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

Diff of /code/trunk/pcre_compile.c

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

revision 1067 by chpe, Tue Oct 16 15:54:22 2012 UTC revision 1396 by ph10, Sun Nov 10 19:04:34 2013 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2012 University of Cambridge             Copyright (c) 1997-2013 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 80  to check them every time. */ Line 80  to check them every time. */
80  /* Definitions to allow mutual recursion */  /* Definitions to allow mutual recursion */
81    
82  static int  static int
83    add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,    add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84      const pcre_uint32 *, unsigned int);      const pcre_uint32 *, unsigned int);
85    
86  static BOOL  static BOOL
87    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL,    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
88      int, int, int *, int *, branch_chain *, compile_data *, int *);      pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
89        compile_data *, int *);
90    
91    
92    
# Line 114  kicks in at the same number of forward r Line 115  kicks in at the same number of forward r
115  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
116  #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)  #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
117    
118    /* This value determines the size of the initial vector that is used for
119    remembering named groups during the pre-compile. It is allocated on the stack,
120    but if it is too small, it is expanded using malloc(), in a similar way to the
121    workspace. The value is the number of slots in the list. */
122    
123    #define NAMED_GROUP_LIST_SIZE  20
124    
125  /* The overrun tests check for a slightly smaller size so that they detect the  /* The overrun tests check for a slightly smaller size so that they detect the
126  overrun before it actually does run off the end of the data block. */  overrun before it actually does run off the end of the data block. */
127    
# Line 121  overrun before it actually does run off Line 129  overrun before it actually does run off
129    
130  /* Private flags added to firstchar and reqchar. */  /* Private flags added to firstchar and reqchar. */
131    
132  #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */  #define REQ_CASELESS    (1 << 0)        /* Indicates caselessness */
133  #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */  #define REQ_VARY        (1 << 1)        /* Reqchar followed non-literal item */
134  #define REQ_MASK       (REQ_CASELESS | REQ_VARY)  /* Negative values for the firstchar and reqchar flags */
135    #define REQ_UNSET       (-2)
136    #define REQ_NONE        (-1)
137    
138  /* Repeated character flags. */  /* Repeated character flags. */
139    
# Line 254  static const int verbcount = sizeof(verb Line 264  static const int verbcount = sizeof(verb
264  now all in a single string, to reduce the number of relocations when a shared  now all in a single string, to reduce the number of relocations when a shared
265  library is dynamically loaded. The list of lengths is terminated by a zero  library is dynamically loaded. The list of lengths is terminated by a zero
266  length entry. The first three must be alpha, lower, upper, as this is assumed  length entry. The first three must be alpha, lower, upper, as this is assumed
267  for handling case independence. */  for handling case independence. The indices for graph, print, and punct are
268    needed, so identify them. */
269    
270  static const char posix_names[] =  static const char posix_names[] =
271    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
# Line 265  static const char posix_names[] = Line 276  static const char posix_names[] =
276  static const pcre_uint8 posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
277    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
278    
279    #define PC_GRAPH  8
280    #define PC_PRINT  9
281    #define PC_PUNCT 10
282    
283    
284  /* Table of class bit maps for each POSIX class. Each class is formed from a  /* Table of class bit maps for each POSIX class. Each class is formed from a
285  base map, with an optional addition or removal of another map. Then, for some  base map, with an optional addition or removal of another map. Then, for some
286  classes, there is some additional tweaking: for [:blank:] the vertical space  classes, there is some additional tweaking: for [:blank:] the vertical space
# Line 292  static const int posix_class_maps[] = { Line 308  static const int posix_class_maps[] = {
308    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
309  };  };
310    
311  /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class  /* Table of substitutes for \d etc when PCRE_UCP is set. They are replaced by
312  substitutes must be in the order of the names, defined above, and there are  Unicode property escapes. */
 both positive and negative cases. NULL means no substitute. */  
313    
314  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
315  static const pcre_uchar string_PNd[]  = {  static const pcre_uchar string_PNd[]  = {
# Line 319  static const pcre_uchar string_pXwd[] = Line 334  static const pcre_uchar string_pXwd[] =
334  static const pcre_uchar *substitutes[] = {  static const pcre_uchar *substitutes[] = {
335    string_PNd,           /* \D */    string_PNd,           /* \D */
336    string_pNd,           /* \d */    string_pNd,           /* \d */
337    string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */    string_PXsp,          /* \S */   /* Xsp is Perl space, but from 8.34, Perl */
338    string_pXsp,          /* \s */    string_pXsp,          /* \s */   /* space and POSIX space are the same. */
339    string_PXwd,          /* \W */    string_PXwd,          /* \W */
340    string_pXwd           /* \w */    string_pXwd           /* \w */
341  };  };
342    
343    /* The POSIX class substitutes must be in the order of the POSIX class names,
344    defined above, and there are both positive and negative cases. NULL means no
345    general substitute of a Unicode property escape (\p or \P). However, for some
346    POSIX classes (e.g. graph, print, punct) a special property code is compiled
347    directly. */
348    
349  static const pcre_uchar string_pL[] =   {  static const pcre_uchar string_pL[] =   {
350    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
351    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
# Line 372  static const pcre_uchar *posix_substitut Line 393  static const pcre_uchar *posix_substitut
393    NULL,                 /* graph */    NULL,                 /* graph */
394    NULL,                 /* print */    NULL,                 /* print */
395    NULL,                 /* punct */    NULL,                 /* punct */
396    string_pXps,          /* space */    /* NOTE: Xps is POSIX space */    string_pXps,          /* space */   /* Xps is POSIX space, but from 8.34 */
397    string_pXwd,          /* word */    string_pXwd,          /* word  */   /* Perl and POSIX space are the same */
398    NULL,                 /* xdigit */    NULL,                 /* xdigit */
399    /* Negated cases */    /* Negated cases */
400    string_PL,            /* ^alpha */    string_PL,            /* ^alpha */
# Line 387  static const pcre_uchar *posix_substitut Line 408  static const pcre_uchar *posix_substitut
408    NULL,                 /* ^graph */    NULL,                 /* ^graph */
409    NULL,                 /* ^print */    NULL,                 /* ^print */
410    NULL,                 /* ^punct */    NULL,                 /* ^punct */
411    string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */    string_PXps,          /* ^space */  /* Xps is POSIX space, but from 8.34 */
412    string_PXwd,          /* ^word */    string_PXwd,          /* ^word */   /* Perl and POSIX space are the same */
413    NULL                  /* ^xdigit */    NULL                  /* ^xdigit */
414  };  };
415  #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))  #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
# Line 452  static const char error_texts[] = Line 473  static const char error_texts[] =
473    "POSIX collating elements are not supported\0"    "POSIX collating elements are not supported\0"
474    "this version of PCRE is compiled without UTF support\0"    "this version of PCRE is compiled without UTF support\0"
475    "spare error\0"  /** DEAD **/    "spare error\0"  /** DEAD **/
476    "character value in \\x{...} sequence is too large\0"    "character value in \\x{} or \\o{} is too large\0"
477    /* 35 */    /* 35 */
478    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
479    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
# Line 484  static const char error_texts[] = Line 505  static const char error_texts[] =
505    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
506    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
507    /* 60 */    /* 60 */
508    "(*VERB) not recognized\0"    "(*VERB) not recognized or malformed\0"
509    "number is too big\0"    "number is too big\0"
510    "subpattern name expected\0"    "subpattern name expected\0"
511    "digit expected after (?+\0"    "digit expected after (?+\0"
# Line 505  static const char error_texts[] = Line 526  static const char error_texts[] =
526    "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"    "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
527    "character value in \\u.... sequence is too large\0"    "character value in \\u.... sequence is too large\0"
528    "invalid UTF-32 string\0"    "invalid UTF-32 string\0"
529      "setting UTF is disabled by the application\0"
530      "non-hex character in \\x{} (closing brace missing?)\0"
531      /* 80 */
532      "non-octal character in \\o{} (closing brace missing?)\0"
533      "missing opening brace after \\o\0"
534      "parentheses are too deeply nested\0"
535      "invalid range in character class\0"
536      "group name must start with a non-digit\0"
537    ;    ;
538    
539  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 644  static const pcre_uint8 ebcdic_chartab[] Line 673  static const pcre_uint8 ebcdic_chartab[]
673  #endif  #endif
674    
675    
676    /* This table is used to check whether auto-possessification is possible
677    between adjacent character-type opcodes. The left-hand (repeated) opcode is
678    used to select the row, and the right-hand opcode is use to select the column.
679    A value of 1 means that auto-possessification is OK. For example, the second
680    value in the first row means that \D+\d can be turned into \D++\d.
681    
682    The Unicode property types (\P and \p) have to be present to fill out the table
683    because of what their opcode values are, but the table values should always be
684    zero because property types are handled separately in the code. The last four
685    columns apply to items that cannot be repeated, so there is no need to have
686    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
687    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
688    
689    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
690    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
691    
692    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
693    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
694      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
695      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
696      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
697      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
698      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
699      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
700      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
701      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
702      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
703      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
704      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
705      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
706      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
707      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
708      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
709      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
710      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
711    };
712    
713    
714    /* This table is used to check whether auto-possessification is possible
715    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
716    left-hand (repeated) opcode is used to select the row, and the right-hand
717    opcode is used to select the column. The values are as follows:
718    
719      0   Always return FALSE (never auto-possessify)
720      1   Character groups are distinct (possessify if both are OP_PROP)
721      2   Check character categories in the same group (general or particular)
722      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
723    
724      4   Check left general category vs right particular category
725      5   Check right general category vs left particular category
726    
727      6   Left alphanum vs right general category
728      7   Left space vs right general category
729      8   Left word vs right general category
730    
731      9   Right alphanum vs left general category
732     10   Right space vs left general category
733     11   Right word vs left general category
734    
735     12   Left alphanum vs right particular category
736     13   Left space vs right particular category
737     14   Left word vs right particular category
738    
739     15   Right alphanum vs left particular category
740     16   Right space vs left particular category
741     17   Right word vs left particular category
742    */
743    
744    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
745    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
746      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
747      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
748      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
749      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
750      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
751      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
752      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
753      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
754      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
755      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
756      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
757    };
758    
759    /* This table is used to check whether auto-possessification is possible
760    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
761    specifies a general category and the other specifies a particular category. The
762    row is selected by the general category and the column by the particular
763    category. The value is 1 if the particular category is not part of the general
764    category. */
765    
766    static const pcre_uint8 catposstab[7][30] = {
767    /* 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 */
768      { 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 */
769      { 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 */
770      { 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 */
771      { 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 */
772      { 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 */
773      { 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 */
774      { 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 */
775    };
776    
777    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
778    a general or particular category. The properties in each row are those
779    that apply to the character set in question. Duplication means that a little
780    unnecessary work is done when checking, but this keeps things much simpler
781    because they can all use the same code. For more details see the comment where
782    this table is used.
783    
784    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
785    "space", but from Perl 5.18 it's included, so both categories are treated the
786    same here. */
787    
788    static const pcre_uint8 posspropstab[3][4] = {
789      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
790      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
791      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
792    };
793    
794    /* This table is used when converting repeating opcodes into possessified
795    versions as a result of an explicit possessive quantifier such as ++. A zero
796    value means there is no possessified version - in those cases the item in
797    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
798    because all relevant opcodes are less than that. */
799    
800    static const pcre_uint8 opcode_possessify[] = {
801      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
802      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
803    
804      0,                       /* NOTI */
805      OP_POSSTAR, 0,           /* STAR, MINSTAR */
806      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
807      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
808      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
809      0,                       /* EXACT */
810      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
811    
812      OP_POSSTARI, 0,          /* STARI, MINSTARI */
813      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
814      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
815      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
816      0,                       /* EXACTI */
817      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
818    
819      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
820      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
821      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
822      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
823      0,                       /* NOTEXACT */
824      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
825    
826      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
827      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
828      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
829      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
830      0,                       /* NOTEXACTI */
831      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
832    
833      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
834      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
835      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
836      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
837      0,                       /* TYPEEXACT */
838      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
839    
840      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
841      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
842      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
843      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
844      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
845    
846      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
847      0, 0,                    /* REF, REFI */
848      0, 0,                    /* DNREF, DNREFI */
849      0, 0                     /* RECURSE, CALLOUT */
850    };
851    
852    
853    
854  /*************************************************  /*************************************************
# Line 665  find_error_text(int n) Line 870  find_error_text(int n)
870  const char *s = error_texts;  const char *s = error_texts;
871  for (; n > 0; n--)  for (; n > 0; n--)
872    {    {
873    while (*s++ != 0) {};    while (*s++ != CHAR_NULL) {};
874    if (*s == 0) return "Error text not found (please report)";    if (*s == CHAR_NULL) return "Error text not found (please report)";
875    }    }
876  return s;  return s;
877  }  }
878    
879    
880    
881  /*************************************************  /*************************************************
882  *           Expand the workspace                 *  *           Expand the workspace                 *
883  *************************************************/  *************************************************/
# Line 749  return (*p == CHAR_RIGHT_CURLY_BRACKET); Line 955  return (*p == CHAR_RIGHT_CURLY_BRACKET);
955  *************************************************/  *************************************************/
956    
957  /* 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
958  positive value for a simple escape such as \n, or 0 for a data character  positive value for a simple escape such as \n, or 0 for a data character which
959  which will be placed in chptr. A backreference to group n is returned as  will be placed in chptr. A backreference to group n is returned as negative n.
960  negative n. When UTF-8 is enabled, a positive value greater than 255 may  When UTF-8 is enabled, a positive value greater than 255 may be returned in
961  be returned in chptr.  chptr. On entry, ptr is pointing at the \. On exit, it is on the final
962  On entry,ptr is pointing at the \. On exit, it is on the final character of the  character of the escape sequence.
 escape sequence.  
963    
964  Arguments:  Arguments:
965    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
966    chptr          points to the data character    chptr          points to a returned data character
967    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
968    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
969    options        the options bits    options        the options bits
# Line 771  Returns:         zero => a data characte Line 976  Returns:         zero => a data characte
976  */  */
977    
978  static int  static int
979  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
980    int bracount, int options, BOOL isclass)    int bracount, int options, BOOL isclass)
981  {  {
982  /* PCRE_UTF16 has the same value as PCRE_UTF8. */  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
# Line 786  ptr--;                            /* Set Line 991  ptr--;                            /* Set
991    
992  /* 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. */
993    
994  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
995    
996  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
997  in a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
# Line 795  Otherwise further processing may be requ Line 1000  Otherwise further processing may be requ
1000  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1001  /* Not alphanumeric */  /* Not alphanumeric */
1002  else if (c < CHAR_0 || c > CHAR_z) {}  else if (c < CHAR_0 || c > CHAR_z) {}
1003  else if ((i = escapes[c - CHAR_0]) != 0) { if (i > 0) c = (pcre_uint32)i; else escape = -i; }  else if ((i = escapes[c - CHAR_0]) != 0)
1004      { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1005    
1006  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1007  /* Not alphanumeric */  /* Not alphanumeric */
# Line 845  else Line 1051  else
1051            }            }
1052    
1053  #if defined COMPILE_PCRE8  #if defined COMPILE_PCRE8
1054          if (c > (utf ? 0x10ffff : 0xff))          if (c > (utf ? 0x10ffffU : 0xffU))
1055  #elif defined COMPILE_PCRE16  #elif defined COMPILE_PCRE16
1056          if (c > (utf ? 0x10ffff : 0xffff))          if (c > (utf ? 0x10ffffU : 0xffffU))
1057  #elif defined COMPILE_PCRE32  #elif defined COMPILE_PCRE32
1058          if (utf && c > 0x10ffff)          if (utf && c > 0x10ffffU)
1059  #endif  #endif
1060            {            {
1061            *errorcodeptr = ERR76;            *errorcodeptr = ERR76;
# Line 896  else Line 1102  else
1102      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1103        {        {
1104        const pcre_uchar *p;        const pcre_uchar *p;
1105        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++)
1106          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1107        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)        if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1108          {          {
1109          escape = ESC_k;          escape = ESC_k;
1110          break;          break;
# Line 961  else Line 1167  else
1167      break;      break;
1168    
1169      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1170      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
1171      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1172        recommended to avoid the ambiguities in the old syntax.
1173    
1174      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
1175      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
1176      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
1177      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
1178      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
1179      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
1180      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1181    
1182        Inside a character class, \ followed by a digit is always either a literal
1183        8 or 9 or an octal number. */
1184    
1185      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:
1186      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
# Line 997  else Line 1207  else
1207          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1208          break;          break;
1209          }          }
1210        if (s < 10 || s <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1211          {          {
1212          escape = -s;          escape = -s;
1213          break;          break;
# Line 1005  else Line 1215  else
1215        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1216        }        }
1217    
1218      /* 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
1219      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
1220      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
1221        changed so as not to insert the binary zero. */
1222    
1223      if ((c = *ptr) >= CHAR_8)      if ((c = *ptr) >= CHAR_8) break;
1224        {  
1225        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1226    
1227      /* \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
1228      larger first octal digit. The original code used just to take the least      larger first octal digit. The original code used just to take the least
# Line 1031  else Line 1239  else
1239  #endif  #endif
1240      break;      break;
1241    
1242      /* \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
1243      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}. */
1244      If not, { is treated as a data character. */  
1245        case CHAR_o:
1246        if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1247          {
1248          ptr += 2;
1249          c = 0;
1250          overflow = FALSE;
1251          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1252            {
1253            register pcre_uint32 cc = *ptr++;
1254            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1255    #ifdef COMPILE_PCRE32
1256            if (c >= 0x20000000l) { overflow = TRUE; break; }
1257    #endif
1258            c = (c << 3) + cc - CHAR_0 ;
1259    #if defined COMPILE_PCRE8
1260            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1261    #elif defined COMPILE_PCRE16
1262            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1263    #elif defined COMPILE_PCRE32
1264            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1265    #endif
1266            }
1267          if (overflow)
1268            {
1269            while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1270            *errorcodeptr = ERR34;
1271            }
1272          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1273            {
1274            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1275            }
1276          else *errorcodeptr = ERR80;
1277          }
1278        break;
1279    
1280        /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1281        numbers. Otherwise it is a lowercase x letter. */
1282    
1283      case CHAR_x:      case CHAR_x:
1284      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1285        {        {
       /* In JavaScript, \x must be followed by two hexadecimal numbers.  
       Otherwise it is a lowercase x letter. */  
1286        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1287          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1288          {          {
# Line 1056  else Line 1299  else
1299  #endif  #endif
1300            }            }
1301          }          }
1302        break;        }    /* End JavaScript handling */
       }  
1303    
1304      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1305        {      greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1306        const pcre_uchar *pt = ptr + 2;      digits. If not, { used to be treated as a data character. However, Perl
1307        seems to read hex digits up to the first non-such, and ignore the rest, so
1308        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1309        now gives an error. */
1310    
1311        c = 0;      else
1312        overflow = FALSE;        {
1313        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1314          {          {
1315          register pcre_uint32 cc = *pt++;          ptr += 2;
1316          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          c = 0;
1317            overflow = FALSE;
1318            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1319              {
1320              register pcre_uint32 cc = *ptr++;
1321              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1322    
1323  #ifdef COMPILE_PCRE32  #ifdef COMPILE_PCRE32
1324          if (c >= 0x10000000l) { overflow = TRUE; break; }            if (c >= 0x10000000l) { overflow = TRUE; break; }
1325  #endif  #endif
1326    
1327  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1328          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1329          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1330  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1331          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 */
1332          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1333  #endif  #endif
1334    
1335  #if defined COMPILE_PCRE8  #if defined COMPILE_PCRE8
1336          if (c > (utf ? 0x10ffff : 0xff)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1337  #elif defined COMPILE_PCRE16  #elif defined COMPILE_PCRE16
1338          if (c > (utf ? 0x10ffff : 0xffff)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1339  #elif defined COMPILE_PCRE32  #elif defined COMPILE_PCRE32
1340          if (utf && c > 0x10ffff) { overflow = TRUE; break; }            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1341  #endif  #endif
1342          }            }
1343    
1344        if (overflow)          if (overflow)
1345          {            {
1346          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1347          *errorcodeptr = ERR34;            *errorcodeptr = ERR34;
1348          }            }
1349    
1350        if (*pt == CHAR_RIGHT_CURLY_BRACKET)          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1351          {            {
1352          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1353          ptr = pt;            }
         break;  
         }  
1354    
1355        /* 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.
1356        recognize this construct; fall through to the normal \x handling. */          We used just to recognize this construct and fall through to the normal
1357        }          \x handling, but nowadays Perl gives an error, which seems much more
1358            sensible, so we do too. */
1359    
1360      /* Read just a single-byte hex-defined char */          else *errorcodeptr = ERR79;
1361            }   /* End of \x{} processing */
1362    
1363      c = 0;        /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1364      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)  
1365        {        else
1366        pcre_uint32 cc;                          /* Some compilers don't like */          {
1367        cc = *(++ptr);                           /* ++ in initializers */          c = 0;
1368            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1369              {
1370              pcre_uint32 cc;                          /* Some compilers don't like */
1371              cc = *(++ptr);                           /* ++ in initializers */
1372  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1373        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1374        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1375  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1376        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */            if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1377        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1378  #endif  #endif
1379        }            }
1380            }     /* End of \xdd handling */
1381          }       /* End of Perl-style \x handling */
1382      break;      break;
1383    
1384      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
# Line 1132  else Line 1388  else
1388    
1389      case CHAR_c:      case CHAR_c:
1390      c = *(++ptr);      c = *(++ptr);
1391      if (c == 0)      if (c == CHAR_NULL)
1392        {        {
1393        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1394        break;        break;
# Line 1188  if ((options & PCRE_UCP) != 0 && escape Line 1444  if ((options & PCRE_UCP) != 0 && escape
1444  return escape;  return escape;
1445  }  }
1446    
1447    
1448    
1449  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
1450  /*************************************************  /*************************************************
1451  *               Handle \P and \p                 *  *               Handle \P and \p                 *
# Line 1201  escape sequence. Line 1459  escape sequence.
1459  Argument:  Argument:
1460    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1461    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
1462    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
1463      pdataptr       points to an unsigned int that is set to the detailed property value
1464    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1465    
1466  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
1467  */  */
1468    
1469  static int  static BOOL
1470  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1471      unsigned int *pdataptr, int *errorcodeptr)
1472  {  {
1473  pcre_uchar c;  pcre_uchar c;
1474  int i, bot, top;  int i, bot, top;
# Line 1216  const pcre_uchar *ptr = *ptrptr; Line 1476  const pcre_uchar *ptr = *ptrptr;
1476  pcre_uchar name[32];  pcre_uchar name[32];
1477    
1478  c = *(++ptr);  c = *(++ptr);
1479  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1480    
1481  *negptr = FALSE;  *negptr = FALSE;
1482    
# Line 1233  if (c == CHAR_LEFT_CURLY_BRACKET) Line 1493  if (c == CHAR_LEFT_CURLY_BRACKET)
1493    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1494      {      {
1495      c = *(++ptr);      c = *(++ptr);
1496      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1497      if (c == CHAR_RIGHT_CURLY_BRACKET) break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1498      name[i] = c;      name[i] = c;
1499      }      }
# Line 1263  while (bot < top) Line 1523  while (bot < top)
1523    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1524    if (r == 0)    if (r == 0)
1525      {      {
1526      *dptr = PRIV(utt)[i].value;      *ptypeptr = PRIV(utt)[i].type;
1527      return PRIV(utt)[i].type;      *pdataptr = PRIV(utt)[i].value;
1528        return TRUE;
1529      }      }
1530    if (r > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1531    }    }
1532    
1533  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1534  *ptrptr = ptr;  *ptrptr = ptr;
1535  return -1;  return FALSE;
1536    
1537  ERROR_RETURN:  ERROR_RETURN:
1538  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1539  *ptrptr = ptr;  *ptrptr = ptr;
1540  return -1;  return FALSE;
1541  }  }
1542  #endif  #endif
1543    
1544    
1545    
   
