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

Diff of /code/trunk/pcre_compile.c

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

revision 1045 by ph10, Sun Sep 23 16:50:00 2012 UTC revision 1393 by ph10, Fri Nov 8 16:37:21 2013 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2012 University of Cambridge             Copyright (c) 1997-2013 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When PCRE_DEBUG is defined, we need the pcre(16)_printint() function, which  /* When PCRE_DEBUG is defined, we need the pcre(16|32)_printint() function, which
57  is also used by pcretest. PCRE_DEBUG is not defined when building a production  is also used by pcretest. PCRE_DEBUG is not defined when building a production
58  library. We do not need to select pcre16_printint.c specially, because the  library. We do not need to select pcre16_printint.c specially, because the
59  COMPILE_PCREx macro will already be appropriately set. */  COMPILE_PCREx macro will already be appropriately set. */
# Line 80  to check them every time. */ Line 80  to check them every time. */
80  /* Definitions to allow mutual recursion */  /* Definitions to allow mutual recursion */
81    
82  static int  static int
83    add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,    add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84      const pcre_uint32 *, unsigned int);      const pcre_uint32 *, unsigned int);
85    
86  static BOOL  static BOOL
87    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL,    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
88      int, int, int *, int *, branch_chain *, compile_data *, int *);      pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
89        compile_data *, int *);
90    
91    
92    
# Line 114  kicks in at the same number of forward r Line 115  kicks in at the same number of forward r
115  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
116  #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)  #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
117    
118    /* This value determines the size of the initial vector that is used for
119    remembering named groups during the pre-compile. It is allocated on the stack,
120    but if it is too small, it is expanded using malloc(), in a similar way to the
121    workspace. The value is the number of slots in the list. */
122    
123    #define NAMED_GROUP_LIST_SIZE  20
124    
125  /* The overrun tests check for a slightly smaller size so that they detect the  /* The overrun tests check for a slightly smaller size so that they detect the
126  overrun before it actually does run off the end of the data block. */  overrun before it actually does run off the end of the data block. */
127    
# Line 121  overrun before it actually does run off Line 129  overrun before it actually does run off
129    
130  /* Private flags added to firstchar and reqchar. */  /* Private flags added to firstchar and reqchar. */
131    
132  #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */  #define REQ_CASELESS    (1 << 0)        /* Indicates caselessness */
133  #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */  #define REQ_VARY        (1 << 1)        /* Reqchar followed non-literal item */
134    /* Negative values for the firstchar and reqchar flags */
135    #define REQ_UNSET       (-2)
136    #define REQ_NONE        (-1)
137    
138  /* Repeated character flags. */  /* Repeated character flags. */
139    
# Line 253  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 264  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 291  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 318  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 371  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 386  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 451  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 483  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 503  static const char error_texts[] = Line 525  static const char error_texts[] =
525    /* 75 */    /* 75 */
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"
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    ;    ;
537    
538  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 642  static const pcre_uint8 ebcdic_chartab[] Line 672  static const pcre_uint8 ebcdic_chartab[]
672  #endif  #endif
673    
674    
675    /* This table is used to check whether auto-possessification is possible
676    between adjacent character-type opcodes. The left-hand (repeated) opcode is
677    used to select the row, and the right-hand opcode is use to select the column.
678    A value of 1 means that auto-possessification is OK. For example, the second
679    value in the first row means that \D+\d can be turned into \D++\d.
680    
681    The Unicode property types (\P and \p) have to be present to fill out the table
682    because of what their opcode values are, but the table values should always be
683    zero because property types are handled separately in the code. The last four
684    columns apply to items that cannot be repeated, so there is no need to have
685    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
686    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
687    
688    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
689    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
690    
691    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
692    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
693      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
694      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
695      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
696      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
697      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
698      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
699      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
700      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 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 },  /* \C */
702      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
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, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
705      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
706      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
707      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
708      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
709      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
710    };
711    
712    
713    /* This table is used to check whether auto-possessification is possible
714    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
715    left-hand (repeated) opcode is used to select the row, and the right-hand
716    opcode is used to select the column. The values are as follows:
717    
718      0   Always return FALSE (never auto-possessify)
719      1   Character groups are distinct (possessify if both are OP_PROP)
720      2   Check character categories in the same group (general or particular)
721      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
722    
723      4   Check left general category vs right particular category
724      5   Check right general category vs left particular category
725    
726      6   Left alphanum vs right general category
727      7   Left space vs right general category
728      8   Left word vs right general category
729    
730      9   Right alphanum vs left general category
731     10   Right space vs left general category
732     11   Right word vs left general category
733    
734     12   Left alphanum vs right particular category
735     13   Left space vs right particular category
736     14   Left word vs right particular category
737    
738     15   Right alphanum vs left particular category
739     16   Right space vs left particular category
740     17   Right word vs left particular category
741    */
742    
743    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
744    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
745      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
746      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
747      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
748      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
749      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
750      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
751      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
752      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
753      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
754      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
755      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
756    };
757    
758    /* This table is used to check whether auto-possessification is possible
759    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
760    specifies a general category and the other specifies a particular category. The
761    row is selected by the general category and the column by the particular
762    category. The value is 1 if the particular category is not part of the general
763    category. */
764    
765    static const pcre_uint8 catposstab[7][30] = {
766    /* 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 */
767      { 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 */
768      { 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 */
769      { 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 */
770      { 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 */
771      { 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 */
772      { 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 */
773      { 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 */
774    };
775    
776    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
777    a general or particular category. The properties in each row are those
778    that apply to the character set in question. Duplication means that a little
779    unnecessary work is done when checking, but this keeps things much simpler
780    because they can all use the same code. For more details see the comment where
781    this table is used.
782    
783    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
784    "space", but from Perl 5.18 it's included, so both categories are treated the
785    same here. */
786    
787    static const pcre_uint8 posspropstab[3][4] = {
788      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
789      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
790      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
791    };
792    
793    /* This table is used when converting repeating opcodes into possessified
794    versions as a result of an explicit possessive quantifier such as ++. A zero
795    value means there is no possessified version - in those cases the item in
796    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
797    because all relevant opcodes are less than that. */
798    
799    static const pcre_uint8 opcode_possessify[] = {
800      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
801      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
802    
803      0,                       /* NOTI */
804      OP_POSSTAR, 0,           /* STAR, MINSTAR */
805      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
806      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
807      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
808      0,                       /* EXACT */
809      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
810    
811      OP_POSSTARI, 0,          /* STARI, MINSTARI */
812      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
813      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
814      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
815      0,                       /* EXACTI */
816      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
817    
818      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
819      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
820      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
821      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
822      0,                       /* NOTEXACT */
823      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
824    
825      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
826      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
827      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
828      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
829      0,                       /* NOTEXACTI */
830      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
831    
832      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
833      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
834      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
835      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
836      0,                       /* TYPEEXACT */
837      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
838    
839      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
840      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
841      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
842      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
843      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
844    
845      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
846      0, 0,                    /* REF, REFI */
847      0, 0,                    /* DNREF, DNREFI */
848      0, 0                     /* RECURSE, CALLOUT */
849    };
850    
851    
852    
853  /*************************************************  /*************************************************
# Line 663  find_error_text(int n) Line 869  find_error_text(int n)
869  const char *s = error_texts;  const char *s = error_texts;
870  for (; n > 0; n--)  for (; n > 0; n--)
871    {    {
872    while (*s++ != 0) {};    while (*s++ != CHAR_NULL) {};
873    if (*s == 0) return "Error text not found (please report)";    if (*s == CHAR_NULL) return "Error text not found (please report)";
874    }    }
875  return s;  return s;
876  }  }
877    
878    
879    
880  /*************************************************  /*************************************************
881  *           Expand the workspace                 *  *           Expand the workspace                 *
882  *************************************************/  *************************************************/
# Line 747  return (*p == CHAR_RIGHT_CURLY_BRACKET); Line 954  return (*p == CHAR_RIGHT_CURLY_BRACKET);
954  *************************************************/  *************************************************/
955    
956  /* 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
957  positive value for a simple escape such as \n, or a negative value which  positive value for a simple escape such as \n, or 0 for a data character which
958  encodes one of the more complicated things such as \d. A backreference to group  will be placed in chptr. A backreference to group n is returned as negative n.
959  n is returned as -(ESC_REF + n); ESC_REF is the highest ESC_xxx macro. When  When UTF-8 is enabled, a positive value greater than 255 may be returned in
960  UTF-8 is enabled, a positive value greater than 255 may be returned. On entry,  chptr. On entry, ptr is pointing at the \. On exit, it is on the final
961  ptr is pointing at the \. On exit, it is on the final character of the escape  character of the escape sequence.
 sequence.  
962    
963  Arguments:  Arguments:
964    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
965      chptr          points to a returned data character
966    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
967    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
968    options        the options bits    options        the options bits
969    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
970    
971  Returns:         zero or positive => a data character  Returns:         zero => a data character
972                   negative => a special escape sequence                   positive => a special escape sequence
973                     negative => a back reference
974                   on error, errorcodeptr is set                   on error, errorcodeptr is set
975  */  */
976    
977  static int  static int
978  check_escape(const pcre_uchar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
979    int options, BOOL isclass)    int bracount, int options, BOOL isclass)
980  {  {
981  /* PCRE_UTF16 has the same value as PCRE_UTF8. */  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
982  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
983  const pcre_uchar *ptr = *ptrptr + 1;  const pcre_uchar *ptr = *ptrptr + 1;
984  pcre_int32 c;  pcre_uint32 c;
985    int escape = 0;
986  int i;  int i;
987    
988  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
# Line 781  ptr--;                            /* Set Line 990  ptr--;                            /* Set
990    
991  /* 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. */
992    
993  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
994    
995  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
996  in a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
# Line 790  Otherwise further processing may be requ Line 999  Otherwise further processing may be requ
999  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1000  /* Not alphanumeric */  /* Not alphanumeric */
1001  else if (c < CHAR_0 || c > CHAR_z) {}  else if (c < CHAR_0 || c > CHAR_z) {}
1002  else if ((i = escapes[c - CHAR_0]) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0)
1003      { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1004    
1005  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1006  /* Not alphanumeric */  /* Not alphanumeric */
1007  else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}  else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
1008  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1009  #endif  #endif
1010    
1011  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
# Line 803  else if ((i = escapes[c - 0x48]) != 0) Line 1013  else if ((i = escapes[c - 0x48]) != 0)
1013  else  else
1014    {    {
1015    const pcre_uchar *oldptr;    const pcre_uchar *oldptr;
1016    BOOL braced, negated;    BOOL braced, negated, overflow;
1017      int s;
1018    
1019    switch (c)    switch (c)
1020      {      {
# Line 828  else Line 1039  else
1039          c = 0;          c = 0;
