = 0;
4794                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                if (!get_ucp(&ptr, &negated, &ptype, &pdata, errorcodeptr))
4795                if (ptype < 0) goto FAILED;                  goto FAILED;
4796                *class_uchardata++ = ((-c == ESC_p) != negated)?                *class_uchardata++ = ((escape == ESC_p) != negated)?
4797                  XCL_PROP : XCL_NOTPROP;                  XCL_PROP : XCL_NOTPROP;
4798                *class_uchardata++ = ptype;                *class_uchardata++ = ptype;
4799                *class_uchardata++ = pdata;                *class_uchardata++ = pdata;
# Line 4344  for (;; ptr++) Line 4820  for (;; ptr++)
4820    
4821          /* Fall through if the escape just defined a single character (c >= 0).          /* Fall through if the escape just defined a single character (c >= 0).
4822          This may be greater than 256. */          This may be greater than 256. */
4823    
4824            escape = 0;
4825    
4826          }   /* End of backslash handling */          }   /* End of backslash handling */
4827    
4828        /* A character may be followed by '-' to form a range. However, Perl does        /* A character may be followed by '-' to form a range. However, Perl does
# Line 4368  for (;; ptr++) Line 4846  for (;; ptr++)
4846    
4847        if (!inescq && ptr[1] == CHAR_MINUS)        if (!inescq && ptr[1] == CHAR_MINUS)
4848          {          {
4849          int d;          pcre_uint32 d;
4850          ptr += 2;          ptr += 2;
4851          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
4852    
# Line 4383  for (;; ptr++) Line 4861  for (;; ptr++)
4861            inescq = TRUE;            inescq = TRUE;
4862            break;            break;
4863            }            }
4864    
4865          /* Minus (hyphen) at the end of a class is treated as a literal, so put          /* Minus (hyphen) at the end of a class is treated as a literal, so put
4866          back the pointer and jump to handle the character that preceded it. */          back the pointer and jump to handle the character that preceded it. */
4867    
4868          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))          if (*ptr == CHAR_NULL || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
4869            {            {
4870            ptr = oldptr;            ptr = oldptr;
4871            goto CLASS_SINGLE_CHARACTER;            goto CLASS_SINGLE_CHARACTER;
4872            }            }
4873    
4874          /* Otherwise, we have a potential range; pick up the next character */          /* Otherwise, we have a potential range; pick up the next character */
4875    
4876  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
# Line 4410  for (;; ptr++) Line 4888  for (;; ptr++)
4888    
4889          if (!inescq && d == CHAR_BACKSLASH)          if (!inescq && d == CHAR_BACKSLASH)
4890            {            {
4891            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            int descape;
4892              descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);
4893            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
4894    
4895            /* \b is backspace; any other special means the '-' was literal. */            /* \b is backspace; any other special means the '-' was literal. */
4896    
4897            if (d < 0)            if (descape != 0)
4898              {              {
4899              if (d == -ESC_b) d = CHAR_BS; else              if (descape == ESC_b) d = CHAR_BS; else
4900                {                {
4901                ptr = oldptr;                ptr = oldptr;
4902                goto CLASS_SINGLE_CHARACTER;  /* A few lines below */                goto CLASS_SINGLE_CHARACTER;  /* A few lines below */
# Line 4438  for (;; ptr++) Line 4917  for (;; ptr++)
4917          /* We have found a character range, so single character optimizations          /* We have found a character range, so single character optimizations
4918          cannot be done anymore. Any value greater than 1 indicates that there          cannot be done anymore. Any value greater than 1 indicates that there
4919          is more than one character. */          is more than one character. */
4920    
4921          class_one_char = 2;          class_one_char = 2;
4922    
4923          /* Remember an explicit \r or \n, and add the range to the class. */          /* Remember an explicit \r or \n, and add the range to the class. */
4924    
4925          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
4926    
4927          class_has_8bitchar +=          class_has_8bitchar +=
4928            add_to_class(classbits, &class_uchardata, options, cd, c, d);            add_to_class(classbits, &class_uchardata, options, cd, c, d);
4929    
4930          continue;   /* Go get the next char in the class */          continue;   /* Go get the next char in the class */
4931          }          }
4932    
4933        /* Handle a single character - we can get here for a normal non-escape        /* Handle a single character - we can get here for a normal non-escape
4934        char, or after \ that introduces a single character or for an apparent        char, or after \ that introduces a single character or for an apparent
4935        range that isn't. Only the value 1 matters for class_one_char, so don't        range that isn't. Only the value 1 matters for class_one_char, so don't
4936        increase it if it is already 2 or more ... just in case there's a class        increase it if it is already 2 or more ... just in case there's a class
4937        with a zillion characters in it. */        with a zillion characters in it. */
4938    
4939        CLASS_SINGLE_CHARACTER:        CLASS_SINGLE_CHARACTER:
# Line 4473  for (;; ptr++) Line 4952  for (;; ptr++)
4952          {          {
4953          ptr++;          ptr++;
4954          zeroreqchar = reqchar;          zeroreqchar = reqchar;
4955            zeroreqcharflags = reqcharflags;
4956    
4957          if (negate_class)          if (negate_class)
4958            {            {
4959            if (firstchar == REQ_UNSET) firstchar = REQ_NONE;  #ifdef SUPPORT_UCP
4960              int d;
4961    #endif
4962              if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4963            zerofirstchar = firstchar;            zerofirstchar = firstchar;
4964            *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;            zerofirstcharflags = firstcharflags;
4965  #ifdef SUPPORT_UTF  
4966            if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)            /* For caseless UTF-8 mode when UCP support is available, check
4967              code += PRIV(ord2utf)(c, code);            whether this character has more than one other case. If so, generate
4968              a special OP_NOTPROP item instead of OP_NOTI. */
4969    
4970    #ifdef SUPPORT_UCP
4971              if (utf && (options & PCRE_CASELESS) != 0 &&
4972                  (d = UCD_CASESET(c)) != 0)
4973                {
4974                *code++ = OP_NOTPROP;
4975                *code++ = PT_CLIST;
4976                *code++ = d;
4977                }
4978            else            else
4979  #endif  #endif
4980              *code++ = c;            /* Char has only one other case, or UCP not available */
4981            goto NOT_CHAR;  
4982                {
4983                *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4984    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
4985                if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)