1546  /*************************************************  /*************************************************
1547  *         Read repeat counts                     *  *         Read repeat counts                     *
1548  *************************************************/  *************************************************/
# Line 1351  return p; Line 1611  return p;
1611    
1612    
1613  /*************************************************  /*************************************************
 *  Subroutine for finding forward reference      *  
 *************************************************/  
   
 /* This recursive function is called only from find_parens() below. The  
 top-level call starts at the beginning of the pattern. All other calls must  
 start at a parenthesis. It scans along a pattern's text looking for capturing  
 subpatterns, and counting them. If it finds a named pattern that matches the  
 name it is given, it returns its number. Alternatively, if the name is NULL, it  
 returns when it reaches a given numbered subpattern. Recursion is used to keep  
 track of subpatterns that reset the capturing group numbers - the (?| feature.  
   
 This function was originally called only from the second pass, in which we know  
 that if (?< or (?' or (?P< is encountered, the name will be correctly  
 terminated because that is checked in the first pass. There is now one call to  
 this function in the first pass, to check for a recursive back reference by  
 name (so that we can make the whole group atomic). In this case, we need check  
 only up to the current position in the pattern, and that is still OK because  
 and previous occurrences will have been checked. To make this work, the test  
 for "end of pattern" is a check against cd->end_pattern in the main loop,  
 instead of looking for a binary zero. This means that the special first-pass  
 call can adjust cd->end_pattern temporarily. (Checks for binary zero while  
 processing items within the loop are OK, because afterwards the main loop will  
 terminate.)  
   
 Arguments:  
   ptrptr       address of the current character pointer (updated)  
   cd           compile background data  
   name         name to seek, or NULL if seeking a numbered subpattern  
   lorn         name length, or subpattern number if name is NULL  
   xmode        TRUE if we are in /x mode  
   utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode  
   count        pointer to the current capturing subpattern number (updated)  
   
 Returns:       the number of the named subpattern, or -1 if not found  
 */  
   
 static int  
 find_parens_sub(pcre_uchar **ptrptr, compile_data *cd, const pcre_uchar *name, int lorn,  
   BOOL xmode, BOOL utf, int *count)  
 {  
 pcre_uchar *ptr = *ptrptr;  
 int start_count = *count;  
 int hwm_count = start_count;  
 BOOL dup_parens = FALSE;  
   
 /* If the first character is a parenthesis, check on the type of group we are  
 dealing with. The very first call may not start with a parenthesis. */  
   
 if (ptr[0] == CHAR_LEFT_PARENTHESIS)  
   {  
   /* Handle specials such as (*SKIP) or (*UTF8) etc. */  
   
   if (ptr[1] == CHAR_ASTERISK) ptr += 2;  
   
   /* Handle a normal, unnamed capturing parenthesis. */  
   
   else if (ptr[1] != CHAR_QUESTION_MARK)  
     {  
     *count += 1;  
     if (name == NULL && *count == lorn) return *count;  
     ptr++;  
     }  
   
   /* All cases now have (? at the start. Remember when we are in a group  
   where the parenthesis numbers are duplicated. */  
   
   else if (ptr[2] == CHAR_VERTICAL_LINE)  
     {  
     ptr += 3;  
     dup_parens = TRUE;  
     }  
   
   /* Handle comments; all characters are allowed until a ket is reached. */  
   
   else if (ptr[2] == CHAR_NUMBER_SIGN)  
     {  
     for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;  
     goto FAIL_EXIT;  
     }  
   
   /* Handle a condition. If it is an assertion, just carry on so that it  
   is processed as normal. If not, skip to the closing parenthesis of the  
   condition (there can't be any nested parens). */  
   
   else if (ptr[2] == CHAR_LEFT_PARENTHESIS)  
     {  
     ptr += 2;  
     if (ptr[1] != CHAR_QUESTION_MARK)  
       {  
       while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;  
       if (*ptr != 0) ptr++;  
       }  
     }  
   
   /* Start with (? but not a condition. */  
   
   else  
     {  
     ptr += 2;  
     if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */  
   
     /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */  
   
     if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&  
         ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)  
       {  
       pcre_uchar term;  
       const pcre_uchar *thisname;  
       *count += 1;  
       if (name == NULL && *count == lorn) return *count;  
       term = *ptr++;  
       if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;  
       thisname = ptr;  
       while (*ptr != term) ptr++;  
       if (name != NULL && lorn == (int)(ptr - thisname) &&  
           STRNCMP_UC_UC(name, thisname, (unsigned int)lorn) == 0)  
         return *count;  
       term++;  
       }  
     }  
   }  
   
 /* Past any initial parenthesis handling, scan for parentheses or vertical  
 bars. Stop if we get to cd->end_pattern. Note that this is important for the  
 first-pass call when this value is temporarily adjusted to stop at the current  
 position. So DO NOT change this to a test for binary zero. */  
   
 for (; ptr < cd->end_pattern; ptr++)  
   {  
   /* Skip over backslashed characters and also entire \Q...\E */  
   
   if (*ptr == CHAR_BACKSLASH)  
     {  
     if (*(++ptr) == 0) goto FAIL_EXIT;  
     if (*ptr == CHAR_Q) for (;;)  
       {  
       while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};  
       if (*ptr == 0) goto FAIL_EXIT;  
       if (*(++ptr) == CHAR_E) break;  
       }  
     continue;  
     }  
   
   /* Skip over character classes; this logic must be similar to the way they  
   are handled for real. If the first character is '^', skip it. Also, if the  
   first few characters (either before or after ^) are \Q\E or \E we skip them  
   too. This makes for compatibility with Perl. Note the use of STR macros to  
   encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */  
   
   if (*ptr == CHAR_LEFT_SQUARE_BRACKET)  
     {  
     BOOL negate_class = FALSE;  
     for (;;)  
       {  
       if (ptr[1] == CHAR_BACKSLASH)  
         {  
         if (ptr[2] == CHAR_E)  
           ptr+= 2;  
         else if (STRNCMP_UC_C8(ptr + 2,  
                  STR_Q STR_BACKSLASH STR_E, 3) == 0)  
           ptr += 4;  
         else  
           break;  
         }  
       else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)  
         {  
         negate_class = TRUE;  
         ptr++;  
         }  
       else break;  
       }  
   
     /* If the next character is ']', it is a data character that must be  
     skipped, except in JavaScript compatibility mode. */  
   
     if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&  
         (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)  
       ptr++;  
   
     while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)  
       {  
       if (*ptr == 0) return -1;  
       if (*ptr == CHAR_BACKSLASH)  
         {  
         if (*(++ptr) == 0) goto FAIL_EXIT;  
         if (*ptr == CHAR_Q) for (;;)  
           {  
           while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};  
           if (*ptr == 0) goto FAIL_EXIT;  
           if (*(++ptr) == CHAR_E) break;  
           }  
         continue;  
         }  
       }  
     continue;  
     }  
   
   /* Skip comments in /x mode */  
   
   if (xmode && *ptr == CHAR_NUMBER_SIGN)  
     {  
     ptr++;  
     while (*ptr != 0)  
       {  
       if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }  
       ptr++;  
 #ifdef SUPPORT_UTF  
       if (utf) FORWARDCHAR(ptr);  
 #endif  
       }  
     if (*ptr == 0) goto FAIL_EXIT;  
     continue;  
     }  
   
   /* Check for the special metacharacters */  
   
   if (*ptr == CHAR_LEFT_PARENTHESIS)  
     {  
     int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);  
     if (rc > 0) return rc;  
     if (*ptr == 0) goto FAIL_EXIT;  
     }  
   
   else if (*ptr == CHAR_RIGHT_PARENTHESIS)  
     {  
     if (dup_parens && *count < hwm_count) *count = hwm_count;  
     goto FAIL_EXIT;  
     }  
   
   else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)  
     {  
     if (*count > hwm_count) hwm_count = *count;  
     *count = start_count;  
     }  
   }  
   
 FAIL_EXIT:  
 *ptrptr = ptr;  
 return -1;  
 }  
   
   
   
   
 /*************************************************  
 *       Find forward referenced subpattern       *  
 *************************************************/  
   
 /* This function scans along a pattern's text looking for capturing  
 subpatterns, and counting them. If it finds a named pattern that matches the  
 name it is given, it returns its number. Alternatively, if the name is NULL, it  
 returns when it reaches a given numbered subpattern. This is used for forward  
 references to subpatterns. We used to be able to start this scan from the  
 current compiling point, using the current count value from cd->bracount, and  
 do it all in a single loop, but the addition of the possibility of duplicate  
 subpattern numbers means that we have to scan from the very start, in order to  
 take account of such duplicates, and to use a recursive function to keep track  
 of the different types of group.  
   