1040          for (i = 0; i < 4; ++i)          for (i = 0; i < 4; ++i)
1041            {            {
1042            register int cc = *(++ptr);            register pcre_uint32 cc = *(++ptr);
1043  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1044            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1045            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
# Line 838  else Line 1049  else
1049  #endif  #endif
1050            }            }
1051    
1052  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
1053          if (c > (utf ? 0x10ffff : 0xff))          if (c > (utf ? 0x10ffffU : 0xffU))
1054  #else  #elif defined COMPILE_PCRE16
1055  #ifdef COMPILE_PCRE16          if (c > (utf ? 0x10ffffU : 0xffffU))
1056          if (c > (utf ? 0x10ffff : 0xffff))  #elif defined COMPILE_PCRE32
1057  #endif          if (utf && c > 0x10ffffU)
1058  #endif  #endif
1059            {            {
1060            *errorcodeptr = ERR76;            *errorcodeptr = ERR76;
# Line 875  else Line 1086  else
1086      (3) For Oniguruma compatibility we also support \g followed by a name or a      (3) For Oniguruma compatibility we also support \g followed by a name or a
1087      number either in angle brackets or in single quotes. However, these are      number either in angle brackets or in single quotes. However, these are
1088      (possibly recursive) subroutine calls, _not_ backreferences. Just return      (possibly recursive) subroutine calls, _not_ backreferences. Just return
1089      the -ESC_g code (cf \k). */      the ESC_g code (cf \k). */
1090    
1091      case CHAR_g:      case CHAR_g:
1092      if (isclass) break;      if (isclass) break;
1093      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1094        {        {
1095        c = -ESC_g;        escape = ESC_g;
1096        break;        break;
1097        }        }
1098    
# Line 890  else Line 1101  else
1101      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1102        {        {
1103        const pcre_uchar *p;        const pcre_uchar *p;
1104        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++)
1105          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1106        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)        if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1107          {          {
1108          c = -ESC_k;          escape = ESC_k;
1109          break;          break;
1110          }          }
1111        braced = TRUE;        braced = TRUE;
# Line 910  else Line 1121  else
1121      else negated = FALSE;      else negated = FALSE;
1122    
1123      /* The integer range is limited by the machine's int representation. */      /* The integer range is limited by the machine's int representation. */
1124      c = 0;      s = 0;
1125        overflow = FALSE;
1126      while (IS_DIGIT(ptr[1]))      while (IS_DIGIT(ptr[1]))
1127        {        {
1128        if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */        if (s > INT_MAX / 10 - 1) /* Integer overflow */
1129          {          {
1130          c = -1;          overflow = TRUE;
1131          break;          break;
1132          }          }
1133        c = c * 10 + *(++ptr) - CHAR_0;        s = s * 10 + (int)(*(++ptr) - CHAR_0);
1134        }        }
1135      if (((unsigned int)c) > INT_MAX) /* Integer overflow */      if (overflow) /* Integer overflow */
1136        {        {
1137        while (IS_DIGIT(ptr[1]))        while (IS_DIGIT(ptr[1]))
1138          ptr++;          ptr++;
# Line 934  else Line 1146  else
1146        break;        break;
1147        }        }
1148    
1149      if (c == 0)      if (s == 0)
1150        {        {
1151        *errorcodeptr = ERR58;        *errorcodeptr = ERR58;
1152        break;        break;
# Line 942  else Line 1154  else
1154    
1155      if (negated)      if (negated)
1156        {        {
1157        if (c > bracount)        if (s > bracount)
1158          {          {
1159          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
1160          break;          break;
1161          }          }
1162        c = bracount - (c - 1);        s = bracount - (s - 1);
1163        }        }
1164    
1165      c = -(ESC_REF + c);      escape = -s;
1166      break;      break;
1167    
1168      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1169      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
1170      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1171        recommended to avoid the ambiguities in the old syntax.
1172    
1173      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
1174      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
1175      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
1176      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
1177      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
1178      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
1179      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1180    
1181        Inside a character class, \ followed by a digit is always either a literal
1182        8 or 9 or an octal number. */
1183    
1184      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:
1185      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
# Line 972  else Line 1188  else
1188        {        {
1189        oldptr = ptr;        oldptr = ptr;
1190        /* The integer range is limited by the machine's int representation. */        /* The integer range is limited by the machine's int representation. */
1191        c -= CHAR_0;        s = (int)(c -CHAR_0);
1192          overflow = FALSE;
1193        while (IS_DIGIT(ptr[1]))        while (IS_DIGIT(ptr[1]))
1194          {          {
1195          if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */          if (s > INT_MAX / 10 - 1) /* Integer overflow */
1196            {            {
1197            c = -1;            overflow = TRUE;
1198            break;            break;
1199            }            }
1200          c = c * 10 + *(++ptr) - CHAR_0;          s = s * 10 + (int)(*(++ptr) - CHAR_0);
1201          }          }
1202        if (((unsigned int)c) > INT_MAX) /* Integer overflow */        if (overflow) /* Integer overflow */
1203          {          {
1204          while (IS_DIGIT(ptr[1]))          while (IS_DIGIT(ptr[1]))
1205            ptr++;            ptr++;
1206          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1207          break;          break;
1208          }          }
1209        if (c < 10 || c <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1210          {          {
1211          c = -(ESC_REF + c);          escape = -s;
1212          break;          break;
1213          }          }
1214        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1215        }        }
1216    
1217      /* 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
1218      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
1219      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
1220        changed so as not to insert the binary zero. */
1221    
1222      if ((c = *ptr) >= CHAR_8)      if ((c = *ptr) >= CHAR_8) break;
1223        {  
1224        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1225    
1226      /* \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
1227      larger first octal digit. The original code used just to take the least      larger first octal digit. The original code used just to take the least
# Line 1023  else Line 1238  else
1238  #endif  #endif
1239      break;      break;
1240    
1241      /* \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
1242      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}. */
1243      If not, { is treated as a data character. */  
1244        case CHAR_o:
1245        if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1246          {
1247          ptr += 2;
1248          c = 0;
1249          overflow = FALSE;
1250          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1251            {
1252            register pcre_uint32 cc = *ptr++;
1253            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1254    #ifdef COMPILE_PCRE32
1255            if (c >= 0x20000000l) { overflow = TRUE; break; }
1256    #endif
1257            c = (c << 3) + cc - CHAR_0 ;
1258    #if defined COMPILE_PCRE8
1259            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1260    #elif defined COMPILE_PCRE16
1261            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1262    #elif defined COMPILE_PCRE32
1263            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1264    #endif
1265            }
1266          if (overflow)
1267            {
1268            while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1269            *errorcodeptr = ERR34;
1270            }
1271          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1272            {
1273            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1274            }
1275          else *errorcodeptr = ERR80;
1276          }
1277        break;
1278    
1279        /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1280        numbers. Otherwise it is a lowercase x letter. */
1281    
1282      case CHAR_x:      case CHAR_x:
1283      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1284        {        {
       /* In JavaScript, \x must be followed by two hexadecimal numbers.  
       Otherwise it is a lowercase x letter. */  
1285        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1286          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1287          {          {
1288          c = 0;          c = 0;
1289          for (i = 0; i < 2; ++i)          for (i = 0; i < 2; ++i)
1290            {            {
1291            register int cc = *(++ptr);            register pcre_uint32 cc = *(++ptr);
1292  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1293            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1294            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
# Line 1048  else Line 1298  else
1298  #endif  #endif
1299            }            }
1300          }          }
1301        break;        }    /* End JavaScript handling */
       }  
1302    
1303      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1304        {      greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1305        const pcre_uchar *pt = ptr + 2;      digits. If not, { used to be treated as a data character. However, Perl
1306        seems to read hex digits up to the first non-such, and ignore the rest, so
1307        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1308        now gives an error. */
1309    
1310        c = 0;      else
1311        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)        {
1312          if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1313          {          {
1314          register int cc = *pt++;          ptr += 2;
1315          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          c = 0;
1316            overflow = FALSE;
1317            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1318              {
1319              register pcre_uint32 cc = *ptr++;
1320              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1321    
1322    #ifdef COMPILE_PCRE32
1323              if (c >= 0x10000000l) { overflow = TRUE; break; }
1324    #endif
1325    
1326  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1327          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1328          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1329  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1330          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 */
1331          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1332  #endif  #endif
1333    
1334  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
1335          if (c > (utf ? 0x10ffff : 0xff)) { c = -1; break; }            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1336  #else  #elif defined COMPILE_PCRE16
1337  #ifdef COMPILE_PCRE16            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1338          if (c > (utf ? 0x10ffff : 0xffff)) { c = -1; break; }  #elif defined COMPILE_PCRE32
1339              if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1340  #endif  #endif
1341  #endif            }
         }  
1342    
1343        if (c < 0)          if (overflow)
1344          {            {
1345          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1346          *errorcodeptr = ERR34;            *errorcodeptr = ERR34;
1347          }            }
1348    
1349        if (*pt == CHAR_RIGHT_CURLY_BRACKET)          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1350          {            {
1351          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1352          ptr = pt;            }
         break;  
         }  
1353    
1354        /* 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.
1355        recognize this construct; fall through to the normal \x handling. */          We used just to recognize this construct and fall through to the normal
1356        }          \x handling, but nowadays Perl gives an error, which seems much more
1357            sensible, so we do too. */
1358    
1359      /* Read just a single-byte hex-defined char */          else *errorcodeptr = ERR79;
1360            }   /* End of \x{} processing */
1361    
1362      c = 0;        /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1363      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)  
1364        {        else
1365        int cc;                                  /* Some compilers don't like */          {
1366        cc = *(++ptr);                           /* ++ in initializers */          c = 0;
1367            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1368              {
1369              pcre_uint32 cc;                          /* Some compilers don't like */
1370              cc = *(++ptr);                           /* ++ in initializers */
1371  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1372        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1373        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1374  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1375        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */            if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1376        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1377  #endif  #endif
1378        }            }
1379            }     /* End of \xdd handling */
1380          }       /* End of Perl-style \x handling */
1381      break;      break;
1382    
1383      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
# Line 1119  else Line 1387  else
1387    
1388      case CHAR_c:      case CHAR_c:
1389      c = *(++ptr);      c = *(++ptr);
1390      if (c == 0)      if (c == CHAR_NULL)
1391        {        {
1392        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1393        break;        break;
# Line 1159  else Line 1427  else
1427  newline". PCRE does not support \N{name}. However, it does support  newline". PCRE does not support \N{name}. However, it does support
1428  quantification such as \N{2,3}. */  quantification such as \N{2,3}. */
1429    
1430  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&  if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1431       !is_counted_repeat(ptr+2))       !is_counted_repeat(ptr+2))
1432    *errorcodeptr = ERR37;    *errorcodeptr = ERR37;
1433    
1434  /* If PCRE_UCP is set, we change the values for \d etc. */  /* If PCRE_UCP is set, we change the values for \d etc. */
1435    
1436  if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)  if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1437    c -= (ESC_DU - ESC_D);    escape += (ESC_DU - ESC_D);
1438    
1439  /* Set the pointer to the final character before returning. */  /* Set the pointer to the final character before returning. */
1440    
1441  *ptrptr = ptr;  *ptrptr = ptr;
1442  return c;  *chptr = c;
1443    return escape;
1444  }  }
1445    
1446    
# Line 1189  escape sequence. Line 1458  escape sequence.
1458  Argument:  Argument:
1459    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1460    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
1461    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
1462      pdataptr       points to an unsigned int that is set to the detailed property value
1463    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1464    
1465  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
1466  */  */
1467    
1468  static int  static BOOL
1469  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1470      unsigned int *pdataptr, int *errorcodeptr)
1471  {  {
1472  int c, i, bot, top;  pcre_uchar c;
1473    int i, bot, top;
1474  const pcre_uchar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
1475  pcre_uchar name[32];  pcre_uchar name[32];
1476    
1477  c = *(++ptr);  c = *(++ptr);
1478  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1479    
1480  *negptr = FALSE;  *negptr = FALSE;
1481    
# Line 1220  if (c == CHAR_LEFT_CURLY_BRACKET) Line 1492  if (c == CHAR_LEFT_CURLY_BRACKET)
1492    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1493      {      {
1494      c = *(++ptr);      c = *(++ptr);
1495      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1496      if (c == CHAR_RIGHT_CURLY_BRACKET) break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1497      name[i] = c;      name[i] = c;
1498      }      }
# Line 1245  top = PRIV(utt_size); Line 1517  top = PRIV(utt_size);
1517    
1518  while (bot < top)  while (bot < top)
1519    {    {
1520      int r;
1521    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1522    c = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1523    if (c == 0)    if (r == 0)
1524      {      {
1525      *dptr = PRIV(utt)[i].value;      *ptypeptr = PRIV(utt)[i].type;
1526      return PRIV(utt)[i].type;      *pdataptr = PRIV(utt)[i].value;
1527        return TRUE;
1528      }      }
1529    if (c > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1530    }    }
1531    
1532  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1533  *ptrptr = ptr;  *ptrptr = ptr;
1534  return -1;  return FALSE;
1535    
1536  ERROR_RETURN:  ERROR_RETURN:
1537  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1538  *ptrptr = ptr;  *ptrptr = ptr;
1539  return -1;  return FALSE;
1540  }  }
1541  #endif  #endif
1542    
1543    
1544    
   
1545  /*************************************************  /*************************************************
1546  *         Read repeat counts                     *  *         Read repeat counts                     *
1547  *************************************************/  *************************************************/
# Line 1297  int max = -1; Line 1570  int max = -1;
1570  /* Read the minimum value and do a paranoid check: a negative value indicates  /* Read the minimum value and do a paranoid check: a negative value indicates
1571  an integer overflow. */  an integer overflow. */
1572    
1573  while (IS_DIGIT(*p)) min = min * 10 + *p++ - CHAR_0;  while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);
1574  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1575    {    {
1576    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 1312  if (*p == CHAR_RIGHT_CURLY_BRACKET) max Line 1585  if (*p == CHAR_RIGHT_CURLY_BRACKET) max
1585    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1586      {      {
1587      max = 0;      max = 0;
1588      while(IS_DIGIT(*p)) max = max * 10 + *p++ - CHAR_0;      while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);
1589      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1590        {        {
1591        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 1337  return p; Line 1610  return p;
1610    
1611    
1612  /*************************************************  /*************************************************
 *  Subroutine for finding forward reference      *  
 *************************************************/  
   
 /* This recursive function is called only from find_parens() below. The  
 top-level call starts at the beginning of the pattern. All other calls must  
 start at a parenthesis. It scans along a pattern's text looking for capturing  
 subpatterns, and counting them. If it finds a named pattern that matches the  
 name it is given, it returns its number. Alternatively, if the name is NULL, it  
 returns when it reaches a given numbered subpattern. Recursion is used to keep  
 track of subpatterns that reset the capturing group numbers - the (?| feature.  
   
 This function was originally called only from the second pass, in which we know  
 that if (?< or (?' or (?P< is encountered, the name will be correctly  
 terminated because that is checked in the first pass. There is now one call to  
 this function in the first pass, to check for a recursive back reference by  
 name (so that we can make the whole group atomic). In this case, we need check  
 only up to the current position in the pattern, and that is still OK because  
 and previous occurrences will have been checked. To make this work, the test  
 for "end of pattern" is a check against cd->end_pattern in the main loop,  
 instead of looking for a binary zero. This means that the special first-pass  
 call can adjust cd->end_pattern temporarily. (Checks for binary zero while  
 processing items within the loop are OK, because afterwards the main loop will  
 terminate.)  
   
 Arguments:  
   ptrptr       address of the current character pointer (updated)  
   cd           compile background data  
   name         name to seek, or NULL if seeking a numbered subpattern  
   lorn         name length, or subpattern number if name is NULL  
   xmode        TRUE if we are in /x mode  
   utf          TRUE if we are in UTF-8 / UTF-16 mode  
   count        pointer to the current capturing subpattern number (updated)  
   
 Returns:       the number of the named subpattern, or -1 if not found  
 */  
   
 static int  
 find_parens_sub(pcre_uchar **ptrptr, compile_data *cd, const pcre_uchar *name, int lorn,  
   BOOL xmode, BOOL utf, int *count)  
 {  
 pcre_uchar *ptr = *ptrptr;  
 int start_count = *count;  
 int hwm_count = start_count;  
 BOOL dup_parens = FALSE;  
   
 /* If the first character is a parenthesis, check on the type of group we are  
 dealing with. The very first call may not start with a parenthesis. */  
   
 if (ptr[0] == CHAR_LEFT_PARENTHESIS)  
   {  
   /* Handle specials such as (*SKIP) or (*UTF8) etc. */  
   
   if (ptr[1] == CHAR_ASTERISK) ptr += 2;  
   
   /* Handle a normal, unnamed capturing parenthesis. */  
   
   else if (ptr[1] != CHAR_QUESTION_MARK)  
     {  
     *count += 1;  
     if (name == NULL && *count == lorn) return *count;  
     ptr++;  
     }  
   
   /* All cases now have (? at the start. Remember when we are in a group  
   where the parenthesis numbers are duplicated. */  
   
   else if (ptr[2] == CHAR_VERTICAL_LINE)  
     {  
     ptr += 3;  
     dup_parens = TRUE;  
     }  
   
   /* Handle comments; all characters are allowed until a ket is reached. */  
   
   else if (ptr[2] == CHAR_NUMBER_SIGN)  
     {  
     for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;  
     goto FAIL_EXIT;  
     }  
   
   /* Handle a condition. If it is an assertion, just carry on so that it  
   is processed as normal. If not, skip to the closing parenthesis of the  
   condition (there can't be any nested parens). */  
   
   else if (ptr[2] == CHAR_LEFT_PARENTHESIS)  
     {  
     ptr += 2;  
     if (ptr[1] != CHAR_QUESTION_MARK)  
       {  
       while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;  
       if (*ptr != 0) ptr++;  
       }  
     }  
   
   /* Start with (? but not a condition. */  
   
   else  
     {  
     ptr += 2;  
     if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */  
   
     /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */  
   
     if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&  
         ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)  
       {  
       int term;  
       const pcre_uchar *thisname;  
       *count += 1;  
       if (name == NULL && *count == lorn) return *count;  
       term = *ptr++;  
       if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;  
       thisname = ptr;  
       while (*ptr != term) ptr++;  
       if (name != NULL && lorn == ptr - thisname &&  
           STRNCMP_UC_UC(name, thisname, lorn) == 0)  
         return *count;  
       term++;  
       }  
     }  
   }  
   
 /* Past any initial parenthesis handling, scan for parentheses or vertical  
 bars. Stop if we get to cd->end_pattern. Note that this is important for the  
 first-pass call when this value is temporarily adjusted to stop at the current  
 position. So DO NOT change this to a test for binary zero. */  
   
 for (; ptr < cd->end_pattern; ptr++)  
   {  
   /* Skip over backslashed characters and also entire \Q...\E */  
   
   if (*ptr == CHAR_BACKSLASH)  
     {  
     if (*(++ptr) == 0) goto FAIL_EXIT;  
     if (*ptr == CHAR_Q) for (;;)  
       {  
       while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};  
       if (*ptr == 0) goto FAIL_EXIT;  
       if (*(++ptr) == CHAR_E) break;  
       }  
     continue;  
     }  
   
   /* Skip over character classes; this logic must be similar to the way they  
   are handled for real. If the first character is '^', skip it. Also, if the  
   first few characters (either before or after ^) are \Q\E or \E we skip them  
   too. This makes for compatibility with Perl. Note the use of STR macros to  
   encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */  
   
   if (*ptr == CHAR_LEFT_SQUARE_BRACKET)  
     {  
     BOOL negate_class = FALSE;  
     for (;;)  
       {  
       if (ptr[1] == CHAR_BACKSLASH)  
         {  
         if (ptr[2] == CHAR_E)  
           ptr+= 2;  
         else if (STRNCMP_UC_C8(ptr + 2,  
                  STR_Q STR_BACKSLASH STR_E, 3) == 0)  
           ptr += 4;  
         else  
           break;  
         }  
       else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)  
         {  
         negate_class = TRUE;  
         ptr++;  
         }  
       else break;  
       }  
   