 Arguments:  
   cd           compile background data  
   name         name to seek, or NULL if seeking a numbered subpattern  
   lorn         name length, or subpattern number if name is NULL  
   xmode        TRUE if we are in /x mode  
   utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode  
   
 Returns:       the number of the found subpattern, or -1 if not found  
 */  
   
 static int  
 find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,  
   BOOL utf)  
 {  
 pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;  
 int count = 0;  
 int rc;  
   
 /* If the pattern does not start with an opening parenthesis, the first call  
 to find_parens_sub() will scan right to the end (if necessary). However, if it  
 does start with a parenthesis, find_parens_sub() will return when it hits the  
 matching closing parens. That is why we have to have a loop. */  
   
 for (;;)  
   {  
   rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);  
   if (rc > 0 || *ptr++ == 0) break;  
   }  
   
 return rc;  
 }  
   
   
   
   
 /*************************************************  
1614  *      Find first significant op code            *  *      Find first significant op code            *
1615  *************************************************/  *************************************************/
1616    
# Line 1684  for (;;) Line 1649  for (;;)
1649    
1650      case OP_CALLOUT:      case OP_CALLOUT:
1651      case OP_CREF:      case OP_CREF:
1652      case OP_NCREF:      case OP_DNCREF:
1653      case OP_RREF:      case OP_RREF:
1654      case OP_NRREF:      case OP_DNRREF:
1655      case OP_DEF:      case OP_DEF:
1656      code += PRIV(OP_lengths)[*code];      code += PRIV(OP_lengths)[*code];
1657      break;      break;
# Line 1700  for (;;) Line 1665  for (;;)
1665    
1666    
1667    
   
1668  /*************************************************  /*************************************************
1669  *        Find the fixed length of a branch       *  *        Find the fixed length of a branch       *
1670  *************************************************/  *************************************************/
# Line 1744  for (;;) Line 1708  for (;;)
1708    {    {
1709    int d;    int d;
1710    pcre_uchar *ce, *cs;    pcre_uchar *ce, *cs;
1711    register int op = *cc;    register pcre_uchar op = *cc;
1712    
1713    switch (op)    switch (op)
1714      {      {
# Line 1824  for (;;) Line 1788  for (;;)
1788      case OP_COMMIT:      case OP_COMMIT:
1789      case OP_CREF:      case OP_CREF:
1790      case OP_DEF:      case OP_DEF:
1791        case OP_DNCREF:
1792        case OP_DNRREF:
1793      case OP_DOLL:      case OP_DOLL:
1794      case OP_DOLLM:      case OP_DOLLM:
1795      case OP_EOD:      case OP_EOD:
1796      case OP_EODN:      case OP_EODN:
1797      case OP_FAIL:      case OP_FAIL:
     case OP_NCREF:  
     case OP_NRREF:  
1798      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1799      case OP_PRUNE:      case OP_PRUNE:
1800      case OP_REVERSE:      case OP_REVERSE:
# Line 1852  for (;;) Line 1816  for (;;)
1816      case OP_NOTI:      case OP_NOTI:
1817      branchlength++;      branchlength++;
1818      cc += 2;      cc += 2;
1819  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #ifdef SUPPORT_UTF
1820      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1821  #endif  #endif
1822      break;      break;
# Line 1864  for (;;) Line 1828  for (;;)
1828      case OP_EXACTI:      case OP_EXACTI:
1829      case OP_NOTEXACT:      case OP_NOTEXACT:
1830      case OP_NOTEXACTI:      case OP_NOTEXACTI:
1831      branchlength += GET2(cc,1);      branchlength += (int)GET2(cc,1);
1832      cc += 2 + IMM2_SIZE;      cc += 2 + IMM2_SIZE;
1833  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #ifdef SUPPORT_UTF
1834      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1835  #endif  #endif
1836      break;      break;
1837    
1838      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1839      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1840      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1841        cc += 2;        cc += 2;
1842      cc += 1 + IMM2_SIZE + 1;      cc += 1 + IMM2_SIZE + 1;
1843      break;      break;
# Line 1909  for (;;) Line 1873  for (;;)
1873    
1874      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1875    
 #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32  
     case OP_XCLASS:  
     cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];  
     /* Fall through */  
 #endif  
   
1876      case OP_CLASS:      case OP_CLASS:
1877      case OP_NCLASS:      case OP_NCLASS:
1878    #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1879        case OP_XCLASS:
1880        /* The original code caused an unsigned overflow in 64 bit systems,
1881        so now we use a conditional statement. */
1882        if (op == OP_XCLASS)
1883          cc += GET(cc, 1);
1884        else
1885          cc += PRIV(OP_lengths)[OP_CLASS];
1886    #else
1887      cc += PRIV(OP_lengths)[OP_CLASS];      cc += PRIV(OP_lengths)[OP_CLASS];
1888    #endif
1889    
1890      switch (*cc)      switch (*cc)
1891        {        {
       case OP_CRPLUS:  
       case OP_CRMINPLUS:  
1892        case OP_CRSTAR:        case OP_CRSTAR:
1893        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1894          case OP_CRPLUS:
1895          case OP_CRMINPLUS:
1896        case OP_CRQUERY:        case OP_CRQUERY:
1897        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1898          case OP_CRPOSSTAR:
1899          case OP_CRPOSPLUS:
1900          case OP_CRPOSQUERY:
1901        return -1;        return -1;
1902    
1903        case OP_CRRANGE:        case OP_CRRANGE:
1904        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1905          case OP_CRPOSRANGE:
1906        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1907        branchlength += GET2(cc,1);        branchlength += (int)GET2(cc,1);
1908        cc += 1 + 2 * IMM2_SIZE;        cc += 1 + 2 * IMM2_SIZE;
1909        break;        break;
1910    
# Line 1999  for (;;) Line 1971  for (;;)
1971      case OP_QUERYI:      case OP_QUERYI:
1972      case OP_REF:      case OP_REF:
1973      case OP_REFI:      case OP_REFI:
1974        case OP_DNREF:
1975        case OP_DNREFI:
1976      case OP_SBRA:      case OP_SBRA:
1977      case OP_SBRAPOS:      case OP_SBRAPOS:
1978      case OP_SCBRA:      case OP_SCBRA:
# Line 2035  for (;;) Line 2009  for (;;)
2009    
2010    
2011    
   
2012  /*************************************************  /*************************************************
2013  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
2014  *************************************************/  *************************************************/
# Line 2059  PRIV(find_bracket)(const pcre_uchar *cod Line 2032  PRIV(find_bracket)(const pcre_uchar *cod
2032  {  {
2033  for (;;)  for (;;)
2034    {    {
2035    register int c = *code;    register pcre_uchar c = *code;
2036    
2037    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2038    
# Line 2082  for (;;) Line 2055  for (;;)
2055    else if (c == OP_CBRA || c == OP_SCBRA ||    else if (c == OP_CBRA || c == OP_SCBRA ||
2056             c == OP_CBRAPOS || c == OP_SCBRAPOS)             c == OP_CBRAPOS || c == OP_SCBRAPOS)
2057      {      {
2058      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
2059      if (n == number) return (pcre_uchar *)code;      if (n == number) return (pcre_uchar *)code;
2060      code += PRIV(OP_lengths)[c];      code += PRIV(OP_lengths)[c];
2061      }      }
# Line 2112  for (;;) Line 2085  for (;;)
2085        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2086        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2087        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2088        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2089          code += 2;          code += 2;
2090        break;        break;
2091    
2092        case OP_MARK:        case OP_MARK:
2093        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2094        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2095        case OP_THEN_ARG:        case OP_THEN_ARG:
2096        code += code[1];        code += code[1];
2097        break;        break;
# Line 2197  find_recurse(const pcre_uchar *code, BOO Line 2167  find_recurse(const pcre_uchar *code, BOO
2167  {  {
2168  for (;;)  for (;;)
2169    {    {
2170    register int c = *code;    register pcre_uchar c = *code;
2171    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2172    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2173    
# Line 2232  for (;;) Line 2202  for (;;)
2202        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2203        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2204        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2205        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2206          code += 2;          code += 2;
2207        break;        break;
2208    
2209        case OP_MARK:        case OP_MARK:
2210        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2211        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2212        case OP_THEN_ARG:        case OP_THEN_ARG:
2213        code += code[1];        code += code[1];
2214        break;        break;
# Line 2343  Arguments: Line 2310  Arguments:
2310    endcode     points to where to stop    endcode     points to where to stop
2311    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2312    cd          contains pointers to tables etc.    cd          contains pointers to tables etc.
2313      recurses    chain of recurse_check to catch mutual recursion
2314    
2315  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2316  */  */
2317    
2318    typedef struct recurse_check {
2319      struct recurse_check *prev;
2320      const pcre_uchar *group;
2321    } recurse_check;
2322    
2323  static BOOL  static BOOL
2324  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2325    BOOL utf, compile_data *cd)    BOOL utf, compile_data *cd, recurse_check *recurses)
2326  {  {
2327  register int c;  register pcre_uchar c;
2328    recurse_check this_recurse;
2329    
2330  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2331       code < endcode;       code < endcode;
2332       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
# Line 2379  for (code = first_significant_code(code Line 2354  for (code = first_significant_code(code
2354    
2355    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2356      {      {
2357      const pcre_uchar *scode;      const pcre_uchar *scode = cd->start_code + GET(code, 1);
2358      BOOL empty_branch;      BOOL empty_branch;
2359    
2360      /* Test for forward reference */      /* Test for forward reference or uncompleted reference. This is disabled
2361        when called to scan a completed pattern by setting cd->start_workspace to
2362        NULL. */
2363    
2364        if (cd->start_workspace != NULL)
2365          {
2366          const pcre_uchar *tcode;
2367          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2368            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2369          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2370          }
2371    
2372        /* If we are scanning a completed pattern, there are no forward references
2373        and all groups are complete. We need to detect whether this is a recursive
2374        call, as otherwise there will be an infinite loop. If it is a recursion,
2375        just skip over it. Simple recursions are easily detected. For mutual
2376        recursions we keep a chain on the stack. */
2377    
2378      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)      else
2379        if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;        {
2380          recurse_check *r = recurses;
2381          const pcre_uchar *endgroup = scode;
2382    
2383      /* Not a forward reference, test for completed backward reference */        do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2384          if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2385    
2386      empty_branch = FALSE;        for (r = recurses; r != NULL; r = r->prev)
2387      scode = cd->start_code + GET(code, 1);          if (r->group == scode) break;
2388      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */        if (r != NULL) continue;   /* Mutual recursion */
2389          }
2390    
2391        /* Completed reference; scan the referenced group, remembering it on the
2392        stack chain to detect mutual recursions. */
2393    
2394      /* Completed backwards reference */      empty_branch = FALSE;
2395        this_recurse.prev = recurses;
2396        this_recurse.group = scode;
2397    
2398      do      do
2399        {        {
2400        if (could_be_empty_branch(scode, endcode, utf, cd))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2401          {          {
2402          empty_branch = TRUE;          empty_branch = TRUE;
2403          break;          break;
# Line 2453  for (code = first_significant_code(code Line 2453  for (code = first_significant_code(code
2453        empty_branch = FALSE;        empty_branch = FALSE;
2454        do        do
2455          {          {
2456          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2457            empty_branch = TRUE;            empty_branch = TRUE;
2458          code += GET(code, 1);          code += GET(code, 1);
2459          }          }
# Line 2495  for (code = first_significant_code(code Line 2495  for (code = first_significant_code(code
2495        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2496        case OP_CRQUERY:        case OP_CRQUERY:
2497        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2498          case OP_CRPOSSTAR:
2499          case OP_CRPOSQUERY:
2500        break;        break;
2501    
2502        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2503        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2504        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2505          case OP_CRPOSPLUS:
2506        return FALSE;        return FALSE;
2507    
2508        case OP_CRRANGE:        case OP_CRRANGE:
2509        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2510          case OP_CRPOSRANGE:
2511        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2512        break;        break;
2513        }        }
# Line 2511  for (code = first_significant_code(code Line 2515  for (code = first_significant_code(code
2515    
2516      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2517    
2518        case OP_ANY:
2519        case OP_ALLANY:
2520        case OP_ANYBYTE:
2521    
2522      case OP_PROP:      case OP_PROP:
2523      case OP_NOTPROP:      case OP_NOTPROP:
2524        case OP_ANYNL:
2525    
2526        case OP_NOT_HSPACE:
2527        case OP_HSPACE:
2528        case OP_NOT_VSPACE:
2529        case OP_VSPACE:
2530      case OP_EXTUNI:      case OP_EXTUNI:
2531    
2532      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2533      case OP_DIGIT:      case OP_DIGIT:
2534      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2535      case OP_WHITESPACE:      case OP_WHITESPACE:
2536      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2537      case OP_WORDCHAR:      case OP_WORDCHAR:
2538      case OP_ANY:  
     case OP_ALLANY:  
     case OP_ANYBYTE:  
2539      case OP_CHAR:      case OP_CHAR:
2540      case OP_CHARI:      case OP_CHARI:
2541      case OP_NOT:      case OP_NOT:
2542      case OP_NOTI:      case OP_NOTI:
2543    
2544      case OP_PLUS:      case OP_PLUS:
2545        case OP_PLUSI:
2546      case OP_MINPLUS:      case OP_MINPLUS:
2547      case OP_POSPLUS:      case OP_MINPLUSI:
2548      case OP_EXACT:  
2549      case OP_NOTPLUS:      case OP_NOTPLUS:
2550        case OP_NOTPLUSI:
2551      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2552        case OP_NOTMINPLUSI:
2553    
2554        case OP_POSPLUS:
2555        case OP_POSPLUSI:
2556      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2557        case OP_NOTPOSPLUSI:
2558    
2559        case OP_EXACT:
2560        case OP_EXACTI:
2561      case OP_NOTEXACT:      case OP_NOTEXACT:
2562        case OP_NOTEXACTI:
2563    
2564      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2565      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2566      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2567      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2568    
2569      return FALSE;      return FALSE;
2570    
2571      /* These are going to continue, as they may be empty, but we have to      /* These are going to continue, as they may be empty, but we have to
# Line 2558  for (code = first_significant_code(code Line 2585  for (code = first_significant_code(code
2585      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2586      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2587      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2588      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2589        code += 2;        code += 2;
2590      break;      break;
2591    
# Line 2572  for (code = first_significant_code(code Line 2599  for (code = first_significant_code(code
2599      return TRUE;      return TRUE;
2600    
2601      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2602      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2603        followed by a multibyte character. */
2604    
2605  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2606      case OP_STAR:      case OP_STAR:
2607      case OP_STARI:      case OP_STARI:
2608        case OP_NOTSTAR:
2609        case OP_NOTSTARI:
2610    
2611      case OP_MINSTAR:      case OP_MINSTAR:
2612      case OP_MINSTARI:      case OP_MINSTARI:
2613        case OP_NOTMINSTAR:
2614        case OP_NOTMINSTARI:
2615    
2616      case OP_POSSTAR:      case OP_POSSTAR:
2617      case OP_POSSTARI:      case OP_POSSTARI:
2618        case OP_NOTPOSSTAR:
2619        case OP_NOTPOSSTARI:
2620    
2621      case OP_QUERY:      case OP_QUERY:
2622      case OP_QUERYI:      case OP_QUERYI:
2623        case OP_NOTQUERY:
2624        case OP_NOTQUERYI:
2625    
2626      case OP_MINQUERY:      case OP_MINQUERY:
2627      case OP_MINQUERYI:      case OP_MINQUERYI:
2628        case OP_NOTMINQUERY:
2629        case OP_NOTMINQUERYI:
2630    
2631      case OP_POSQUERY:      case OP_POSQUERY:
2632      case OP_POSQUERYI:      case OP_POSQUERYI:
2633        case OP_NOTPOSQUERY:
2634        case OP_NOTPOSQUERYI:
2635    
2636      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2637      break;      break;
2638    
2639      case OP_UPTO:      case OP_UPTO:
2640      case OP_UPTOI:      case OP_UPTOI:
2641        case OP_NOTUPTO:
2642        case OP_NOTUPTOI:
2643    
2644      case OP_MINUPTO:      case OP_MINUPTO:
2645      case OP_MINUPTOI:      case OP_MINUPTOI:
2646        case OP_NOTMINUPTO:
2647        case OP_NOTMINUPTOI:
2648    
2649      case OP_POSUPTO:      case OP_POSUPTO:
2650      case OP_POSUPTOI:      case OP_POSUPTOI:
2651        case OP_NOTPOSUPTO:
2652        case OP_NOTPOSUPTOI:
2653    
2654      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]);
2655      break;      break;
2656  #endif  #endif
# Line 2606  for (code = first_significant_code(code Line 2661  for (code = first_significant_code(code
2661      case OP_MARK:      case OP_MARK:
2662      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
2663      case OP_SKIP_ARG:      case OP_SKIP_ARG:
     code += code[1];  
     break;  
   