     /* If the next character is ']', it is a data character that must be  
     skipped, except in JavaScript compatibility mode. */  
   
     if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&  
         (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)  
       ptr++;  
   
     while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)  
       {  
       if (*ptr == 0) return -1;  
       if (*ptr == CHAR_BACKSLASH)  
         {  
         if (*(++ptr) == 0) goto FAIL_EXIT;  
         if (*ptr == CHAR_Q) for (;;)  
           {  
           while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};  
           if (*ptr == 0) goto FAIL_EXIT;  
           if (*(++ptr) == CHAR_E) break;  
           }  
         continue;  
         }  
       }  
     continue;  
     }  
   
   /* Skip comments in /x mode */  
   
   if (xmode && *ptr == CHAR_NUMBER_SIGN)  
     {  
     ptr++;  
     while (*ptr != 0)  
       {  
       if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }  
       ptr++;  
 #ifdef SUPPORT_UTF  
       if (utf) FORWARDCHAR(ptr);  
 #endif  
       }  
     if (*ptr == 0) goto FAIL_EXIT;  
     continue;  
     }  
   
   /* Check for the special metacharacters */  
   
   if (*ptr == CHAR_LEFT_PARENTHESIS)  
     {  
     int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);  
     if (rc > 0) return rc;  
     if (*ptr == 0) goto FAIL_EXIT;  
     }  
   
   else if (*ptr == CHAR_RIGHT_PARENTHESIS)  
     {  
     if (dup_parens && *count < hwm_count) *count = hwm_count;  
     goto FAIL_EXIT;  
     }  
   
   else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)  
     {  
     if (*count > hwm_count) hwm_count = *count;  
     *count = start_count;  
     }  
   }  
   
 FAIL_EXIT:  
 *ptrptr = ptr;  
 return -1;  
 }  
   
   
   
   
 /*************************************************  
 *       Find forward referenced subpattern       *  
 *************************************************/  
   
 /* This function scans along a pattern's text looking for capturing  
 subpatterns, and counting them. If it finds a named pattern that matches the  
 name it is given, it returns its number. Alternatively, if the name is NULL, it  
 returns when it reaches a given numbered subpattern. This is used for forward  
 references to subpatterns. We used to be able to start this scan from the  
 current compiling point, using the current count value from cd->bracount, and  
 do it all in a single loop, but the addition of the possibility of duplicate  
 subpattern numbers means that we have to scan from the very start, in order to  
 take account of such duplicates, and to use a recursive function to keep track  
 of the different types of group.  
   
 Arguments:  
   cd           compile background data  
   name         name to seek, or NULL if seeking a numbered subpattern  
   lorn         name length, or subpattern number if name is NULL  
   xmode        TRUE if we are in /x mode  
   utf          TRUE if we are in UTF-8 / UTF-16 mode  
   
 Returns:       the number of the found subpattern, or -1 if not found  
 */  
   
 static int  
 find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,  
   BOOL utf)  
 {  
 pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;  
 int count = 0;  
 int rc;  
   
 /* If the pattern does not start with an opening parenthesis, the first call  
 to find_parens_sub() will scan right to the end (if necessary). However, if it  
 does start with a parenthesis, find_parens_sub() will return when it hits the  
 matching closing parens. That is why we have to have a loop. */  
   
 for (;;)  
   {  
   rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);  
   if (rc > 0 || *ptr++ == 0) break;  
   }  
   
 return rc;  
 }  
   
   
   
   
 /*************************************************  
1613  *      Find first significant op code            *  *      Find first significant op code            *
1614  *************************************************/  *************************************************/
1615    
# Line 1670  for (;;) Line 1648  for (;;)
1648    
1649      case OP_CALLOUT:      case OP_CALLOUT:
1650      case OP_CREF:      case OP_CREF:
1651      case OP_NCREF:      case OP_DNCREF:
1652      case OP_RREF:      case OP_RREF:
1653      case OP_NRREF:      case OP_DNRREF:
1654      case OP_DEF:      case OP_DEF:
1655      code += PRIV(OP_lengths)[*code];      code += PRIV(OP_lengths)[*code];
1656      break;      break;
# Line 1686  for (;;) Line 1664  for (;;)
1664    
1665    
1666    
   
1667  /*************************************************  /*************************************************
1668  *        Find the fixed length of a branch       *  *        Find the fixed length of a branch       *
1669  *************************************************/  *************************************************/
# Line 1704  and doing the check at the end; a flag s Line 1681  and doing the check at the end; a flag s
1681    
1682  Arguments:  Arguments:
1683    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1684    utf      TRUE in UTF-8 / UTF-16 mode    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1685    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1686    cd       the "compile data" structure    cd       the "compile data" structure
1687    
# Line 1730  for (;;) Line 1707  for (;;)
1707    {    {
1708    int d;    int d;
1709    pcre_uchar *ce, *cs;    pcre_uchar *ce, *cs;
1710    register int op = *cc;    register pcre_uchar op = *cc;
1711    
1712    switch (op)    switch (op)
1713      {      {
# Line 1810  for (;;) Line 1787  for (;;)
1787      case OP_COMMIT:      case OP_COMMIT:
1788      case OP_CREF:      case OP_CREF:
1789      case OP_DEF:      case OP_DEF:
1790        case OP_DNCREF:
1791        case OP_DNRREF:
1792      case OP_DOLL:      case OP_DOLL:
1793      case OP_DOLLM:      case OP_DOLLM:
1794      case OP_EOD:      case OP_EOD:
1795      case OP_EODN:      case OP_EODN:
1796      case OP_FAIL:      case OP_FAIL:
     case OP_NCREF:  
     case OP_NRREF:  
1797      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1798      case OP_PRUNE:      case OP_PRUNE:
1799      case OP_REVERSE:      case OP_REVERSE:
# Line 1850  for (;;) Line 1827  for (;;)
1827      case OP_EXACTI:      case OP_EXACTI:
1828      case OP_NOTEXACT:      case OP_NOTEXACT:
1829      case OP_NOTEXACTI:      case OP_NOTEXACTI:
1830      branchlength += GET2(cc,1);      branchlength += (int)GET2(cc,1);
1831      cc += 2 + IMM2_SIZE;      cc += 2 + IMM2_SIZE;
1832  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
1833      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
# Line 1859  for (;;) Line 1836  for (;;)
1836    
1837      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1838      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1839      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP) cc += 2;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1840          cc += 2;
1841      cc += 1 + IMM2_SIZE + 1;      cc += 1 + IMM2_SIZE + 1;
1842      break;      break;
1843    
# Line 1894  for (;;) Line 1872  for (;;)
1872    
1873      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1874    
 #if defined SUPPORT_UTF || defined COMPILE_PCRE16  
     case OP_XCLASS:  
     cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];  
     /* Fall through */  
 #endif  
   
1875      case OP_CLASS:      case OP_CLASS:
1876      case OP_NCLASS:      case OP_NCLASS:
1877    #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1878        case OP_XCLASS:
1879        /* The original code caused an unsigned overflow in 64 bit systems,
1880        so now we use a conditional statement. */
1881        if (op == OP_XCLASS)
1882          cc += GET(cc, 1);
1883        else
1884          cc += PRIV(OP_lengths)[OP_CLASS];
1885    #else
1886      cc += PRIV(OP_lengths)[OP_CLASS];      cc += PRIV(OP_lengths)[OP_CLASS];
1887    #endif
1888    
1889      switch (*cc)      switch (*cc)
1890        {        {
       case OP_CRPLUS:  
       case OP_CRMINPLUS:  
1891        case OP_CRSTAR:        case OP_CRSTAR:
1892        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1893          case OP_CRPLUS:
1894          case OP_CRMINPLUS:
1895        case OP_CRQUERY:        case OP_CRQUERY:
1896        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1897          case OP_CRPOSSTAR:
1898          case OP_CRPOSPLUS:
1899          case OP_CRPOSQUERY:
1900        return -1;        return -1;
1901    
1902        case OP_CRRANGE:        case OP_CRRANGE:
1903        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1904          case OP_CRPOSRANGE:
1905        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1906        branchlength += GET2(cc,1);        branchlength += (int)GET2(cc,1);
1907        cc += 1 + 2 * IMM2_SIZE;        cc += 1 + 2 * IMM2_SIZE;
1908        break;        break;
1909    
# Line 1984  for (;;) Line 1970  for (;;)
1970      case OP_QUERYI:      case OP_QUERYI:
1971      case OP_REF:      case OP_REF:
1972      case OP_REFI:      case OP_REFI:
1973        case OP_DNREF:
1974        case OP_DNREFI:
1975      case OP_SBRA:      case OP_SBRA:
1976      case OP_SBRAPOS:      case OP_SBRAPOS:
1977      case OP_SCBRA:      case OP_SCBRA:
# Line 2020  for (;;) Line 2008  for (;;)
2008    
2009    
2010    
   
2011  /*************************************************  /*************************************************
2012  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
2013  *************************************************/  *************************************************/
# Line 2033  length. Line 2020  length.
2020    
2021  Arguments:  Arguments:
2022    code        points to start of expression    code        points to start of expression
2023    utf         TRUE in UTF-8 / UTF-16 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2024    number      the required bracket number or negative to find a lookbehind    number      the required bracket number or negative to find a lookbehind
2025    
2026  Returns:      pointer to the opcode for the bracket, or NULL if not found  Returns:      pointer to the opcode for the bracket, or NULL if not found
# Line 2044  PRIV(find_bracket)(const pcre_uchar *cod Line 2031  PRIV(find_bracket)(const pcre_uchar *cod
2031  {  {
2032  for (;;)  for (;;)
2033    {    {
2034    register int c = *code;    register pcre_uchar c = *code;
2035    
2036    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2037    
# Line 2067  for (;;) Line 2054  for (;;)
2054    else if (c == OP_CBRA || c == OP_SCBRA ||    else if (c == OP_CBRA || c == OP_SCBRA ||
2055             c == OP_CBRAPOS || c == OP_SCBRAPOS)             c == OP_CBRAPOS || c == OP_SCBRAPOS)
2056      {      {
2057      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
2058      if (n == number) return (pcre_uchar *)code;      if (n == number) return (pcre_uchar *)code;
2059      code += PRIV(OP_lengths)[c];      code += PRIV(OP_lengths)[c];
2060      }      }
# Line 2097  for (;;) Line 2084  for (;;)
2084        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2085        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2086        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2087        if (code[1 + IMM2_SIZE] == OP_PROP        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2088          || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;          code += 2;
2089        break;        break;
2090    
2091        case OP_MARK:        case OP_MARK:
2092        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2093        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2094        case OP_THEN_ARG:        case OP_THEN_ARG:
2095        code += code[1];        code += code[1];
2096        break;        break;
# Line 2120  for (;;) Line 2104  for (;;)
2104    a multi-byte character. The length in the table is a minimum, so we have to    a multi-byte character. The length in the table is a minimum, so we have to
2105    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2106    
2107  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2108      if (utf) switch(c)      if (utf) switch(c)
2109        {        {
2110        case OP_CHAR:        case OP_CHAR:
# Line 2172  instance of OP_RECURSE. Line 2156  instance of OP_RECURSE.
2156    
2157  Arguments:  Arguments:
2158    code        points to start of expression    code        points to start of expression
2159    utf         TRUE in UTF-8 / UTF-16 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2160    
2161  Returns:      pointer to the opcode for OP_RECURSE, or NULL if not found  Returns:      pointer to the opcode for OP_RECURSE, or NULL if not found
2162  */  */
# Line 2182  find_recurse(const pcre_uchar *code, BOO Line 2166  find_recurse(const pcre_uchar *code, BOO
2166  {  {
2167  for (;;)  for (;;)
2168    {    {
2169    register int c = *code;    register pcre_uchar c = *code;
2170    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2171    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2172    
# Line 2217  for (;;) Line 2201  for (;;)
2201        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2202        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2203        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2204        if (code[1 + IMM2_SIZE] == OP_PROP        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2205          || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;          code += 2;
2206        break;        break;
2207    
2208        case OP_MARK:        case OP_MARK:
2209        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2210        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2211        case OP_THEN_ARG:        case OP_THEN_ARG:
2212        code += code[1];        code += code[1];
2213        break;        break;
# Line 2240  for (;;) Line 2221  for (;;)
2221      by a multi-byte character. The length in the table is a minimum, so we have      by a multi-byte character. The length in the table is a minimum, so we have
2222      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2223    
2224  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2225      if (utf) switch(c)      if (utf) switch(c)
2226        {        {
2227        case OP_CHAR:        case OP_CHAR:
# Line 2326  bracket whose current branch will alread Line 2307  bracket whose current branch will alread
2307  Arguments:  Arguments:
2308    code        points to start of search    code        points to start of search
2309    endcode     points to where to stop    endcode     points to where to stop
2310    utf         TRUE if in UTF-8 / UTF-16 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2311    cd          contains pointers to tables etc.    cd          contains pointers to tables etc.
2312      recurses    chain of recurse_check to catch mutual recursion
2313    
2314  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2315  */  */
2316    
2317    typedef struct recurse_check {
2318      struct recurse_check *prev;
2319      const pcre_uchar *group;
2320    } recurse_check;
2321    
2322  static BOOL  static BOOL
2323  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2324    BOOL utf, compile_data *cd)    BOOL utf, compile_data *cd, recurse_check *recurses)
2325  {  {
2326  register int c;  register pcre_uchar c;
2327    recurse_check this_recurse;
2328    
2329  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2330       code < endcode;       code < endcode;
2331       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
# Line 2364  for (code = first_significant_code(code Line 2353  for (code = first_significant_code(code
2353    
2354    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2355      {      {
2356      const pcre_uchar *scode;      const pcre_uchar *scode = cd->start_code + GET(code, 1);
2357      BOOL empty_branch;      BOOL empty_branch;
2358    
2359      /* Test for forward reference */      /* Test for forward reference or uncompleted reference. This is disabled
2360        when called to scan a completed pattern by setting cd->start_workspace to
2361        NULL. */
2362    
2363        if (cd->start_workspace != NULL)
2364          {
2365          const pcre_uchar *tcode;
2366          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2367            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2368          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2369          }
2370    
2371        /* If we are scanning a completed pattern, there are no forward references
2372        and all groups are complete. We need to detect whether this is a recursive
2373        call, as otherwise there will be an infinite loop. If it is a recursion,
2374        just skip over it. Simple recursions are easily detected. For mutual
2375        recursions we keep a chain on the stack. */
2376    
2377      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)      else
2378        if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;        {
2379          recurse_check *r = recurses;
2380          const pcre_uchar *endgroup = scode;
2381    
2382      /* Not a forward reference, test for completed backward reference */        do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2383          if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2384    
2385      empty_branch = FALSE;        for (r = recurses; r != NULL; r = r->prev)
2386      scode = cd->start_code + GET(code, 1);          if (r->group == scode) break;
2387      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */        if (r != NULL) continue;   /* Mutual recursion */
2388          }
2389    
2390        /* Completed reference; scan the referenced group, remembering it on the
2391        stack chain to detect mutual recursions. */
2392    
2393      /* Completed backwards reference */      empty_branch = FALSE;
2394        this_recurse.prev = recurses;
2395        this_recurse.group = scode;
2396    
2397      do      do
2398        {        {
2399        if (could_be_empty_branch(scode, endcode, utf, cd))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2400          {          {
2401          empty_branch = TRUE;          empty_branch = TRUE;
2402          break;          break;
# Line 2438  for (code = first_significant_code(code Line 2452  for (code = first_significant_code(code
2452        empty_branch = FALSE;        empty_branch = FALSE;
2453        do        do
2454          {          {
2455          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2456            empty_branch = TRUE;            empty_branch = TRUE;
2457          code += GET(code, 1);          code += GET(code, 1);
2458          }          }
# Line 2480  for (code = first_significant_code(code Line 2494  for (code = first_significant_code(code
2494        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2495        case OP_CRQUERY:        case OP_CRQUERY:
2496        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2497          case OP_CRPOSSTAR:
2498          case OP_CRPOSQUERY:
2499        break;        break;
2500    
2501        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2502        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2503        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2504          case OP_CRPOSPLUS:
2505        return FALSE;        return FALSE;
2506    
2507        case OP_CRRANGE:        case OP_CRRANGE:
2508        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2509          case OP_CRPOSRANGE:
2510        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2511        break;        break;
2512        }        }
# Line 2496  for (code = first_significant_code(code Line 2514  for (code = first_significant_code(code
2514    
2515      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2516    
2517        case OP_ANY:
2518        case OP_ALLANY:
2519        case OP_ANYBYTE:
2520    
2521      case OP_PROP:      case OP_PROP:
2522      case OP_NOTPROP:      case OP_NOTPROP:
2523        case OP_ANYNL:
2524    
2525        case OP_NOT_HSPACE:
2526        case OP_HSPACE:
2527        case OP_NOT_VSPACE:
2528        case OP_VSPACE:
2529      case OP_EXTUNI:      case OP_EXTUNI:
2530    
2531      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2532      case OP_DIGIT:      case OP_DIGIT:
2533      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2534      case OP_WHITESPACE:      case OP_WHITESPACE:
2535      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2536      case OP_WORDCHAR:      case OP_WORDCHAR:
2537      case OP_ANY:  
     case OP_ALLANY:  
     case OP_ANYBYTE:  
2538      case OP_CHAR:      case OP_CHAR:
2539      case OP_CHARI:      case OP_CHARI:
2540      case OP_NOT:      case OP_NOT:
2541      case OP_NOTI:      case OP_NOTI:
2542    
2543      case OP_PLUS:      case OP_PLUS:
2544        case OP_PLUSI:
2545      case OP_MINPLUS:      case OP_MINPLUS:
2546      case OP_POSPLUS:      case OP_MINPLUSI:
2547      case OP_EXACT:  
2548      case OP_NOTPLUS:      case OP_NOTPLUS:
2549        case OP_NOTPLUSI:
2550      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2551        case OP_NOTMINPLUSI:
2552    
2553        case OP_POSPLUS:
2554        case OP_POSPLUSI:
2555      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2556        case OP_NOTPOSPLUSI:
2557    
2558        case OP_EXACT:
2559        case OP_EXACTI:
2560      case OP_NOTEXACT:      case OP_NOTEXACT:
2561        case OP_NOTEXACTI:
2562    
2563      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2564      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2565      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2566      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2567    
2568      return FALSE;      return FALSE;
2569    
2570      /* These are going to continue, as they may be empty, but we have to      /* These are going to continue, as they may be empty, but we have to
# Line 2543  for (code = first_significant_code(code Line 2584  for (code = first_significant_code(code
2584      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2585      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2586      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2587      if (code[1 + IMM2_SIZE] == OP_PROP      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2588        || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;        code += 2;
2589      break;      break;
2590    
2591      /* End of branch */      /* End of branch */
# Line 2557  for (code = first_significant_code(code Line 2598  for (code = first_significant_code(code
2598      return TRUE;      return TRUE;
2599    
2600      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2601      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2602        followed by a multibyte character. */
2603    
2604  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2605      case OP_STAR:      case OP_STAR:
2606      case OP_STARI:      case OP_STARI:
2607        case OP_NOTSTAR:
2608        case OP_NOTSTARI:
2609    
2610      case OP_MINSTAR:      case OP_MINSTAR:
2611      case OP_MINSTARI:      case OP_MINSTARI:
2612        case OP_NOTMINSTAR:
2613        case OP_NOTMINSTARI:
2614    
2615      case OP_POSSTAR:      case OP_POSSTAR:
2616      case OP_POSSTARI:      case OP_POSSTARI:
2617        case OP_NOTPOSSTAR:
2618        case OP_NOTPOSSTARI:
2619    
2620      case OP_QUERY:      case OP_QUERY:
2621      case OP_QUERYI:      case OP_QUERYI:
2622        case OP_NOTQUERY:
2623        case OP_NOTQUERYI:
2624    
2625      case OP_MINQUERY:      case OP_MINQUERY:
2626      case OP_MINQUERYI:      case OP_MINQUERYI:
2627        case OP_NOTMINQUERY:
2628        case OP_NOTMINQUERYI:
2629    
2630      case OP_POSQUERY:      case OP_POSQUERY:
2631      case OP_POSQUERYI:      case OP_POSQUERYI:
2632        case OP_NOTPOSQUERY:
2633        case OP_NOTPOSQUERYI:
2634    
2635      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2636      break;      break;
2637    
2638      case OP_UPTO:      case OP_UPTO:
2639      case OP_UPTOI:      case OP_UPTOI:
2640        case OP_NOTUPTO:
2641        case OP_NOTUPTOI:
2642    
2643      case OP_MINUPTO:      case OP_MINUPTO:
2644      case OP_MINUPTOI:      case OP_MINUPTOI:
2645        case OP_NOTMINUPTO:
2646        case OP_NOTMINUPTOI:
2647    
2648      case OP_POSUPTO:      case OP_POSUPTO:
2649      case OP_POSUPTOI:      case OP_POSUPTOI:
2650        case OP_NOTPOSUPTO:
2651        case OP_NOTPOSUPTOI:
2652    
2653      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]);
2654      break;      break;
2655  #endif  #endif
# Line 2591  for (code = first_significant_code(code Line 2660  for (code = first_significant_code(code
2660      case OP_MARK:      case OP_MARK:
2661      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
2662      case OP_SKIP_ARG:      case OP_SKIP_ARG:
     code += code[1];  
     break;  
   