2664      case OP_THEN_ARG:      case OP_THEN_ARG:
2665      code += code[1];      code += code[1];
2666      break;      break;
# Line 2652  could_be_empty(const pcre_uchar *code, c Line 2704  could_be_empty(const pcre_uchar *code, c
2704  {  {
2705  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2706    {    {
2707    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2708      return FALSE;      return FALSE;
2709    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2710    }    }
# Line 2662  return TRUE; Line 2714  return TRUE;
2714    
2715    
2716  /*************************************************  /*************************************************
2717  *           Check for POSIX class syntax         *  *        Base opcode of repeated opcodes         *
2718  *************************************************/  *************************************************/
2719    
2720  /* This function is called when the sequence "[:" or "[." or "[=" is  /* Returns the base opcode for repeated single character type opcodes. If the
2721  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  
2722    
2723  Returns:   TRUE or FALSE  Arguments:  c opcode
2724    Returns:    base opcode for the type
2725  */  */
2726    
2727  static BOOL  static pcre_uchar
2728  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)  get_repeat_base(pcre_uchar c)
2729  {  {
2730  int terminator;          /* Don't combine these lines; the Solaris cc */  return (c > OP_TYPEPOSUPTO)? c :
2731  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */         (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2732  for (++ptr; *ptr != 0; ptr++)         (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2733    {         (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2734    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)         (c >= OP_STARI)?      OP_STARI :
2735      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;  
2736  }  }
2737    
2738    
2739    
2740    #ifdef SUPPORT_UCP
2741  /*************************************************  /*************************************************
2742  *          Check POSIX class name                *  *        Check a character and a property        *
2743  *************************************************/  *************************************************/
2744    
2745  /* 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
2746  such as [:alnum:].  is adjacent to a fixed character.
2747    
2748  Arguments:  Arguments:
2749    ptr        points to the first letter    c            the character
2750    len        the length of the name    ptype        the property type
2751      pdata        the data for the type
2752      negated      TRUE if it's a negated property (\P or \p{^)
2753    
2754  Returns:     a value representing the name, or -1 if unknown  Returns:       TRUE if auto-possessifying is OK
2755  */  */
2756    
2757  static int  static BOOL
2758  check_posix_name(const pcre_uchar *ptr, int len)  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2759      BOOL negated)
2760  {  {
2761  const char *pn = posix_names;  const pcre_uint32 *p;
2762  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   *  
 *************************************************/  
2763    
2764  /* OP_RECURSE items contain an offset from the start of the regex to the group  switch(ptype)
2765  that is referenced. This means that groups can be replicated for fixed    {
2766  repetition simply by copying (because the recursion is allowed to refer to    case PT_LAMP:
2767  earlier groups that are outside the current group). However, when a group is    return (prop->chartype == ucp_Lu ||
2768  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is            prop->chartype == ucp_Ll ||
2769  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.  
2770    
2771  This function has been extended with the possibility of forward references for    case PT_GC:
2772  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).  
2773    
2774  Arguments:    case PT_PC:
2775    group      points to the start of the group    return (pdata == prop->chartype) == negated;
   adjust     the amount by which the group is to be moved  
   utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode  
   cd         contains pointers to tables etc.  
   save_hwm   the hwm forward reference pointer at the start of the group  
2776    
2777  Returns:     nothing    case PT_SC:
2778  */    return (pdata == prop->script) == negated;
2779    
2780  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;  
2781    
2782  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)    case PT_ALNUM:
2783    {    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2784    int offset;            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
   pcre_uchar *hc;  
2785    
2786    /* 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
2787    reference. */    means that Perl space and POSIX space are now identical. PCRE was changed
2788      at release 8.34. */
2789    
2790    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    case PT_SPACE:    /* Perl space */
2791      case PT_PXSPACE:  /* POSIX space */
2792      switch(c)
2793      {      {
2794      offset = GET(hc, 0);      HSPACE_CASES:
2795      if (cd->start_code + offset == ptr + 1)      VSPACE_CASES:
2796        {      return negated;
2797        PUT(hc, 0, offset + adjust);  
2798        break;      default:
2799        }      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2800      }      }
2801      break;  /* Control never reaches here */
2802    
2803    /* Otherwise, adjust the recursion offset if it's after the start of this    case PT_WORD:
2804    group. */    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2805              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2806              c == CHAR_UNDERSCORE) == negated;
2807    
2808    if (hc >= cd->hwm)    case PT_CLIST:
2809      p = PRIV(ucd_caseless_sets) + prop->caseset;
2810      for (;;)
2811      {      {
2812      offset = GET(ptr, 1);      if (c < *p) return !negated;
2813      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      if (c == *p++) return negated;
2814      }      }
2815      break;  /* Control never reaches here */
   ptr += 1 + LINK_SIZE;  
2816    }    }
2817    
2818    return FALSE;
2819  }  }
2820    #endif  /* SUPPORT_UCP */
2821    
2822    
2823    
2824  /*************************************************  /*************************************************
2825  *        Insert an automatic callout point       *  *        Fill the character property list        *
2826  *************************************************/  *************************************************/
2827    
2828  /* 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-
2829  callout points before each pattern item.  possessification, and if so, fills a list with its properties.
2830    
2831  Arguments:  Arguments:
2832    code           current code pointer    code        points to start of expression
2833    ptr            current pattern pointer    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2834    cd             pointers to tables etc    fcc         points to case-flipping table
2835      list        points to output list
2836                  list[0] will be filled with the opcode
2837                  list[1] will be non-zero if this opcode
2838                    can match an empty character string
2839                  list[2..7] depends on the opcode
2840    
2841  Returns:         new code pointer  Returns:      points to the start of the next opcode if *code is accepted
2842                  NULL if *code is not accepted
2843  */  */
2844    
2845  static pcre_uchar *  static const pcre_uchar *
2846  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)  get_chr_property_list(const pcre_uchar *code, BOOL utf,
2847      const pcre_uint8 *fcc, pcre_uint32 *list)
2848  {  {
2849  *code++ = OP_CALLOUT;  pcre_uchar c = *code;
2850  *code++ = 255;  const pcre_uchar *end;
2851  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */  const pcre_uint32 *clist_src;
2852  PUT(code, LINK_SIZE, 0);                       /* Default length */  pcre_uint32 *clist_dest;
2853  return code + 2 * LINK_SIZE;  pcre_uint32 chr;
2854  }  pcre_uchar base;
2855    
2856    list[0] = c;
2857    list[1] = FALSE;
2858    code++;
2859    
2860    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2861      {
2862      base = get_repeat_base(c);
2863      c -= (base - OP_STAR);
2864    
2865  /*************************************************    if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2866  *         Complete a callout item                *      code += IMM2_SIZE;
 *************************************************/  
2867    
2868  /* 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.  
2869    
2870  Arguments:    switch(base)
2871    previous_callout   points to previous callout item      {
2872    ptr                current pattern pointer      case OP_STAR:
2873    cd                 pointers to tables etc      list[0] = OP_CHAR;
2874        break;
2875    
2876  Returns:             nothing      case OP_STARI:
2877  */      list[0] = OP_CHARI;
2878        break;
2879    
2880  static void      case OP_NOTSTAR:
2881  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)      list[0] = OP_NOT;
2882  {      break;
 int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));  
 PUT(previous_callout, 2 + LINK_SIZE, length);  
 }  
2883    
2884        case OP_NOTSTARI:
2885        list[0] = OP_NOTI;
2886        break;
2887    
2888        case OP_TYPESTAR:
2889        list[0] = *code;
2890        code++;
2891        break;
2892        }
2893      c = list[0];
2894      }
2895    
2896  #ifdef SUPPORT_UCP  switch(c)
2897  /*************************************************    {
2898  *           Get othercase range                  *    case OP_NOT_DIGIT:
2899  *************************************************/    case OP_DIGIT:
2900      case OP_NOT_WHITESPACE:
2901      case OP_WHITESPACE:
2902      case OP_NOT_WORDCHAR:
2903      case OP_WORDCHAR:
2904      case OP_ANY:
2905      case OP_ALLANY:
2906      case OP_ANYNL:
2907      case OP_NOT_HSPACE:
2908      case OP_HSPACE:
2909      case OP_NOT_VSPACE:
2910      case OP_VSPACE:
2911      case OP_EXTUNI:
2912      case OP_EODN:
2913      case OP_EOD:
2914      case OP_DOLL:
2915      case OP_DOLLM:
2916      return code;
2917    
2918  /* This function is passed the start and end of a class range, in UTF-8 mode    case OP_CHAR:
2919  with UCP support. It searches up the characters, looking for ranges of    case OP_NOT:
2920  characters in the "other" case. Each call returns the next one, updating the    GETCHARINCTEST(chr, code);
2921  start address. A character with multiple other cases is returned on its own    list[2] = chr;
2922  with a special return value.    list[3] = NOTACHAR;
2923      return code;
2924    
2925  Arguments:    case OP_CHARI:
2926    cptr        points to starting character value; updated    case OP_NOTI:
2927    d           end value    list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2928    ocptr       where to put start of othercase range    GETCHARINCTEST(chr, code);
2929    odptr       where to put end of othercase range    list[2] = chr;
2930    
2931  Yield:        -1 when no more  #ifdef SUPPORT_UCP
2932                 0 when a range is returned    if (chr < 128 || (chr < 256 && !utf))
2933                >0 the CASESET offset for char with multiple other cases      list[3] = fcc[chr];
2934                  in this case, ocptr contains the original    else
2935  */      list[3] = UCD_OTHERCASE(chr);
2936    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2937      list[3] = (chr < 256) ? fcc[chr] : chr;
2938    #else
2939      list[3] = fcc[chr];
2940    #endif
2941    
2942  static int    /* The othercase might be the same value. */
 get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,  
   pcre_uint32 *odptr)  
 {  
 pcre_uint32 c, othercase, next;  
 int co;  
2943    
2944  /* Find the first character that has an other case. If it has multiple other    if (chr == list[3])
2945  cases, return its case offset value. */      list[3] = NOTACHAR;
2946      else
2947        list[4] = NOTACHAR;
2948      return code;
2949    
2950  for (c = *cptr; c <= d; c++)  #ifdef SUPPORT_UCP
2951    {    case OP_PROP:
2952    if ((co = UCD_CASESET(c)) != 0)    case OP_NOTPROP:
2953      if (code[0] != PT_CLIST)
2954      {      {
2955      *ocptr = c++;   /* Character that has the set */      list[2] = code[0];
2956      *cptr = c;      /* Rest of input range */      list[3] = code[1];
2957      return co;      return code + 2;
2958      }      }
   if ((othercase = UCD_OTHERCASE(c)) != c) break;  
   }  
2959    
2960  if (c > d) return -1;  /* Reached end of range */    /* Convert only if we have enough space. */
2961    
2962  *ocptr = othercase;    clist_src = PRIV(ucd_caseless_sets) + code[1];
2963  next = othercase + 1;    clist_dest = list + 2;
2964      code += 2;
2965    
2966  for (++c; c <= d; c++)    do {
2967    {       if (clist_dest >= list + 8)
2968    if (UCD_OTHERCASE(c) != next) break;         {
2969    next++;         /* Early return if there is not enough space. This should never
2970    }         happen, since all clists are shorter than 5 character now. */
2971           list[2] = code[0];
2972           list[3] = code[1];
2973           return code;
2974           }
2975         *clist_dest++ = *clist_src;
2976         }
2977      while(*clist_src++ != NOTACHAR);
2978    
2979  *odptr = next - 1;     /* End of othercase range */    /* All characters are stored. The terminating NOTACHAR
2980  *cptr = c;             /* Rest of input range */    is copied form the clist itself. */
2981  return 0;  
2982      list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
2983      return code;
2984    #endif
2985    
2986      case OP_NCLASS:
2987      case OP_CLASS:
2988    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2989      case OP_XCLASS:
2990      if (c == OP_XCLASS)
2991        end = code + GET(code, 0) - 1;
2992      else
2993    #endif
2994        end = code + 32 / sizeof(pcre_uchar);
2995    
2996      switch(*end)
2997        {
2998        case OP_CRSTAR:
2999        case OP_CRMINSTAR:
3000        case OP_CRQUERY:
3001        case OP_CRMINQUERY:
3002        case OP_CRPOSSTAR:
3003        case OP_CRPOSQUERY:
3004        list[1] = TRUE;
3005        end++;
3006        break;
3007    
3008        case OP_CRPLUS:
3009        case OP_CRMINPLUS:
3010        case OP_CRPOSPLUS:
3011        end++;
3012        break;
3013    
3014        case OP_CRRANGE:
3015        case OP_CRMINRANGE:
3016        case OP_CRPOSRANGE:
3017        list[1] = (GET2(end, 1) == 0);
3018        end += 1 + 2 * IMM2_SIZE;
3019        break;
3020        }
3021      list[2] = end - code;
3022      return end;
3023      }
3024    return NULL;    /* Opcode not accepted */
3025  }  }
3026    
3027    
3028    
3029  /*************************************************  /*************************************************
3030  *        Check a character and a property        *  *    Scan further character sets for match       *
3031  *************************************************/  *************************************************/
3032    
3033  /* 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
3034  is adjacent to a fixed character.  which case the base cannot be possessified.
3035    
3036  Arguments:  Arguments:
3037    c            the character    code        points to the byte code
3038    ptype        the property type    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3039    pdata        the data for the type    cd          static compile data
3040    negated      TRUE if it's a negated property (\P or \p{^)    base_list   the data list of the base opcode
3041    
3042  Returns:       TRUE if auto-possessifying is OK  Returns:      TRUE if the auto-possessification is possible
3043  */  */
3044    
3045  static BOOL  static BOOL
3046  check_char_prop(pcre_uint32 c, int ptype, int pdata, BOOL negated)  compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3047      const pcre_uint32 *base_list, const pcre_uchar *base_end)
3048  {  {
3049  #ifdef SUPPORT_UCP  pcre_uchar c;
3050  const pcre_uint32 *p;  pcre_uint32 list[8];
3051  #endif  const pcre_uint32 *chr_ptr;
3052    const pcre_uint32 *ochr_ptr;
3053  const ucd_record *prop = GET_UCD(c);  const pcre_uint32 *list_ptr;
3054    const pcre_uchar *next_code;
3055    const pcre_uint8 *class_bitset;
3056    const pcre_uint32 *set1, *set2, *set_end;
3057    pcre_uint32 chr;
3058    BOOL accepted, invert_bits;
3059    
3060    /* Note: the base_list[1] contains whether the current opcode has greedy
3061    (represented by a non-zero value) quantifier. This is a different from
3062    other character type lists, which stores here that the character iterator
3063    matches to an empty string (also represented by a non-zero value). */
3064    
3065  switch(ptype)  for(;;)
3066    {    {
3067    case PT_LAMP:    /* All operations move the code pointer forward.
3068    return (prop->chartype == ucp_Lu ||    Therefore infinite recursions are not possible. */
           prop->chartype == ucp_Ll ||  
           prop->chartype == ucp_Lt) == negated;  
   
   case PT_GC:  
   return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;  
3069    
3070    case PT_PC:    c = *code;
   return (pdata == prop->chartype) == negated;  
3071    
3072    case PT_SC:    /* Skip over callouts */
   return (pdata == prop->script) == negated;  
3073    
3074    /* These are specials */    if (c == OP_CALLOUT)
3075        {
3076        code += PRIV(OP_lengths)[c];
3077        continue;
3078        }
3079    
3080    case PT_ALNUM:    if (c == OP_ALT)
3081    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||      {
3082            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;      do code += GET(code, 1); while (*code == OP_ALT);
3083        c = *code;
3084        }
3085    
3086    case PT_SPACE:    /* Perl space */    switch(c)
3087    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||      {
3088            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)      case OP_END:
3089            == negated;      case OP_KETRPOS:
3090        /* TRUE only in greedy case. The non-greedy case could be replaced by
3091        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3092        uses more memory, which we cannot get at this stage.) */
3093    
3094    case PT_PXSPACE:  /* POSIX space */      return base_list[1] != 0;
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||  
           c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||  
           c == CHAR_FF || c == CHAR_CR)  
           == negated;  
3095    
3096    case PT_WORD:      case OP_KET:
3097    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||      /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3098            PRIV(ucp_gentype)[prop->chartype] == ucp_N ||      it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3099            c == CHAR_UNDERSCORE) == negated;      cannot be converted to a possessive form. */
3100    
3101  #ifdef SUPPORT_UCP      if (base_list[1] == 0) return FALSE;
3102    case PT_CLIST:  
3103    p = PRIV(ucd_caseless_sets) + prop->caseset;      switch(*(code - GET(code, 1)))
3104    for (;;)        {
3105      {        case OP_ASSERT:
3106      if ((unsigned int)c < *p) return !negated;        case OP_ASSERT_NOT:
3107      if ((unsigned int)c == *p++) return negated;        case OP_ASSERTBACK:
3108      }        case OP_ASSERTBACK_NOT:
3109    break;  /* Control never reaches here */        case OP_ONCE:
3110  #endif        case OP_ONCE_NC:
3111    }        /* Atomic sub-patterns and assertions can always auto-possessify their
3112          last iterator. */
3113          return TRUE;
3114          }
3115    
3116  return FALSE;      code += PRIV(OP_lengths)[c];
3117  }      continue;
 #endif  /* SUPPORT_UCP */  
3118    
3119        case OP_ONCE:
3120        case OP_ONCE_NC:
3121        case OP_BRA:
3122        case OP_CBRA:
3123        next_code = code + GET(code, 1);
3124        code += PRIV(OP_lengths)[c];
3125    
3126        while (*next_code == OP_ALT)
3127          {
3128          if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3129          code = next_code + 1 + LINK_SIZE;
3130          next_code += GET(next_code, 1);
3131          }
3132        continue;
3133    
3134  /*************************************************      case OP_BRAZERO:
3135  *     Check if auto-possessifying is possible    *      case OP_BRAMINZERO:
 *************************************************/  
3136    
3137  /* This function is called for unlimited repeats of certain items, to see      next_code = code + 1;
3138  whether the next thing could possibly match the repeated item. If not, it makes      if (*next_code != OP_BRA && *next_code != OP_CBRA
3139  sense to automatically possessify the repeated item.          && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3140    
3141        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3142    
3143        /* The bracket content will be checked by the
3144        OP_BRA/OP_CBRA case above. */
3145        next_code += 1 + LINK_SIZE;
3146        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3147          return FALSE;
3148    
3149  Arguments:      code += PRIV(OP_lengths)[c];
3150    previous      pointer to the repeated opcode      continue;
3151    utf           TRUE in UTF-8 / UTF-16 / UTF-32 mode      }
   ptr           next character in pattern  
   options       options bits  
   cd            contains pointers to tables etc.  
3152    
3153  Returns:        TRUE if possessifying is wanted    /* Check for a supported opcode, and load its properties. */
 */  
3154    
3155  static BOOL    code = get_chr_property_list(code, utf, cd->fcc, list);
3156  check_auto_possessive(const pcre_uchar *previous, BOOL utf,    if (code == NULL) return FALSE;    /* Unsupported */
   const pcre_uchar *ptr, int options, compile_data *cd)  
 {  
 pcre_uint32 c = NOTACHAR;  
 pcre_uint32 next;  
 int escape;  
 int op_code = *previous++;  
3157    
3158  /* Skip whitespace and comments in extended mode */    /* If either opcode is a small character list, set pointers for comparing
3159      characters from that list with another list, or with a property. */
3160    
3161  if ((options & PCRE_EXTENDED) != 0)    if (base_list[0] == OP_CHAR)
   {  
   for (;;)  
3162      {      {
3163      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      chr_ptr = base_list + 2;
3164      if (*ptr == CHAR_NUMBER_SIGN)      list_ptr = list;
3165        {      }
3166        ptr++;    else if (list[0] == OP_CHAR)
3167        while (*ptr != 0)      {
3168          {      chr_ptr = list + 2;
3169          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }      list_ptr = base_list;
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
       }  
     else break;  
3170      }      }
   }  
3171    
3172  /* If the next item is one that we can handle, get its value. A non-negative    /* Character bitsets can also be compared to certain opcodes. */
 value is a character, a negative value is an escape value. */  
3173    
3174  if (*ptr == CHAR_BACKSLASH)    else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3175    {  #ifdef COMPILE_PCRE8
3176    int temperrorcode = 0;        /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3177    escape = check_escape(&ptr, &next, &temperrorcode, cd->bracount, options, FALSE);        || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
   if (temperrorcode != 0) return FALSE;  
   ptr++;    /* Point after the escape sequence */  
   }  
 else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)  
   {  
   escape = 0;  
 #ifdef SUPPORT_UTF  
   if (utf) { GETCHARINC(next, ptr); } else  
3178  #endif  #endif
3179    next = *ptr++;        )
   }  
 else return FALSE;  
   
 /* Skip whitespace and comments in extended mode */  
   
 if ((options & PCRE_EXTENDED) != 0)  
   {  
   for (;;)  
3180      {      {
3181      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;  #ifdef COMPILE_PCRE8
3182      if (*ptr == CHAR_NUMBER_SIGN)      if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3183        {  #else
3184        ptr++;      if (base_list[0] == OP_CLASS)
       while (*ptr != 0)  
         {  
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
3185  #endif  #endif
3186          }        {
3187          set1 = (pcre_uint32 *)(base_end - base_list[2]);
3188          list_ptr = list;
3189          }
3190        else
3191          {
3192          set1 = (pcre_uint32 *)(code - list[2]);
3193          list_ptr = base_list;
3194        }        }
     else break;  
     }  
   }  
   
 /* If the next thing is itself optional, we have to give up. */  
3195    
3196  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||      invert_bits = FALSE;
3197    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)      switch(list_ptr[0])
3198      return FALSE;        {
3199          case OP_CLASS:
3200          case OP_NCLASS:
3201          set2 = (pcre_uint32 *)
3202            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3203          break;
3204    
3205  /* If the previous item is a character, get its value. */        /* OP_XCLASS cannot be supported here, because its bitset
3206          is not necessarily complete. E.g: [a-\0x{200}] is stored
3207          as a character range, and the appropriate bits are not set. */
3208    
3209  if (op_code == OP_CHAR || op_code == OP_CHARI ||        case OP_NOT_DIGIT:
3210      op_code == OP_NOT || op_code == OP_NOTI)          invert_bits = TRUE;
3211    //if (escape == 0) switch(op_code)          /* Fall through */
3212    {        case OP_DIGIT:
3213  #ifdef SUPPORT_UTF          set2 = (pcre_uint32 *)(cd->cbits + cbit_digit);
3214    GETCHARTEST(c, previous);          break;
 #else  
   c = *previous;  
 #endif  
   }  
3215    
3216  /* Now compare the next item with the previous opcode. First, handle cases when        case OP_NOT_WHITESPACE:
3217  the next item is a character. */          invert_bits = TRUE;
3218            /* Fall through */
3219          case OP_WHITESPACE:
3220            set2 = (pcre_uint32 *)(cd->cbits + cbit_space);
3221            break;
3222    
3223  if (escape == 0)        case OP_NOT_WORDCHAR:
3224    {          invert_bits = TRUE;
3225    /* For a caseless UTF match, the next character may have more than one other          /* Fall through */
3226    case, which maps to the special PT_CLIST property. Check this first. */        case OP_WORDCHAR:
3227            set2 = (pcre_uint32 *)(cd->cbits + cbit_word);
3228  #ifdef SUPPORT_UCP          break;
   if (utf && c != NOTACHAR && (options & PCRE_CASELESS) != 0)  
     {  
     int ocs = UCD_CASESET(next);  
     if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);  
     }  
 #endif  
   
   switch(op_code)  
     {  
     case OP_CHAR:  
     return c != next;  
3229    
3230      /* For CHARI (caseless character) we must check the other case. If we have        default:
3231      Unicode property support, we can use it to test the other case of        return FALSE;
3232      high-valued characters. We know that next can have only one other case,        }
     because multi-other-case characters are dealt with above. */  
3233    
3234      case OP_CHARI:      /* Compare 4 bytes to improve speed. */
3235      if (c == next) return FALSE;      set_end = set1 + (32 / 4);
3236  #ifdef SUPPORT_UTF      if (invert_bits)
     if (utf)  
3237        {        {
3238        pcre_uint32 othercase;        do
3239        if (next < 128) othercase = cd->fcc[next]; else          {
3240  #ifdef SUPPORT_UCP          if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3241        othercase = UCD_OTHERCASE(next);          }
3242  #else        while (set1 < set_end);
       othercase = NOTACHAR;  
 #endif  
       return c != othercase;  
3243        }        }
3244      else      else
 #endif  /* SUPPORT_UTF */  
     return (c != TABLE_GET(next, cd->fcc, next));  /* Not UTF */  
   
     case OP_NOT:  
     return c == next;  
   