2663      case OP_THEN_ARG:      case OP_THEN_ARG:
2664      code += code[1];      code += code[1];
2665      break;      break;
# Line 2625  Arguments: Line 2691  Arguments:
2691    code        points to start of the recursion    code        points to start of the recursion
2692    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2693    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2694    utf         TRUE if in UTF-8 / UTF-16 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2695    cd          pointers to tables etc    cd          pointers to tables etc
2696    
2697  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
# Line 2637  could_be_empty(const pcre_uchar *code, c Line 2703  could_be_empty(const pcre_uchar *code, c
2703  {  {
2704  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2705    {    {
2706    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2707      return FALSE;      return FALSE;
2708    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2709    }    }
# Line 2647  return TRUE; Line 2713  return TRUE;
2713    
2714    
2715  /*************************************************  /*************************************************
2716  *           Check for POSIX class syntax         *  *        Base opcode of repeated opcodes         *
2717  *************************************************/  *************************************************/
2718    
2719  /* This function is called when the sequence "[:" or "[." or "[=" is  /* Returns the base opcode for repeated single character type opcodes. If the
2720  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  
2721    
2722  Returns:   TRUE or FALSE  Arguments:  c opcode
2723    Returns:    base opcode for the type
2724  */  */
2725    
2726  static BOOL  static pcre_uchar
2727  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)  get_repeat_base(pcre_uchar c)
2728  {  {
2729  int terminator;          /* Don't combine these lines; the Solaris cc */  return (c > OP_TYPEPOSUPTO)? c :
2730  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */         (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2731  for (++ptr; *ptr != 0; ptr++)         (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2732    {         (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2733    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)         (c >= OP_STARI)?      OP_STARI :
2734      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;  
2735  }  }
2736    
2737    
2738    
2739    #ifdef SUPPORT_UCP
2740  /*************************************************  /*************************************************
2741  *          Check POSIX class name                *  *        Check a character and a property        *
2742  *************************************************/  *************************************************/
2743    
2744  /* 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
2745  such as [:alnum:].  is adjacent to a fixed character.
2746    
2747  Arguments:  Arguments:
2748    ptr        points to the first letter    c            the character
2749    len        the length of the name    ptype        the property type
2750      pdata        the data for the type
2751      negated      TRUE if it's a negated property (\P or \p{^)
2752    
2753  Returns:     a value representing the name, or -1 if unknown  Returns:       TRUE if auto-possessifying is OK
2754  */  */
2755    
2756  static int  static BOOL
2757  check_posix_name(const pcre_uchar *ptr, int len)  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2758      BOOL negated)
2759  {  {
2760  const char *pn = posix_names;  const pcre_uint32 *p;
2761  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   *  
 *************************************************/  
2762    
2763  /* OP_RECURSE items contain an offset from the start of the regex to the group  switch(ptype)
2764  that is referenced. This means that groups can be replicated for fixed    {
2765  repetition simply by copying (because the recursion is allowed to refer to    case PT_LAMP:
2766  earlier groups that are outside the current group). However, when a group is    return (prop->chartype == ucp_Lu ||
2767  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is            prop->chartype == ucp_Ll ||
2768  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.  
2769    
2770  This function has been extended with the possibility of forward references for    case PT_GC:
2771  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).  
2772    
2773  Arguments:    case PT_PC:
2774    group      points to the start of the group    return (pdata == prop->chartype) == negated;
   adjust     the amount by which the group is to be moved  
   utf        TRUE in UTF-8 / UTF-16 mode  
   cd         contains pointers to tables etc.  
   save_hwm   the hwm forward reference pointer at the start of the group  
2775    
2776  Returns:     nothing    case PT_SC:
2777  */    return (pdata == prop->script) == negated;
2778    
2779  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;  
2780    
2781  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)    case PT_ALNUM:
2782    {    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2783    int offset;            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
   pcre_uchar *hc;  
2784    
2785    /* 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
2786    reference. */    means that Perl space and POSIX space are now identical. PCRE was changed
2787      at release 8.34. */
2788    
2789    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    case PT_SPACE:    /* Perl space */
2790      case PT_PXSPACE:  /* POSIX space */
2791      switch(c)
2792      {      {
2793      offset = GET(hc, 0);      HSPACE_CASES:
2794      if (cd->start_code + offset == ptr + 1)      VSPACE_CASES:
2795        {      return negated;
2796        PUT(hc, 0, offset + adjust);  
2797        break;      default:
2798        }      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2799      }      }
2800      break;  /* Control never reaches here */
2801    
2802    /* Otherwise, adjust the recursion offset if it's after the start of this    case PT_WORD:
2803    group. */    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2804              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2805              c == CHAR_UNDERSCORE) == negated;
2806    
2807    if (hc >= cd->hwm)    case PT_CLIST:
2808      p = PRIV(ucd_caseless_sets) + prop->caseset;
2809      for (;;)
2810      {      {
2811      offset = GET(ptr, 1);      if (c < *p) return !negated;
2812      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      if (c == *p++) return negated;
2813      }      }
2814      break;  /* Control never reaches here */
   ptr += 1 + LINK_SIZE;  
2815    }    }
2816    
2817    return FALSE;
2818  }  }
2819    #endif  /* SUPPORT_UCP */
2820    
2821    
2822    
2823  /*************************************************  /*************************************************
2824  *        Insert an automatic callout point       *  *        Fill the character property list        *
2825  *************************************************/  *************************************************/
2826    
2827  /* 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-
2828  callout points before each pattern item.  possessification, and if so, fills a list with its properties.
2829    
2830  Arguments:  Arguments:
2831    code           current code pointer    code        points to start of expression
2832    ptr            current pattern pointer    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2833    cd             pointers to tables etc    fcc         points to case-flipping table
2834      list        points to output list
2835                  list[0] will be filled with the opcode
2836                  list[1] will be non-zero if this opcode
2837                    can match an empty character string
2838                  list[2..7] depends on the opcode
2839    
2840  Returns:         new code pointer  Returns:      points to the start of the next opcode if *code is accepted
2841                  NULL if *code is not accepted
2842  */  */
2843    
2844  static pcre_uchar *  static const pcre_uchar *
2845  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)  get_chr_property_list(const pcre_uchar *code, BOOL utf,
2846      const pcre_uint8 *fcc, pcre_uint32 *list)
2847  {  {
2848  *code++ = OP_CALLOUT;  pcre_uchar c = *code;
2849  *code++ = 255;  const pcre_uchar *end;
2850  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */  const pcre_uint32 *clist_src;
2851  PUT(code, LINK_SIZE, 0);                       /* Default length */  pcre_uint32 *clist_dest;
2852  return code + 2 * LINK_SIZE;  pcre_uint32 chr;
2853  }  pcre_uchar base;
2854    
2855    list[0] = c;
2856    list[1] = FALSE;
2857    code++;
2858    
2859    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2860      {
2861      base = get_repeat_base(c);
2862      c -= (base - OP_STAR);
2863    
2864  /*************************************************    if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2865  *         Complete a callout item                *      code += IMM2_SIZE;
 *************************************************/  
2866    
2867  /* 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.  
2868    
2869  Arguments:    switch(base)
2870    previous_callout   points to previous callout item      {
2871    ptr                current pattern pointer      case OP_STAR:
2872    cd                 pointers to tables etc      list[0] = OP_CHAR;
2873        break;
2874    
2875  Returns:             nothing      case OP_STARI:
2876  */      list[0] = OP_CHARI;
2877        break;
2878    
2879  static void      case OP_NOTSTAR:
2880  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)      list[0] = OP_NOT;
2881  {      break;
 int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));  
 PUT(previous_callout, 2 + LINK_SIZE, length);  
 }  
2882    
2883        case OP_NOTSTARI:
2884        list[0] = OP_NOTI;
2885        break;
2886    
2887        case OP_TYPESTAR:
2888        list[0] = *code;
2889        code++;
2890        break;
2891        }
2892      c = list[0];
2893      }
2894    
2895  #ifdef SUPPORT_UCP  switch(c)
2896  /*************************************************    {
2897  *           Get othercase range                  *    case OP_NOT_DIGIT:
2898  *************************************************/    case OP_DIGIT:
2899      case OP_NOT_WHITESPACE:
2900      case OP_WHITESPACE:
2901      case OP_NOT_WORDCHAR:
2902      case OP_WORDCHAR:
2903      case OP_ANY:
2904      case OP_ALLANY:
2905      case OP_ANYNL:
2906      case OP_NOT_HSPACE:
2907      case OP_HSPACE:
2908      case OP_NOT_VSPACE:
2909      case OP_VSPACE:
2910      case OP_EXTUNI:
2911      case OP_EODN:
2912      case OP_EOD:
2913      case OP_DOLL:
2914      case OP_DOLLM:
2915      return code;
2916    
2917  /* This function is passed the start and end of a class range, in UTF-8 mode    case OP_CHAR:
2918  with UCP support. It searches up the characters, looking for ranges of    case OP_NOT:
2919  characters in the "other" case. Each call returns the next one, updating the    GETCHARINCTEST(chr, code);
2920  start address. A character with multiple other cases is returned on its own    list[2] = chr;
2921  with a special return value.    list[3] = NOTACHAR;
2922      return code;
2923    
2924  Arguments:    case OP_CHARI:
2925    cptr        points to starting character value; updated    case OP_NOTI:
2926    d           end value    list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2927    ocptr       where to put start of othercase range    GETCHARINCTEST(chr, code);