     case OP_NOTI:  
     if (c == next) return TRUE;  
 #ifdef SUPPORT_UTF  
     if (utf)  
3245        {        {
3246        pcre_uint32 othercase;        do
3247        if (next < 128) othercase = cd->fcc[next]; else          {
3248  #ifdef SUPPORT_UCP          if ((*set1++ & *set2++) != 0) return FALSE;
3249        othercase = UCD_OTHERCASE(next);          }
3250  #else        while (set1 < set_end);
       othercase = NOTACHAR;  
 #endif  
       return c == othercase;  
3251        }        }
     else  
 #endif  /* SUPPORT_UTF */  
     return (c == TABLE_GET(next, cd->fcc, next));  /* Not UTF */  
3252    
3253      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.      if (list[1] == 0) return TRUE;
3254      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */      /* Might be an empty repeat. */
3255        continue;
3256        }
3257    
3258      case OP_DIGIT:    /* Some property combinations also acceptable. Unicode property opcodes are
3259      return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;    processed specially; the rest can be handled with a lookup table. */
3260    
3261      case OP_NOT_DIGIT:    else
3262      return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;      {
3263        pcre_uint32 leftop, rightop;
3264    
3265      case OP_WHITESPACE:      leftop = base_list[0];
3266      return next > 255 || (cd->ctypes[next] & ctype_space) == 0;      rightop = list[0];
3267    
3268      case OP_NOT_WHITESPACE:  #ifdef SUPPORT_UCP
3269      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;      accepted = FALSE; /* Always set in non-unicode case. */
3270        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3271          {
3272          if (rightop == OP_EOD)
3273            accepted = TRUE;
3274          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3275            {
3276            int n;
3277            const pcre_uint8 *p;
3278            BOOL same = leftop == rightop;
3279            BOOL lisprop = leftop == OP_PROP;
3280            BOOL risprop = rightop == OP_PROP;
3281            BOOL bothprop = lisprop && risprop;
3282    
3283            /* There's a table that specifies how each combination is to be
3284            processed:
3285              0   Always return FALSE (never auto-possessify)
3286              1   Character groups are distinct (possessify if both are OP_PROP)
3287              2   Check character categories in the same group (general or particular)
3288              3   Return TRUE if the two opcodes are not the same
3289              ... see comments below
3290            */
3291    
3292            n = propposstab[base_list[2]][list[2]];
3293            switch(n)
3294              {
3295              case 0: break;
3296              case 1: accepted = bothprop; break;
3297              case 2: accepted = (base_list[3] == list[3]) != same; break;
3298              case 3: accepted = !same; break;
3299    
3300      case OP_WORDCHAR:            case 4:  /* Left general category, right particular category */
3301      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;            accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3302              break;
3303    
3304      case OP_NOT_WORDCHAR:            case 5:  /* Right general category, left particular category */
3305      return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;            accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3306              break;
3307    
3308      case OP_HSPACE:            /* This code is logically tricky. Think hard before fiddling with it.
3309      case OP_NOT_HSPACE:            The posspropstab table has four entries per row. Each row relates to
3310      switch(next)            one of PCRE's special properties such as ALNUM or SPACE or WORD.
3311        {            Only WORD actually needs all four entries, but using repeats for the
3312        HSPACE_CASES:            others means they can all use the same code below.
3313        return op_code == OP_NOT_HSPACE;  
3314              The first two entries in each row are Unicode general categories, and
3315              apply always, because all the characters they include are part of the
3316              PCRE character set. The third and fourth entries are a general and a
3317              particular category, respectively, that include one or more relevant
3318              characters. One or the other is used, depending on whether the check
3319              is for a general or a particular category. However, in both cases the
3320              category contains more characters than the specials that are defined
3321              for the property being tested against. Therefore, it cannot be used
3322              in a NOTPROP case.
3323    
3324              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3325              Underscore is covered by ucp_P or ucp_Po. */
3326    
3327              case 6:  /* Left alphanum vs right general category */
3328              case 7:  /* Left space vs right general category */
3329              case 8:  /* Left word vs right general category */
3330              p = posspropstab[n-6];
3331              accepted = risprop && lisprop ==
3332                (list[3] != p[0] &&
3333                 list[3] != p[1] &&
3334                (list[3] != p[2] || !lisprop));
3335              break;
3336    
3337        default:            case 9:   /* Right alphanum vs left general category */
3338        return op_code != OP_NOT_HSPACE;            case 10:  /* Right space vs left general category */
3339              case 11:  /* Right word vs left general category */
3340              p = posspropstab[n-9];
3341              accepted = lisprop && risprop ==
3342                (base_list[3] != p[0] &&
3343                 base_list[3] != p[1] &&
3344                (base_list[3] != p[2] || !risprop));
3345              break;
3346    
3347              case 12:  /* Left alphanum vs right particular category */
3348              case 13:  /* Left space vs right particular category */
3349              case 14:  /* Left word vs right particular category */
3350              p = posspropstab[n-12];
3351              accepted = risprop && lisprop ==
3352                (catposstab[p[0]][list[3]] &&
3353                 catposstab[p[1]][list[3]] &&
3354                (list[3] != p[3] || !lisprop));
3355              break;
3356    
3357              case 15:  /* Right alphanum vs left particular category */
3358              case 16:  /* Right space vs left particular category */
3359              case 17:  /* Right word vs left particular category */
3360              p = posspropstab[n-15];
3361              accepted = lisprop && risprop ==
3362                (catposstab[p[0]][base_list[3]] &&
3363                 catposstab[p[1]][base_list[3]] &&
3364                (base_list[3] != p[3] || !risprop));
3365              break;
3366              }
3367            }
3368        }        }
3369    
3370      case OP_ANYNL:      else
3371      case OP_VSPACE:  #endif  /* SUPPORT_UCP */
3372      case OP_NOT_VSPACE:  
3373      switch(next)      accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3374               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3375               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3376    
3377        if (!accepted)
3378          return FALSE;
3379    
3380        if (list[1] == 0) return TRUE;
3381        /* Might be an empty repeat. */
3382        continue;
3383        }
3384    
3385      /* Control reaches here only if one of the items is a small character list.
3386      All characters are checked against the other side. */
3387    
3388      do
3389        {
3390        chr = *chr_ptr;
3391    
3392        switch(list_ptr[0])
3393        {        {
3394        VSPACE_CASES:        case OP_CHAR:
3395        return op_code == OP_NOT_VSPACE;        ochr_ptr = list_ptr + 2;
3396          do
3397            {
3398            if (chr == *ochr_ptr) return FALSE;
3399            ochr_ptr++;
3400            }
3401          while(*ochr_ptr != NOTACHAR);
3402          break;
3403    
3404        default:        case OP_NOT:
3405        return op_code != OP_NOT_VSPACE;        ochr_ptr = list_ptr + 2;
3406        }        do
3407            {
3408            if (chr == *ochr_ptr)
3409              break;
3410            ochr_ptr++;
3411            }
3412          while(*ochr_ptr != NOTACHAR);
3413          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3414          break;
3415    
3416          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3417          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3418    
3419          case OP_DIGIT:
3420          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3421          break;
3422    
3423          case OP_NOT_DIGIT:
3424          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3425          break;
3426    
3427          case OP_WHITESPACE:
3428          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3429          break;
3430    
3431          case OP_NOT_WHITESPACE:
3432          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3433          break;
3434    
3435          case OP_WORDCHAR:
3436          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3437          break;
3438    
3439          case OP_NOT_WORDCHAR:
3440          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3441          break;
3442    
3443          case OP_HSPACE:
3444          switch(chr)
3445            {
3446            HSPACE_CASES: return FALSE;
3447            default: break;
3448            }
3449          break;
3450    
3451          case OP_NOT_HSPACE:
3452          switch(chr)
3453            {
3454            HSPACE_CASES: break;
3455            default: return FALSE;
3456            }
3457          break;
3458    
3459          case OP_ANYNL:
3460          case OP_VSPACE:
3461          switch(chr)
3462            {
3463            VSPACE_CASES: return FALSE;
3464            default: break;
3465            }
3466          break;
3467    
3468          case OP_NOT_VSPACE:
3469          switch(chr)
3470            {
3471            VSPACE_CASES: break;
3472            default: return FALSE;
3473            }
3474          break;
3475    
3476          case OP_DOLL:
3477          case OP_EODN:
3478          switch (chr)
3479            {
3480            case CHAR_CR:
3481            case CHAR_LF:
3482            case CHAR_VT:
3483            case CHAR_FF:
3484            case CHAR_NEL:
3485    #ifndef EBCDIC
3486            case 0x2028:
3487            case 0x2029:
3488    #endif  /* Not EBCDIC */
3489            return FALSE;
3490            }
3491          break;
3492    
3493          case OP_EOD:    /* Can always possessify before \z */
3494          break;
3495    
3496  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3497      case OP_PROP:        case OP_PROP:
3498      return check_char_prop(next, previous[0], previous[1], FALSE);        case OP_NOTPROP:
3499          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3500                list_ptr[0] == OP_NOTPROP))
3501            return FALSE;
3502          break;
3503    #endif
3504    
3505      case OP_NOTPROP:        case OP_NCLASS:
3506      return check_char_prop(next, previous[0], previous[1], TRUE);        if (chr > 255) return FALSE;
3507          /* Fall through */
3508    
3509          case OP_CLASS:
3510          if (chr > 255) break;
3511          class_bitset = (pcre_uint8 *)
3512            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3513          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3514          break;
3515    
3516    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3517          case OP_XCLASS:
3518          if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3519              list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3520          break;
3521  #endif  #endif
3522    
3523      default:        default:
3524      return FALSE;        return FALSE;
3525          }
3526    
3527        chr_ptr++;
3528      }      }
3529      while(*chr_ptr != NOTACHAR);
3530    
3531      /* At least one character must be matched from this opcode. */
3532    
3533      if (list[1] == 0) return TRUE;
3534    }    }
3535    
3536  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP  return FALSE;
3537  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. */  
3538    
 switch(op_code)  
   {  
   case OP_CHAR:  
   case OP_CHARI:  
   switch(escape)  
     {  
     case ESC_d:  
     return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;  
3539    
     case ESC_D:  
     return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;  
3540    
3541      case ESC_s:  /*************************************************
3542      return c > 255 || (cd->ctypes[c] & ctype_space) == 0;  *    Scan compiled regex for auto-possession     *
3543    *************************************************/
3544    
3545      case ESC_S:  /* Replaces single character iterations with their possessive alternatives
3546      return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;  if appropriate. This function modifies the compiled opcode!
3547    
3548      case ESC_w:  Arguments:
3549      return c > 255 || (cd->ctypes[c] & ctype_word) == 0;    code        points to start of the byte code
3550      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3551      cd          static compile data
3552    
3553      case ESC_W:  Returns:      nothing
3554      return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;  */
3555    
3556      case ESC_h:  static void
3557      case ESC_H:  auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3558      switch(c)  {
3559        {  register pcre_uchar c;
3560        HSPACE_CASES:  const pcre_uchar *end;
3561        return escape != ESC_h;  pcre_uchar *repeat_opcode;
3562    pcre_uint32 list[8];
       default:  
       return escape == ESC_h;  
       }  
3563    
3564      case ESC_v:  for (;;)
3565      case ESC_V:    {
3566      switch(c)    c = *code;
3567    
3568      if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3569        {
3570        c -= get_repeat_base(c) - OP_STAR;
3571        end = (c <= OP_MINUPTO) ?
3572          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3573        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3574    
3575        if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3576        {        {
3577        VSPACE_CASES:        switch(c)
3578        return escape != ESC_v;          {
3579            case OP_STAR:
3580            *code += OP_POSSTAR - OP_STAR;
3581            break;
3582    
3583        default:          case OP_MINSTAR:
3584        return escape == ESC_v;          *code += OP_POSSTAR - OP_MINSTAR;
3585        }          break;
3586    
3587      /* When PCRE_UCP is set, these values get generated for \d etc. Find          case OP_PLUS:
3588      their substitutions and process them. The result will always be either          *code += OP_POSPLUS - OP_PLUS;
3589      ESC_p or ESC_P. Then fall through to process those values. */          break;
3590    
3591  #ifdef SUPPORT_UCP          case OP_MINPLUS:
3592      case ESC_du:          *code += OP_POSPLUS - OP_MINPLUS;
3593      case ESC_DU:          break;
3594      case ESC_wu:  
3595      case ESC_WU:          case OP_QUERY:
3596      case ESC_su:          *code += OP_POSQUERY - OP_QUERY;
3597      case ESC_SU:          break;
3598        {  
3599        int temperrorcode = 0;          case OP_MINQUERY:
3600        ptr = substitutes[escape - ESC_DU];          *code += OP_POSQUERY - OP_MINQUERY;
3601        escape = check_escape(&ptr, &next, &temperrorcode, 0, options, FALSE);          break;
3602        if (temperrorcode != 0) return FALSE;  
3603        ptr++;    /* For compatibility */          case OP_UPTO:
3604            *code += OP_POSUPTO - OP_UPTO;
3605            break;
3606    
3607            case OP_MINUPTO:
3608            *code += OP_MINUPTO - OP_UPTO;
3609            break;
3610            }
3611        }        }
3612      /* Fall through */      c = *code;
3613        }
3614      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3615        {
3616    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3617        if (c == OP_XCLASS)
3618          repeat_opcode = code + GET(code, 1);
3619        else
3620    #endif
3621          repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3622    
3623      case ESC_p:      c = *repeat_opcode;
3624      case ESC_P:      if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3625        {        {
3626        int ptype, pdata, errorcodeptr;        /* end must not be NULL. */
3627        BOOL negated;        end = get_chr_property_list(code, utf, cd->fcc, list);
3628    
3629        ptr--;      /* Make ptr point at the p or P */        list[1] = (c & 1) == 0;
       ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);  
       if (ptype < 0) return FALSE;  
       ptr++;      /* Point past the final curly ket */  
3630    
3631        /* If the property item is optional, we have to give up. (When generated        if (compare_opcodes(end, utf, cd, list, end))
3632        from \d etc by PCRE_UCP, this test will have been applied much earlier,          {
3633        to the original \d etc. At this point, ptr will point to a zero byte. */          switch (c)
3634              {
3635              case OP_CRSTAR:
3636              case OP_CRMINSTAR:
3637              *repeat_opcode = OP_CRPOSSTAR;
3638              break;
3639    
3640        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||            case OP_CRPLUS:
3641          STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)            case OP_CRMINPLUS:
3642            return FALSE;            *repeat_opcode = OP_CRPOSPLUS;
3643              break;
3644    
3645        /* Do the property check. */            case OP_CRQUERY:
3646              case OP_CRMINQUERY:
3647              *repeat_opcode = OP_CRPOSQUERY;
3648              break;
3649    
3650        return check_char_prop(c, ptype, pdata, (escape == ESC_P) != negated);            case OP_CRRANGE:
3651              case OP_CRMINRANGE:
3652              *repeat_opcode = OP_CRPOSRANGE;
3653              break;
3654              }
3655            }
3656        }        }
3657        c = *code;
3658        }
3659    
3660      switch(c)
3661        {
3662        case OP_END:
3663        return;
3664    
3665        case OP_TYPESTAR:
3666        case OP_TYPEMINSTAR:
3667        case OP_TYPEPLUS:
3668        case OP_TYPEMINPLUS:
3669        case OP_TYPEQUERY:
3670        case OP_TYPEMINQUERY:
3671        case OP_TYPEPOSSTAR:
3672        case OP_TYPEPOSPLUS:
3673        case OP_TYPEPOSQUERY:
3674        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3675        break;
3676    
3677        case OP_TYPEUPTO:
3678        case OP_TYPEMINUPTO:
3679        case OP_TYPEEXACT:
3680        case OP_TYPEPOSUPTO:
3681        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3682          code += 2;
3683        break;
3684    
3685    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3686        case OP_XCLASS:
3687        code += GET(code, 1);
3688        break;
3689  #endif  #endif
3690    
3691      default:      case OP_MARK:
3692      return FALSE;      case OP_PRUNE_ARG:
3693        case OP_SKIP_ARG:
3694        case OP_THEN_ARG:
3695        code += code[1];
3696        break;
3697      }      }
3698    
3699    /* In principle, support for Unicode properties should be integrated here as    /* Add in the fixed length from the table */
3700    well. It means re-organizing the above code so as to get hold of the property  
3701    values before switching on the op-code. However, I wonder how many patterns    code += PRIV(OP_lengths)[c];
3702    combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,  
3703    these op-codes are never generated.) */    /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3704      a multi-byte character. The length in the table is a minimum, so we have to
3705      arrange to skip the extra bytes. */
3706    
3707    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3708      if (utf) switch(c)
3709        {
3710        case OP_CHAR:
3711        case OP_CHARI:
3712        case OP_NOT:
3713        case OP_NOTI:
3714        case OP_STAR:
3715        case OP_MINSTAR:
3716        case OP_PLUS:
3717        case OP_MINPLUS:
3718        case OP_QUERY:
3719        case OP_MINQUERY:
3720        case OP_UPTO:
3721        case OP_MINUPTO:
3722        case OP_EXACT:
3723        case OP_POSSTAR:
3724        case OP_POSPLUS:
3725        case OP_POSQUERY:
3726        case OP_POSUPTO:
3727        case OP_STARI:
3728        case OP_MINSTARI:
3729        case OP_PLUSI:
3730        case OP_MINPLUSI:
3731        case OP_QUERYI:
3732        case OP_MINQUERYI:
3733        case OP_UPTOI:
3734        case OP_MINUPTOI:
3735        case OP_EXACTI:
3736        case OP_POSSTARI:
3737        case OP_POSPLUSI:
3738        case OP_POSQUERYI:
3739        case OP_POSUPTOI:
3740        case OP_NOTSTAR:
3741        case OP_NOTMINSTAR:
3742        case OP_NOTPLUS:
3743        case OP_NOTMINPLUS:
3744        case OP_NOTQUERY:
3745        case OP_NOTMINQUERY:
3746        case OP_NOTUPTO:
3747        case OP_NOTMINUPTO:
3748        case OP_NOTEXACT:
3749        case OP_NOTPOSSTAR:
3750        case OP_NOTPOSPLUS:
3751        case OP_NOTPOSQUERY:
3752        case OP_NOTPOSUPTO:
3753        case OP_NOTSTARI:
3754        case OP_NOTMINSTARI:
3755        case OP_NOTPLUSI:
3756        case OP_NOTMINPLUSI:
3757        case OP_NOTQUERYI:
3758        case OP_NOTMINQUERYI:
3759        case OP_NOTUPTOI:
3760        case OP_NOTMINUPTOI:
3761        case OP_NOTEXACTI:
3762        case OP_NOTPOSSTARI:
3763        case OP_NOTPOSPLUSI:
3764        case OP_NOTPOSQUERYI:
3765        case OP_NOTPOSUPTOI:
3766        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3767        break;
3768        }
3769    #else
3770      (void)(utf);  /* Keep compiler happy by referencing function argument */
3771    #endif
3772      }
3773    }
3774    
3775    
3776    
3777    /*************************************************
3778    *           Check for POSIX class syntax         *
3779    *************************************************/
3780    
3781    /* This function is called when the sequence "[:" or "[." or "[=" is
3782    encountered in a character class. It checks whether this is followed by a
3783    sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3784    reach an unescaped ']' without the special preceding character, return FALSE.
3785    
3786    Originally, this function only recognized a sequence of letters between the
3787    terminators, but it seems that Perl recognizes any sequence of characters,
3788    though of course unknown POSIX names are subsequently rejected. Perl gives an
3789    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3790    didn't consider this to be a POSIX class. Likewise for [:1234:].
3791    
3792    The problem in trying to be exactly like Perl is in the handling of escapes. We
3793    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3794    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3795    below handles the special case of \], but does not try to do any other escape
3796    processing. This makes it different from Perl for cases such as [:l\ower:]
3797    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3798    "l\ower". This is a lesser evil than not diagnosing bad classes when Perl does,
3799    I think.
3800    
3801    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3802    It seems that the appearance of a nested POSIX class supersedes an apparent
3803    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3804    a digit.
3805    
3806    In Perl, unescaped square brackets may also appear as part of class names. For
3807    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3808    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3809    seem right at all. PCRE does not allow closing square brackets in POSIX class
3810    names.
3811    
3812    Arguments:
3813      ptr      pointer to the initial [
3814      endptr   where to return the end pointer
3815    
3816    Returns:   TRUE or FALSE
3817    */
3818    
3819    static BOOL
3820    check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3821    {
3822    pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
3823    terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
3824    for (++ptr; *ptr != CHAR_NULL; ptr++)
3825      {
3826      if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3827        ptr++;
3828      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3829      else
3830        {
3831        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3832          {
3833          *endptr = ptr;
3834          return TRUE;
3835          }
3836        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3837             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3838              ptr[1] == CHAR_EQUALS_SIGN) &&
3839            check_posix_syntax(ptr, endptr))
3840          return FALSE;
3841        }
3842      }
3843    return FALSE;
3844    }
3845    
3846    
3847    
3848    
3849    /*************************************************
3850    *          Check POSIX class name                *
3851    *************************************************/
3852    
3853    /* This function is called to check the name given in a POSIX-style class entry
3854    such as [:alnum:].
3855    
3856    Arguments:
3857      ptr        points to the first letter