2928    odptr       where to put end of othercase range    list[2] = chr;
2929    
2930  Yield:        -1 when no more  #ifdef SUPPORT_UCP
2931                 0 when a range is returned    if (chr < 128 || (chr < 256 && !utf))
2932                >0 the CASESET offset for char with multiple other cases      list[3] = fcc[chr];
2933                  in this case, ocptr contains the original    else
2934  */      list[3] = UCD_OTHERCASE(chr);
2935    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2936      list[3] = (chr < 256) ? fcc[chr] : chr;
2937    #else
2938      list[3] = fcc[chr];
2939    #endif
2940    
2941  static int    /* The othercase might be the same value. */
 get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,  
   unsigned int *odptr)  
 {  
 unsigned int c, othercase, next;  
 int co;  
2942    
2943  /* Find the first character that has an other case. If it has multiple other    if (chr == list[3])
2944  cases, return its case offset value. */      list[3] = NOTACHAR;
2945      else
2946        list[4] = NOTACHAR;
2947      return code;
2948    
2949  for (c = *cptr; c <= d; c++)  #ifdef SUPPORT_UCP
2950    {    case OP_PROP:
2951    if ((co = UCD_CASESET(c)) != 0)    case OP_NOTPROP:
2952      if (code[0] != PT_CLIST)
2953      {      {
2954      *ocptr = c++;   /* Character that has the set */      list[2] = code[0];
2955      *cptr = c;      /* Rest of input range */      list[3] = code[1];
2956      return co;      return code + 2;
2957      }      }
   if ((othercase = UCD_OTHERCASE(c)) != c) break;  
   }  
2958    
2959  if (c > d) return -1;  /* Reached end of range */    /* Convert only if we have enough space. */
2960    
2961  *ocptr = othercase;    clist_src = PRIV(ucd_caseless_sets) + code[1];
2962  next = othercase + 1;    clist_dest = list + 2;
2963      code += 2;
2964    
2965  for (++c; c <= d; c++)    do {
2966    {       if (clist_dest >= list + 8)
2967    if (UCD_OTHERCASE(c) != next) break;         {
2968    next++;         /* Early return if there is not enough space. This should never
2969    }         happen, since all clists are shorter than 5 character now. */
2970           list[2] = code[0];
2971           list[3] = code[1];
2972           return code;
2973           }
2974         *clist_dest++ = *clist_src;
2975         }
2976      while(*clist_src++ != NOTACHAR);
2977    
2978  *odptr = next - 1;     /* End of othercase range */    /* All characters are stored. The terminating NOTACHAR
2979  *cptr = c;             /* Rest of input range */    is copied form the clist itself. */
2980  return 0;  
2981      list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
2982      return code;
2983    #endif
2984    
2985      case OP_NCLASS:
2986      case OP_CLASS:
2987    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2988      case OP_XCLASS:
2989      if (c == OP_XCLASS)
2990        end = code + GET(code, 0) - 1;
2991      else
2992    #endif
2993        end = code + 32 / sizeof(pcre_uchar);
2994    
2995      switch(*end)
2996        {
2997        case OP_CRSTAR:
2998        case OP_CRMINSTAR:
2999        case OP_CRQUERY:
3000        case OP_CRMINQUERY:
3001        case OP_CRPOSSTAR:
3002        case OP_CRPOSQUERY:
3003        list[1] = TRUE;
3004        end++;
3005        break;
3006    
3007        case OP_CRPLUS:
3008        case OP_CRMINPLUS:
3009        case OP_CRPOSPLUS:
3010        end++;
3011        break;
3012    
3013        case OP_CRRANGE:
3014        case OP_CRMINRANGE:
3015        case OP_CRPOSRANGE:
3016        list[1] = (GET2(end, 1) == 0);
3017        end += 1 + 2 * IMM2_SIZE;
3018        break;
3019        }
3020      list[2] = end - code;
3021      return end;
3022      }
3023    return NULL;    /* Opcode not accepted */
3024  }  }
3025    
3026    
3027    
3028  /*************************************************  /*************************************************
3029  *        Check a character and a property        *  *    Scan further character sets for match       *
3030  *************************************************/  *************************************************/
3031    
3032  /* 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
3033  is adjacent to a fixed character.  which case the base cannot be possessified.
3034    
3035  Arguments:  Arguments:
3036    c            the character    code        points to the byte code
3037    ptype        the property type    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3038    pdata        the data for the type    cd          static compile data
3039    negated      TRUE if it's a negated property (\P or \p{^)    base_list   the data list of the base opcode
3040    
3041  Returns:       TRUE if auto-possessifying is OK  Returns:      TRUE if the auto-possessification is possible
3042  */  */
3043    
3044  static BOOL  static BOOL
3045  check_char_prop(int c, int ptype, int pdata, BOOL negated)  compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3046      const pcre_uint32 *base_list, const pcre_uchar *base_end)
3047  {  {
3048  const ucd_record *prop = GET_UCD(c);  pcre_uchar c;
3049  switch(ptype)  pcre_uint32 list[8];
3050    const pcre_uint32 *chr_ptr;
3051    const pcre_uint32 *ochr_ptr;
3052    const pcre_uint32 *list_ptr;
3053    const pcre_uchar *next_code;
3054    const pcre_uint8 *class_bitset;
3055    const pcre_uint32 *set1, *set2, *set_end;
3056    pcre_uint32 chr;
3057    BOOL accepted, invert_bits;
3058    
3059    /* Note: the base_list[1] contains whether the current opcode has greedy
3060    (represented by a non-zero value) quantifier. This is a different from
3061    other character type lists, which stores here that the character iterator
3062    matches to an empty string (also represented by a non-zero value). */
3063    
3064    for(;;)
3065    {    {
3066    case PT_LAMP:    /* All operations move the code pointer forward.
3067    return (prop->chartype == ucp_Lu ||    Therefore infinite recursions are not possible. */
           prop->chartype == ucp_Ll ||  
           prop->chartype == ucp_Lt) == negated;  
3068    
3069    case PT_GC:    c = *code;
   return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;  
3070    
3071    case PT_PC:    /* Skip over callouts */
   return (pdata == prop->chartype) == negated;  
3072    
3073    case PT_SC:    if (c == OP_CALLOUT)
3074    return (pdata == prop->script) == negated;      {
3075        code += PRIV(OP_lengths)[c];
3076        continue;
3077        }
3078    
3079    /* These are specials */    if (c == OP_ALT)
3080        {
3081        do code += GET(code, 1); while (*code == OP_ALT);
3082        c = *code;
3083        }
3084    
3085    case PT_ALNUM:    switch(c)
3086    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||      {
3087            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;      case OP_END:
3088        case OP_KETRPOS:
3089        /* TRUE only in greedy case. The non-greedy case could be replaced by
3090        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3091        uses more memory, which we cannot get at this stage.) */
3092    
3093    case PT_SPACE:    /* Perl space */      return base_list[1] != 0;
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||  
           c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)  
           == negated;  
3094    
3095    case PT_PXSPACE:  /* POSIX space */      case OP_KET:
3096    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||      /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3097            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||      it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3098            c == CHAR_FF || c == CHAR_CR)      cannot be converted to a possessive form. */
3099            == negated;  
3100        if (base_list[1] == 0) return FALSE;
3101    
3102        switch(*(code - GET(code, 1)))
3103          {
3104          case OP_ASSERT:
3105          case OP_ASSERT_NOT:
3106          case OP_ASSERTBACK:
3107          case OP_ASSERTBACK_NOT:
3108          case OP_ONCE:
3109          case OP_ONCE_NC:
3110          /* Atomic sub-patterns and assertions can always auto-possessify their
3111          last iterator. */
3112          return TRUE;
3113          }
3114    
3115    case PT_WORD:      code += PRIV(OP_lengths)[c];
3116    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||      continue;
           PRIV(ucp_gentype)[prop->chartype] == ucp_N ||  
           c == CHAR_UNDERSCORE) == negated;  
   }  
 return FALSE;  
 }  
 #endif  /* SUPPORT_UCP */  
3117    
3118        case OP_ONCE:
3119        case OP_ONCE_NC:
3120        case OP_BRA:
3121        case OP_CBRA:
3122        next_code = code + GET(code, 1);
3123        code += PRIV(OP_lengths)[c];
3124    
3125        while (*next_code == OP_ALT)
3126          {
3127          if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3128          code = next_code + 1 + LINK_SIZE;
3129          next_code += GET(next_code, 1);
3130          }
3131        continue;
3132    
3133  /*************************************************      case OP_BRAZERO:
3134  *     Check if auto-possessifying is possible    *      case OP_BRAMINZERO:
 *************************************************/  
3135    
3136  /* This function is called for unlimited repeats of certain items, to see      next_code = code + 1;
3137  whether the next thing could possibly match the repeated item. If not, it makes      if (*next_code != OP_BRA && *next_code != OP_CBRA
3138  sense to automatically possessify the repeated item.          && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3139    
3140        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3141    
3142        /* The bracket content will be checked by the
3143        OP_BRA/OP_CBRA case above. */
3144        next_code += 1 + LINK_SIZE;
3145        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3146          return FALSE;
3147    
3148  Arguments:      code += PRIV(OP_lengths)[c];
3149    previous      pointer to the repeated opcode      continue;
3150    utf           TRUE in UTF-8 / UTF-16 mode      }
   ptr           next character in pattern  
   options       options bits  
   cd            contains pointers to tables etc.  
3151    
3152  Returns:        TRUE if possessifying is wanted    /* Check for a supported opcode, and load its properties. */
 */  
3153    
3154  static BOOL    code = get_chr_property_list(code, utf, cd->fcc, list);
3155  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_int32 c, next;  
 int op_code = *previous++;  
3156    
3157  /* Skip whitespace and comments in extended mode */    /* If either opcode is a small character list, set pointers for comparing
3158      characters from that list with another list, or with a property. */
3159    
3160  if ((options & PCRE_EXTENDED) != 0)    if (base_list[0] == OP_CHAR)
   {  
   for (;;)  
3161      {      {
3162      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      chr_ptr = base_list + 2;
3163      if (*ptr == CHAR_NUMBER_SIGN)      list_ptr = list;
3164        {      }
3165        ptr++;    else if (list[0] == OP_CHAR)
3166        while (*ptr != 0)      {
3167          {      chr_ptr = list + 2;
3168          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }      list_ptr = base_list;
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
       }  
     else break;  
3169      }      }
   }  
3170    
3171  /* 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. */  
3172    
3173  if (*ptr == CHAR_BACKSLASH)    else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3174    {  #ifdef COMPILE_PCRE8
3175    int temperrorcode = 0;        /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3176    next = check_escape(&ptr, &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)  