3858      len        the length of the name
3859    
3860    Returns:     a value representing the name, or -1 if unknown
3861    */
3862    
3863    static int
3864    check_posix_name(const pcre_uchar *ptr, int len)
3865    {
3866    const char *pn = posix_names;
3867    register int yield = 0;
3868    while (posix_name_lengths[yield] != 0)
3869      {
3870      if (len == posix_name_lengths[yield] &&
3871        STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3872      pn += posix_name_lengths[yield] + 1;
3873      yield++;
3874      }
3875    return -1;
3876    }
3877    
3878    
3879    /*************************************************
3880    *    Adjust OP_RECURSE items in repeated group   *
3881    *************************************************/
3882    
3883    /* OP_RECURSE items contain an offset from the start of the regex to the group
3884    that is referenced. This means that groups can be replicated for fixed
3885    repetition simply by copying (because the recursion is allowed to refer to
3886    earlier groups that are outside the current group). However, when a group is
3887    optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3888    inserted before it, after it has been compiled. This means that any OP_RECURSE
3889    items within it that refer to the group itself or any contained groups have to
3890    have their offsets adjusted. That one of the jobs of this function. Before it
3891    is called, the partially compiled regex must be temporarily terminated with
3892    OP_END.
3893    
3894    This function has been extended with the possibility of forward references for
3895    recursions and subroutine calls. It must also check the list of such references
3896    for the group we are dealing with. If it finds that one of the recursions in
3897    the current group is on this list, it adjusts the offset in the list, not the
3898    value in the reference (which is a group number).
3899    
3900    Arguments:
3901      group      points to the start of the group
3902      adjust     the amount by which the group is to be moved
3903      utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
3904      cd         contains pointers to tables etc.
3905      save_hwm   the hwm forward reference pointer at the start of the group
3906    
3907    Returns:     nothing
3908    */
3909    
3910    static void
3911    adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3912      pcre_uchar *save_hwm)
3913    {
3914    pcre_uchar *ptr = group;
3915    
3916    while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3917      {
3918      int offset;
3919      pcre_uchar *hc;
3920    
3921      /* See if this recursion is on the forward reference list. If so, adjust the
3922      reference. */
3923    
3924      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3925        {
3926        offset = (int)GET(hc, 0);
3927        if (cd->start_code + offset == ptr + 1)
3928          {
3929          PUT(hc, 0, offset + adjust);
3930          break;
3931          }
3932        }
3933    
3934      /* Otherwise, adjust the recursion offset if it's after the start of this
3935      group. */
3936    
3937      if (hc >= cd->hwm)
3938        {
3939        offset = (int)GET(ptr, 1);
3940        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3941        }
3942    
3943      ptr += 1 + LINK_SIZE;
3944      }
3945    }
3946    
3947    
3948    
3949    /*************************************************
3950    *        Insert an automatic callout point       *
3951    *************************************************/
3952    
3953    /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
3954    callout points before each pattern item.
3955    
3956    Arguments:
3957      code           current code pointer
3958      ptr            current pattern pointer
3959      cd             pointers to tables etc
3960    
3961    Returns:         new code pointer
3962    */
3963    
3964    static pcre_uchar *
3965    auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
3966    {
3967    *code++ = OP_CALLOUT;
3968    *code++ = 255;
3969    PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
3970    PUT(code, LINK_SIZE, 0);                       /* Default length */
3971    return code + 2 * LINK_SIZE;
3972    }
3973    
3974    
3975    
3976    /*************************************************
3977    *         Complete a callout item                *
3978    *************************************************/
3979    
3980    /* A callout item contains the length of the next item in the pattern, which
3981    we can't fill in till after we have reached the relevant point. This is used
3982    for both automatic and manual callouts.
3983    
3984    Arguments:
3985      previous_callout   points to previous callout item
3986      ptr                current pattern pointer
3987      cd                 pointers to tables etc
3988    
3989    Returns:             nothing
3990    */
3991    
3992    static void
3993    complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
3994    {
3995    int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
3996    PUT(previous_callout, 2 + LINK_SIZE, length);
3997    }
3998    
3999    
   case OP_DIGIT:  
   return escape == ESC_D || escape == ESC_s || escape == ESC_W ||  
          escape == ESC_h || escape == ESC_v || escape == ESC_R;  
4000    
4001    case OP_NOT_DIGIT:  #ifdef SUPPORT_UCP
4002    return escape == ESC_d;  /*************************************************
4003    *           Get othercase range                  *
4004    *************************************************/
4005    
4006    case OP_WHITESPACE:  /* This function is passed the start and end of a class range, in UTF-8 mode
4007    return escape == ESC_S || escape == ESC_d || escape == ESC_w;  with UCP support. It searches up the characters, looking for ranges of
4008    characters in the "other" case. Each call returns the next one, updating the
4009    start address. A character with multiple other cases is returned on its own
4010    with a special return value.
4011    
4012    case OP_NOT_WHITESPACE:  Arguments:
4013    return escape == ESC_s || escape == ESC_h || escape == ESC_v || escape == ESC_R;    cptr        points to starting character value; updated
4014      d           end value
4015      ocptr       where to put start of othercase range
4016      odptr       where to put end of othercase range
4017    
4018    case OP_HSPACE:  Yield:        -1 when no more
4019    return escape == ESC_S || escape == ESC_H || escape == ESC_d ||                 0 when a range is returned
4020           escape == ESC_w || escape == ESC_v || escape == ESC_R;                >0 the CASESET offset for char with multiple other cases
4021                    in this case, ocptr contains the original
4022    */
4023    
4024    case OP_NOT_HSPACE:  static int
4025    return escape == ESC_h;  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4026      pcre_uint32 *odptr)
4027    {
4028    pcre_uint32 c, othercase, next;
4029    unsigned int co;
4030    
4031    /* 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
4032    case OP_ANYNL:  cases, return its case offset value. */
   case OP_VSPACE:  
   return escape == ESC_V || escape == ESC_d || escape == ESC_w;  
4033    
4034    case OP_NOT_VSPACE:  for (c = *cptr; c <= d; c++)
4035    return escape == ESC_v || escape == ESC_R;    {
4036      if ((co = UCD_CASESET(c)) != 0)
4037        {
4038        *ocptr = c++;   /* Character that has the set */
4039        *cptr = c;      /* Rest of input range */
4040        return (int)co;
4041        }
4042      if ((othercase = UCD_OTHERCASE(c)) != c) break;
4043      }
4044    
4045    case OP_WORDCHAR:  if (c > d) return -1;  /* Reached end of range */
   return escape == ESC_W || escape == ESC_s || escape == ESC_h ||  
          escape == ESC_v || escape == ESC_R;  
4046    
4047    case OP_NOT_WORDCHAR:  *ocptr = othercase;
4048    return escape == ESC_w || escape == ESC_d;  next = othercase + 1;
4049    
4050    default:  for (++c; c <= d; c++)
4051    return FALSE;    {
4052      if (UCD_OTHERCASE(c) != next) break;
4053      next++;
4054    }    }
4055    
4056  /* Control does not reach here */  *odptr = next - 1;     /* End of othercase range */
4057    *cptr = c;             /* Rest of input range */
4058    return 0;
4059  }  }
4060    #endif  /* SUPPORT_UCP */
4061    
4062    
4063    
# Line 3418  switch(op_code) Line 4066  switch(op_code)
4066  *************************************************/  *************************************************/
4067    
4068  /* 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
4069  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
4070  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
4071  mutually recursive with the function immediately below.  mutually recursive with the function immediately below.
4072    
4073  Arguments:  Arguments:
4074    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4075    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4076    options       the options word    options       the options word
4077    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4078    start         start of range character    start         start of range character
4079    end           end of range character    end           end of range character
4080    
4081  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4082                  the pointer to extra data is updated                  the pointer to extra data is updated
4083  */  */
# Line 3441  add_to_class(pcre_uint8 *classbits, pcre Line 4089  add_to_class(pcre_uint8 *classbits, pcre
4089  pcre_uint32 c;  pcre_uint32 c;
4090  int n8 = 0;  int n8 = 0;
4091    
4092  /* If caseless matching is required, scan the range and process alternate  /* If caseless matching is required, scan the range and process alternate
4093  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
4094  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
4095  range. */  range. */
4096    
4097  if ((options & PCRE_CASELESS) != 0)  if ((options & PCRE_CASELESS) != 0)
4098    {    {
4099  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4100    if ((options & PCRE_UTF8) != 0)    if ((options & PCRE_UTF8) != 0)
4101      {      {
4102      int rc;      int rc;
4103      pcre_uint32 oc, od;      pcre_uint32 oc, od;
4104    
4105      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
4106      c = start;      c = start;
4107    
4108      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
4109        {        {
4110        /* Handle a single character that has more than one other case. */        /* Handle a single character that has more than one other case. */
4111    
4112        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4113          PRIV(ucd_caseless_sets) + rc, oc);          PRIV(ucd_caseless_sets) + rc, oc);
4114    
4115        /* Do nothing if the other case range is within the original range. */        /* Do nothing if the other case range is within the original range. */
4116    
4117        else if (oc >= start && od <= end) continue;        else if (oc >= start && od <= end) continue;
4118    
4119        /* 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
4120        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
4121        range. Otherwise, use a recursive call to add the additional range. */        range. Otherwise, use a recursive call to add the additional range. */
4122    
4123        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4124        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
4125        else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);        else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
# Line 3481  if ((options & PCRE_CASELESS) != 0) Line 4129  if ((options & PCRE_CASELESS) != 0)
4129  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
4130    
4131    /* Not UTF-mode, or no UCP */    /* Not UTF-mode, or no UCP */
4132    
4133    for (c = start; c <= end && c < 256; c++)    for (c = start; c <= end && c < 256; c++)
4134      {      {
4135      SETBIT(classbits, cd->fcc[c]);      SETBIT(classbits, cd->fcc[c]);
4136      n8++;      n8++;
4137      }      }
4138    }    }
4139    
4140  /* 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
4141  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
4142  in all cases. */  in all cases. */
# Line 3514  if (end < 0x100) Line 4162  if (end < 0x100)
4162    {    {
4163    for (c = start; c <= end; c++)    for (c = start; c <= end; c++)
4164      {      {
4165      n8++;      n8++;
4166      SETBIT(classbits, c);      SETBIT(classbits, c);
4167      }      }
4168    }    }
4169    
4170  else  else
4171    {    {
4172    pcre_uchar *uchardata = *uchardptr;    pcre_uchar *uchardata = *uchardptr;
4173    
4174  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4175    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
4176      {      {
4177      if (start < end)      if (start < end)
4178        {        {
4179        *uchardata++ = XCL_RANGE;        *uchardata++ = XCL_RANGE;
4180        uchardata += PRIV(ord2utf)(start, uchardata);        uchardata += PRIV(ord2utf)(start, uchardata);
4181        uchardata += PRIV(ord2utf)(end, uchardata);        uchardata += PRIV(ord2utf)(end, uchardata);
4182        }        }
4183      else if (start == end)      else if (start == end)
4184        {        {
4185        *uchardata++ = XCL_SINGLE;        *uchardata++ = XCL_SINGLE;
4186        uchardata += PRIV(ord2utf)(start, uchardata);        uchardata += PRIV(ord2utf)(start, uchardata);
4187        }        }
4188      }      }
4189    else    else
4190  #endif  /* SUPPORT_UTF */  #endif  /* SUPPORT_UTF */
4191    
4192    /* Without UTF support, character values are constrained by the bit length,    /* Without UTF support, character values are constrained by the bit length,
4193    and can only be > 256 for 16-bit and 32-bit libraries. */    and can only be > 256 for 16-bit and 32-bit libraries. */
4194    
4195  #ifdef COMPILE_PCRE8  #ifdef COMPILE_PCRE8
4196      {}      {}
4197  #else  #else
4198    if (start < end)    if (start < end)
4199      {      {
4200      *uchardata++ = XCL_RANGE;      *uchardata++ = XCL_RANGE;
# Line 3557  else Line 4205  else
4205      {      {
4206      *uchardata++ = XCL_SINGLE;      *uchardata++ = XCL_SINGLE;
4207      *uchardata++ = start;      *uchardata++ = start;
4208      }      }
4209  #endif  #endif
4210    
4211    *uchardptr = uchardata;   /* Updata extra data pointer */    *uchardptr = uchardata;   /* Updata extra data pointer */
4212    }    }
4213    
4214  return n8;    /* Number of 8-bit characters */  return n8;    /* Number of 8-bit characters */
4215  }  }
4216    
4217    
4218    
4219    
4220  /*************************************************  /*************************************************
4221  *        Add a list of characters to a class     *  *        Add a list of characters to a class     *
4222  *************************************************/  *************************************************/
4223    
4224  /* 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
4225  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
4226  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
4227  handled appropriately. This function is mutually recursive with the function  handled appropriately. This function is mutually recursive with the function
# Line 3583  Arguments: Line 4231  Arguments:
4231    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4232    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4233    options       the options word    options       the options word
4234    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4235    p             points to row of 32-bit values, terminated by NOTACHAR    p             points to row of 32-bit values, terminated by NOTACHAR
4236    except        character to omit; this is used when adding lists of    except        character to omit; this is used when adding lists of
4237                    case-equivalent characters to avoid including the one we                    case-equivalent characters to avoid including the one we
4238                    already know about                    already know about
4239    
4240  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4241                  the pointer to extra data is updated                  the pointer to extra data is updated
4242  */  */
# Line 3602  while (p[0] < NOTACHAR) Line 4250  while (p[0] < NOTACHAR)
4250    {    {
4251    int n = 0;    int n = 0;
4252    if (p[0] != except)    if (p[0] != except)
4253      {      {
4254      while(p[n+1] == p[0] + n + 1) n++;      while(p[n+1] == p[0] + n + 1) n++;
4255      n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);      n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4256      }      }
4257    p += n + 1;    p += n + 1;
4258    }    }
4259  return n8;  return n8;
4260  }  }
4261    
4262    
4263    
# Line 3624  Arguments: Line 4272  Arguments:
4272    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4273    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4274    options       the options word    options       the options word
4275    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4276    p             points to row of 32-bit values, terminated by NOTACHAR    p             points to row of 32-bit values, terminated by NOTACHAR
4277    
4278  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4279                  the pointer to extra data is updated                  the pointer to extra data is updated
4280  */  */
4281    
4282  static int  static int
4283  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
4284    int options, compile_data *cd, const pcre_uint32 *p)    int options, compile_data *cd, const pcre_uint32 *p)
4285  {  {
4286  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
# Line 3644  while (p[0] < NOTACHAR) Line 4292  while (p[0] < NOTACHAR)
4292    while (p[1] == p[0] + 1) p++;    while (p[1] == p[0] + 1) p++;
4293    n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,    n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
4294      (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);      (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
4295    p++;    p++;
4296    }    }
4297  return n8;  return n8;
4298  }  }
4299    
4300    
4301    
# Line 3662  to find out the amount of memory needed, Line 4310  to find out the amount of memory needed,
4310  phase. The value of lengthptr distinguishes the two phases.  phase. The value of lengthptr distinguishes the two phases.
4311    
4312  Arguments:  Arguments:
4313    optionsptr     pointer to the option bits    optionsptr        pointer to the option bits
4314    codeptr        points to the pointer to the current code point    codeptr           points to the pointer to the current code point
4315    ptrptr         points to the current pattern pointer    ptrptr            points to the current pattern pointer
4316    errorcodeptr   points to error code variable    errorcodeptr      points to error code variable
4317    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr      place to put the first required character
4318    reqcharptr     set to the last literal character required, else < 0    firstcharflagsptr place to put the first character flags, or a negative number
4319    bcptr          points to current branch chain    reqcharptr        place to put the last required character
4320    cond_depth     conditional nesting depth    reqcharflagsptr   place to put the last required character flags, or a negative number
4321    cd             contains pointers to tables etc.    bcptr             points to current branch chain
4322    lengthptr      NULL during the real compile phase    cond_depth        conditional nesting depth
4323                   points to length accumulator during pre-compile phase    cd                contains pointers to tables etc.
4324      lengthptr         NULL during the real compile phase
4325                        points to length accumulator during pre-compile phase
4326    
4327  Returns:         TRUE on success  Returns:            TRUE on success
4328                   FALSE, with *errorcodeptr set non-zero on error                      FALSE, with *errorcodeptr set non-zero on error
4329  */  */
4330    
4331  static BOOL  static BOOL
4332  compile_branch(int *optionsptr, pcre_uchar **codeptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
4333    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,    const pcre_uchar **ptrptr, int *errorcodeptr,
4334    pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,    pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
4335      pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
4336      branch_chain *bcptr, int cond_depth,
4337    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
4338  {  {
4339  int repeat_type, op_type;  int repeat_type, op_type;
4340  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
4341  int bravalue = 0;  int bravalue = 0;
4342  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
4343  pcre_int32 firstchar, reqchar;  pcre_uint32 firstchar, reqchar;
4344  pcre_int32 zeroreqchar, zerofirstchar;  pcre_int32 firstcharflags, reqcharflags;
4345    pcre_uint32 zeroreqchar, zerofirstchar;
4346    pcre_int32 zeroreqcharflags, zerofirstcharflags;
4347  pcre_int32 req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
4348  int options = *optionsptr;               /* May change dynamically */  int options = *optionsptr;               /* May change dynamically */
4349  int after_manual_callout = 0;  int after_manual_callout = 0;
# Line 3717  dynamically as we process the pattern. * Line 4371  dynamically as we process the pattern. *
4371  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4372  /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */  /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
4373  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
4374    #ifndef COMPILE_PCRE32
4375  pcre_uchar utf_chars[6];  pcre_uchar utf_chars[6];
4376    #endif
4377  #else  #else
4378  BOOL utf = FALSE;  BOOL utf = FALSE;
4379  #endif  #endif
4380    
4381  /* Helper variables for OP_XCLASS opcode (for characters > 255). We define  /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
4382  class_uchardata always so that it can be passed to add_to_class() always,  class_uchardata always so that it can be passed to add_to_class() always,
4383  though it will not be used in non-UTF 8-bit cases. This avoids having to supply  though it will not be used in non-UTF 8-bit cases. This avoids having to supply
4384  alternative calls for the different cases. */  alternative calls for the different cases. */
4385    
4386  pcre_uchar *class_uchardata;  pcre_uchar *class_uchardata;
# Line 3752  to take the zero repeat into account. Th Line 4408  to take the zero repeat into account. Th
4408  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
4409  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
4410    
4411  firstchar = reqchar = zerofirstchar = zeroreqchar = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
4412    firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
4413    
4414  /* The variable req_caseopt contains either the REQ_CASELESS value  /* The variable req_caseopt contains either the REQ_CASELESS value
4415  or zero, according to the current setting of the caseless flag. The  or zero, according to the current setting of the caseless flag. The
# Line 3778  for (;; ptr++) Line 4435  for (;; ptr++)
4435    int recno;    int recno;
4436    int refsign;    int refsign;
4437    int skipbytes;    int skipbytes;
4438    int subreqchar;    pcre_uint32 subreqchar, subfirstchar;
4439    int subfirstchar;    pcre_int32 subreqcharflags, subfirstcharflags;
4440    int terminator;    int terminator;
4441    int mclength;    unsigned int mclength;
4442    int tempbracount;    unsigned int tempbracount;
4443    int ec; // FIXMEchpe pcre_uint32    pcre_uint32 ec;
4444    pcre_uchar mcbuffer[8];    pcre_uchar mcbuffer[8];
4445    
4446    /* Get next character in the pattern */    /* Get next character in the pattern */
4447    
4448    c = *ptr;    c = *ptr;
4449    
4450    /* 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
4451    string. Nesting only happens one level deep. */    string. Nesting only happens one level deep. */
4452    
4453    if (c == 0 && nestptr != NULL)    if (c == CHAR_NULL && nestptr != NULL)
4454      {      {
4455      ptr = nestptr;      ptr = nestptr;
4456      nestptr = NULL;      nestptr = NULL;
# Line 3868  for (;; ptr++) Line 4525  for (;; ptr++)
4525    
4526    /* 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 */
4527    
4528    if (inescq && c != 0)    if (inescq && c != CHAR_NULL)
4529      {      {
4530      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4531        {        {
# Line 3891  for (;; ptr++) Line 4548  for (;; ptr++)
4548          }          }
4549        goto NORMAL_CHAR;        goto NORMAL_CHAR;
4550        }        }
4551        /* Control does not reach here. */
4552      }      }
4553    
4554    /* Fill in length of a previous callout, except when the next thing is    /* In extended mode, skip white space and comments. We need a loop in order
4555    a quantifier. */    to check for more white space and more comments after a comment. */
4556    
   is_quantifier =  
     c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||  
     (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));  
   
   if (!is_quantifier && previous_callout != NULL &&  
        after_manual_callout-- <= 0)  
     {  
     if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */  
       complete_callout(previous_callout, ptr, cd);  
     previous_callout = NULL;  
     }  
   
   /* In extended mode, skip white space and comments. */  
   
4557    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
4558      {      {
4559      if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;      for (;;)
     if (c == CHAR_NUMBER_SIGN)  
4560        {        {
4561          while (MAX_255(c) && (cd->ctypes[c] & ctype_space) != 0) c = *(++ptr);
4562          if (c != CHAR_NUMBER_SIGN) break;
4563        ptr++;        ptr++;
4564        while (*ptr != 0)        while (*ptr != CHAR_NULL)
4565          {          {
4566          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr))         /* For non-fixed-length newline cases, */
4567              {                          /* IS_NEWLINE sets cd->nllen. */
4568              ptr += cd->nllen;
4569              break;
4570              }
4571          ptr++;          ptr++;
4572  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4573          if (utf) FORWARDCHAR(ptr);          if (utf) FORWARDCHAR(ptr);
4574  #endif  #endif
4575          }          }
4576        if (*ptr != 0) continue;        c = *ptr;     /* Either NULL or the char after a newline */
4577          }
4578        }
4579    
4580        /* Else fall through to handle end of string */    /* See if the next thing is a quantifier. */
4581        c = 0;  
4582        }    is_quantifier =
4583        c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4584        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4585    
4586      /* Fill in length of a previous callout, except when the next thing is a
4587      quantifier or when processing a property substitution string in UCP mode. */
4588    
4589      if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4590           after_manual_callout-- <= 0)
4591        {
4592        if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
4593          complete_callout(previous_callout, ptr, cd);
4594        previous_callout = NULL;
4595      }      }
4596    
4597    /* No auto callout for quantifiers. */    /* Create auto callout, except for quantifiers, or while processing property
4598      strings that are substituted for \w etc in UCP mode. */
4599    
4600    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4601      {      {
4602      previous_callout = code;      previous_callout = code;
4603      code = auto_callout(code, ptr, cd);      code = auto_callout(code, ptr, cd);
4604      }      }
4605    
4606      /* Process the next pattern item. */
4607    
4608    switch(c)    switch(c)
4609      {      {
4610      /* ===================================================================*/      /* ===================================================================*/
4611      case 0:                        /* The branch terminates at string end */      case CHAR_NULL:                /* The branch terminates at string end */
4612      case CHAR_VERTICAL_LINE:       /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
4613      case CHAR_RIGHT_PARENTHESIS:      case CHAR_RIGHT_PARENTHESIS:
4614      *firstcharptr = firstchar;      *firstcharptr = firstchar;
4615        *firstcharflagsptr = firstcharflags;
4616      *reqcharptr = reqchar;      *reqcharptr = reqchar;
4617        *reqcharflagsptr = reqcharflags;
4618      *codeptr = code;      *codeptr = code;
4619      *ptrptr = ptr;      *ptrptr = ptr;
4620      if (lengthptr != NULL)      if (lengthptr != NULL)
# Line 3970  for (;; ptr++) Line 4638  for (;; ptr++)
4638      previous = NULL;      previous = NULL;
4639      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
4640        {        {
4641        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;        if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4642        *code++ = OP_CIRCM;        *code++ = OP_CIRCM;
4643        }        }
4644      else *code++ = OP_CIRC;      else *code++ = OP_CIRC;
# Line 3985  for (;; ptr++) Line 4653  for (;; ptr++)
4653      repeats. The value of reqchar doesn't change either. */      repeats. The value of reqchar doesn't change either. */
4654    
4655      case CHAR_DOT:      case CHAR_DOT:
4656      if (firstchar == REQ_UNSET) firstchar = REQ_NONE;      if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4657      zerofirstchar = firstchar;      zerofirstchar = firstchar;
4658        zerofirstcharflags = firstcharflags;
4659      zeroreqchar = reqchar;      zeroreqchar = reqchar;
4660        zeroreqcharflags = reqcharflags;
4661      previous = code;      previous = code;
4662      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
4663      break;      break;
# Line 4061  for (;; ptr++) Line 4731  for (;; ptr++)
4731          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4732        {        {
4733        *code++ = negate_class? OP_ALLANY : OP_FAIL;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
4734        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;        if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4735        zerofirstchar = firstchar;        zerofirstchar = firstchar;
4736          zerofirstcharflags = firstcharflags;
4737        break;        break;
4738        }        }
4739    
# Line 4097  for (;; ptr++) Line 4768  for (;; ptr++)
4768      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
4769      loop, c contains the first byte of the character. */      loop, c contains the first byte of the character. */
4770    
4771      if (c != 0) do      if (c != CHAR_NULL) do
4772        {        {
4773        const pcre_uchar *oldptr;        const pcre_uchar *oldptr;
4774    
4775  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #ifdef SUPPORT_UTF
4776        if (utf && HAS_EXTRALEN(c))        if (utf && HAS_EXTRALEN(c))
4777          {                           /* Braces are required because the */          {                           /* Braces are required because the */
4778          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
# Line 4112  for (;; ptr++) Line 4783  for (;; ptr++)
4783        /* In the pre-compile phase, accumulate the length of any extra        /* In the pre-compile phase, accumulate the length of any extra
4784        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
4785        contain a zillion > 255 characters no longer overwrite the work space        contain a zillion > 255 characters no longer overwrite the work space
4786        (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,
4787        however. */        however. */
4788    
4789        if (lengthptr != NULL && class_uchardata > class_uchardata_base)        if (lengthptr != NULL && class_uchardata > class_uchardata_base)
# Line 4176  for (;; ptr++) Line 4847  for (;; ptr++)
4847          alpha. This relies on the fact that the class table starts with          alpha. This relies on the fact that the class table starts with
4848          alpha, lower, upper as the first 3 entries. */          alpha, lower, upper as the first 3 entries. */
4849    
4850          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
4851            posix_class = 0;            posix_class = 0;
4852    
4853          /* When PCRE_UCP is set, some of the POSIX classes are converted to          /* When PCRE_UCP is set, some of the POSIX classes are converted to
4854          different escape sequences that use Unicode properties. */          different escape sequences that use Unicode properties \p or \P. Others
4855            that are not available via \p or \P generate XCL_PROP/XCL_NOTPROP
4856            directly. */
4857    
4858  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4859          if ((options & PCRE_UCP) != 0)          if ((options & PCRE_UCP) != 0)
4860            {            {
4861              unsigned int ptype = 0;
4862            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
4863    
4864              /* The posix_substitutes table specifies which POSIX classes can be
4865              converted to \p or \P items. */
4866    
4867            if (posix_substitutes[pc] != NULL)            if (posix_substitutes[pc] != NULL)
4868              {              {
4869              nestptr = tempptr + 1;              nestptr = tempptr + 1;
4870              ptr = posix_substitutes[pc] - 1;              ptr = posix_substitutes[pc] - 1;
4871              continue;              continue;
4872              }              }
4873    
4874              /* There are three other classes that generate special property calls
4875              that are recognized only in an XCLASS. */
4876    
4877              else switch(posix_class)
4878                {
4879                case PC_GRAPH:
4880                ptype = PT_PXGRAPH;
4881                /* Fall through */
4882                case PC_PRINT:
4883                if (ptype == 0) ptype = PT_PXPRINT;
4884                /* Fall through */
4885                case PC_PUNCT:
4886                if (ptype == 0) ptype = PT_PXPUNCT;
4887                *class_uchardata++ = local_negate? XCL_NOTPROP : XCL_PROP;
4888                *class_uchardata++ = ptype;
4889                *class_uchardata++ = 0;
4890                ptr = tempptr + 1;
4891                continue;
4892    
4893                /* For all other POSIX classes, no special action is taken in UCP
4894                mode. Fall through to the non_UCP case. */
4895    
4896                default:
4897                break;
4898                }
4899            }            }
4900  #endif  #endif
4901          /* In the non-UCP case, we build the bit map for the POSIX class in a          /* In the non-UCP case, or when UCP makes no difference, we build the
4902          chunk of local store because we may be adding and subtracting from it,          bit map for the POSIX class in a chunk of local store because we may be
4903          and we don't want to subtract bits that may be in the main map already.          adding and subtracting from it, and we don't want to subtract bits that
4904          At the end we or the result into the bit map that is being built. */          may be in the main map already. At the end we or the result into the
4905            bit map that is being built. */
4906    
4907          posix_class *= 3;          posix_class *= 3;
4908    
# Line 4253  for (;; ptr++) Line 4958  for (;; ptr++)
4958    
4959        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
4960          {          {
4961          escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options, TRUE);          escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options,
4962              TRUE);
4963          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
4964            if (escape == 0) c = ec;
         if (escape == 0)  
           c = ec;  
4965          else if (escape == ESC_b) c = CHAR_BS; /* \b is backspace in a class */          else if (escape == ESC_b) c = CHAR_BS; /* \b is backspace in a class */
4966          else if (escape == ESC_N)            /* \N is not supported in a class */          else if (escape == ESC_N)          /* \N is not supported in a class */
4967            {            {
4968            *errorcodeptr = ERR71;            *errorcodeptr = ERR71;
4969            goto FAILED;            goto FAILED;
# Line 4316  for (;; ptr++) Line 5019  for (;; ptr++)
5019              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
5020              continue;              continue;
5021    
5022              /* 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
5023              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
5024              class. Luckily, the value of CHAR_VT is 0x0b in both ASCII and              previously set by something earlier in the character class.
5025              EBCDIC, so we lazily just adjust the appropriate bit. */              Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
5026                we could just adjust the appropriate bit. From PCRE 8.34 we no
5027                longer treat \s and \S specially. */
5028    
5029              case ESC_s:              case ESC_s:
5030              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];  
5031              continue;              continue;
5032    
5033              case ESC_S:              case ESC_S:
5034              should_flip_negation = TRUE;              should_flip_negation = TRUE;
5035              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 */  
5036              continue;              continue;
5037    
5038              /* The rest apply in both UCP and non-UCP cases. */              /* The rest apply in both UCP and non-UCP cases. */
5039    
5040              case ESC_h:              case ESC_h:
5041              (void)add_list_to_class(classbits, &class_uchardata, options, cd,              (void)add_list_to_class(classbits, &class_uchardata, options, cd,
5042                PRIV(hspace_list), NOTACHAR);                PRIV(hspace_list), NOTACHAR);
5043              continue;              continue;
5044    
5045              case ESC_H:              case ESC_H:
5046              (void)add_not_list_to_class(classbits, &class_uchardata, options,              (void)add_not_list_to_class(classbits, &class_uchardata, options,
5047                cd, PRIV(hspace_list));                cd, PRIV(hspace_list));
5048              continue;              continue;
5049    
5050              case ESC_v:              case ESC_v:
5051              (void)add_list_to_class(classbits, &class_uchardata, options, cd,              (void)add_list_to_class(classbits, &class_uchardata, options, cd,
5052                PRIV(vspace_list), NOTACHAR);                PRIV(vspace_list), NOTACHAR);
5053              continue;              continue;
5054    
5055              case ESC_V:              case ESC_V:
5056              (void)add_not_list_to_class(classbits, &class_uchardata, options,              (void)add_not_list_to_class(classbits, &class_uchardata, options,
5057                cd, PRIV(vspace_list));                cd, PRIV(vspace_list));
5058              continue;              continue;
5059    
5060  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
# Line 4360  for (;; ptr++) Line 5062  for (;; ptr++)
5062              case ESC_P:              case ESC_P:
5063                {                {
5064                BOOL negated;                BOOL negated;
5065                int pdata;                unsigned int ptype = 0, pdata = 0;
5066                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                if (!get_ucp(&ptr, &negated, &ptype, &pdata, errorcodeptr))
5067                if (ptype < 0) goto FAILED;                  goto FAILED;
5068                *class_uchardata++ = ((escape == ESC_p) != negated)?                *class_uchardata++ = ((escape == ESC_p) != negated)?
5069                  XCL_PROP : XCL_NOTPROP;                  XCL_PROP : XCL_NOTPROP;
5070                *class_uchardata++ = ptype;                *class_uchardata++ = ptype;
# Line 4390  for (;; ptr++) Line 5092  for (;; ptr++)
5092    
5093          /* Fall through if the escape just defined a single character (c >= 0).          /* Fall through if the escape just defined a single character (c >= 0).
5094          This may be greater than 256. */          This may be greater than 256. */
5095    
5096          escape = 0;          escape = 0;
5097    
5098          }   /* End of backslash handling */ &n