   {  
 #ifdef SUPPORT_UTF  
   if (utf) { GETCHARINC(next, ptr); } else  
3177  #endif  #endif
3178    next = *ptr++;        )
3179    }      {
3180  else return FALSE;  #ifdef COMPILE_PCRE8
3181        if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3182    #else
3183        if (base_list[0] == OP_CLASS)
3184    #endif
3185          {
3186          set1 = (pcre_uint32 *)(base_end - base_list[2]);
3187          list_ptr = list;
3188          }
3189        else
3190          {
3191          set1 = (pcre_uint32 *)(code - list[2]);
3192          list_ptr = base_list;
3193          }
3194    
3195  /* Skip whitespace and comments in extended mode */      invert_bits = FALSE;
3196        switch(list_ptr[0])
3197          {
3198          case OP_CLASS:
3199          case OP_NCLASS:
3200          set2 = (pcre_uint32 *)
3201            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3202          break;
3203    
3204  if ((options & PCRE_EXTENDED) != 0)        /* OP_XCLASS cannot be supported here, because its bitset
3205    {        is not necessarily complete. E.g: [a-\0x{200}] is stored
3206    for (;;)        as a character range, and the appropriate bits are not set. */
3207      {  
3208      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;        case OP_NOT_DIGIT:
3209      if (*ptr == CHAR_NUMBER_SIGN)          invert_bits = TRUE;
3210            /* Fall through */
3211          case OP_DIGIT:
3212            set2 = (pcre_uint32 *)(cd->cbits + cbit_digit);
3213            break;
3214    
3215          case OP_NOT_WHITESPACE:
3216            invert_bits = TRUE;
3217            /* Fall through */
3218          case OP_WHITESPACE:
3219            set2 = (pcre_uint32 *)(cd->cbits + cbit_space);
3220            break;
3221    
3222          case OP_NOT_WORDCHAR:
3223            invert_bits = TRUE;
3224            /* Fall through */
3225          case OP_WORDCHAR:
3226            set2 = (pcre_uint32 *)(cd->cbits + cbit_word);
3227            break;
3228    
3229          default:
3230          return FALSE;
3231          }
3232    
3233        /* Compare 4 bytes to improve speed. */
3234        set_end = set1 + (32 / 4);
3235        if (invert_bits)
3236        {        {
3237        ptr++;        do
       while (*ptr != 0)  
3238          {          {
3239          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3240          ptr++;          }
3241  #ifdef SUPPORT_UTF        while (set1 < set_end);
3242          if (utf) FORWARDCHAR(ptr);        }
3243  #endif      else
3244          {
3245          do
3246            {
3247            if ((*set1++ & *set2++) != 0) return FALSE;
3248          }          }
3249          while (set1 < set_end);
3250        }        }
3251      else break;  
3252        if (list[1] == 0) return TRUE;
3253        /* Might be an empty repeat. */
3254        continue;
3255      }      }
   }  
3256    
3257  /* If the next thing is itself optional, we have to give up. */    /* Some property combinations also acceptable. Unicode property opcodes are
3258      processed specially; the rest can be handled with a lookup table. */
3259    
3260  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||    else
3261    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)      {
3262      return FALSE;      pcre_uint32 leftop, rightop;
3263    
3264  /* Now compare the next item with the previous opcode. First, handle cases when      leftop = base_list[0];
3265  the next item is a character. */      rightop = list[0];
3266    
3267  if (next >= 0) switch(op_code)  #ifdef SUPPORT_UCP
3268    {      accepted = FALSE; /* Always set in non-unicode case. */
3269    case OP_CHAR:      if (leftop == OP_PROP || leftop == OP_NOTPROP)
3270  #ifdef SUPPORT_UTF        {
3271    GETCHARTEST(c, previous);        if (rightop == OP_EOD)
3272  #else          accepted = TRUE;
3273    c = *previous;        else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3274  #endif          {
3275    return c != next;          int n;
3276            const pcre_uint8 *p;
3277            BOOL same = leftop == rightop;
3278            BOOL lisprop = leftop == OP_PROP;
3279            BOOL risprop = rightop == OP_PROP;
3280            BOOL bothprop = lisprop && risprop;
3281    
3282            /* There's a table that specifies how each combination is to be
3283            processed:
3284              0   Always return FALSE (never auto-possessify)
3285              1   Character groups are distinct (possessify if both are OP_PROP)
3286              2   Check character categories in the same group (general or particular)
3287              3   Return TRUE if the two opcodes are not the same
3288              ... see comments below
3289            */
3290    
3291            n = propposstab[base_list[2]][list[2]];
3292            switch(n)
3293              {
3294              case 0: break;
3295              case 1: accepted = bothprop; break;
3296              case 2: accepted = (base_list[3] == list[3]) != same; break;
3297              case 3: accepted = !same; break;
3298    
3299    /* For CHARI (caseless character) we must check the other case. If we have            case 4:  /* Left general category, right particular category */
3300    Unicode property support, we can use it to test the other case of            accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3301    high-valued characters. */            break;
3302    
3303    case OP_CHARI:            case 5:  /* Right general category, left particular category */
3304  #ifdef SUPPORT_UTF            accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3305    GETCHARTEST(c, previous);            break;
3306  #else  
3307    c = *previous;            /* This code is logically tricky. Think hard before fiddling with it.
3308  #endif            The posspropstab table has four entries per row. Each row relates to
3309    if (c == next) return FALSE;            one of PCRE's special properties such as ALNUM or SPACE or WORD.
3310  #ifdef SUPPORT_UTF            Only WORD actually needs all four entries, but using repeats for the
3311    if (utf)            others means they can all use the same code below.
3312    
3313              The first two entries in each row are Unicode general categories, and
3314              apply always, because all the characters they include are part of the
3315              PCRE character set. The third and fourth entries are a general and a
3316              particular category, respectively, that include one or more relevant
3317              characters. One or the other is used, depending on whether the check
3318              is for a general or a particular category. However, in both cases the
3319              category contains more characters than the specials that are defined
3320              for the property being tested against. Therefore, it cannot be used
3321              in a NOTPROP case.
3322    
3323              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3324              Underscore is covered by ucp_P or ucp_Po. */
3325    
3326              case 6:  /* Left alphanum vs right general category */
3327              case 7:  /* Left space vs right general category */
3328              case 8:  /* Left word vs right general category */
3329              p = posspropstab[n-6];
3330              accepted = risprop && lisprop ==
3331                (list[3] != p[0] &&
3332                 list[3] != p[1] &&
3333                (list[3] != p[2] || !lisprop));
3334              break;
3335    
3336              case 9:   /* Right alphanum vs left general category */
3337              case 10:  /* Right space vs left general category */
3338              case 11:  /* Right word vs left general category */
3339              p = posspropstab[n-9];
3340              accepted = lisprop && risprop ==
3341                (base_list[3] != p[0] &&
3342                 base_list[3] != p[1] &&
3343                (base_list[3] != p[2] || !risprop));
3344              break;
3345    
3346              case 12:  /* Left alphanum vs right particular category */
3347              case 13:  /* Left space vs right particular category */
3348              case 14:  /* Left word vs right particular category */
3349              p = posspropstab[n-12];
3350              accepted = risprop && lisprop ==
3351                (catposstab[p[0]][list[3]] &&
3352                 catposstab[p[1]][list[3]] &&
3353                (list[3] != p[3] || !lisprop));
3354              break;
3355    
3356              case 15:  /* Right alphanum vs left particular category */
3357              case 16:  /* Right space vs left particular category */
3358              case 17:  /* Right word vs left particular category */
3359              p = posspropstab[n-15];
3360              accepted = lisprop && risprop ==
3361                (catposstab[p[0]][base_list[3]] &&
3362                 catposstab[p[1]][base_list[3]] &&
3363                (base_list[3] != p[3] || !risprop));
3364              break;
3365              }
3366            }
3367          }
3368    
3369        else
3370    #endif  /* SUPPORT_UCP */
3371    
3372        accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3373               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3374               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3375    
3376        if (!accepted)
3377          return FALSE;
3378    
3379        if (list[1] == 0) return TRUE;
3380        /* Might be an empty repeat. */
3381        continue;
3382        }
3383    
3384      /* Control reaches here only if one of the items is a small character list.
3385      All characters are checked against the other side. */
3386    
3387      do
3388      {      {
3389      unsigned int othercase;      chr = *chr_ptr;
3390      if (next < 128) othercase = cd->fcc[next]; else  
3391        switch(list_ptr[0])
3392          {
3393          case OP_CHAR:
3394          ochr_ptr = list_ptr + 2;
3395          do
3396            {
3397            if (chr == *ochr_ptr) return FALSE;
3398            ochr_ptr++;
3399            }
3400          while(*ochr_ptr != NOTACHAR);
3401          break;
3402    
3403          case OP_NOT:
3404          ochr_ptr = list_ptr + 2;
3405          do
3406            {
3407            if (chr == *ochr_ptr)
3408              break;
3409            ochr_ptr++;
3410            }
3411          while(*ochr_ptr != NOTACHAR);
3412          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3413          break;
3414    
3415          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3416          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3417    
3418          case OP_DIGIT:
3419          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3420          break;
3421    
3422          case OP_NOT_DIGIT:
3423          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3424          break;
3425    
3426          case OP_WHITESPACE:
3427          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3428          break;
3429    
3430          case OP_NOT_WHITESPACE:
3431          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3432          break;
3433    
3434          case OP_WORDCHAR:
3435          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3436          break;
3437    
3438          case OP_NOT_WORDCHAR:
3439          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3440          break;
3441    
3442          case OP_HSPACE:
3443          switch(chr)
3444            {
3445            HSPACE_CASES: return FALSE;
3446            default: break;
3447            }
3448          break;
3449    
3450          case OP_NOT_HSPACE:
3451          switch(chr)
3452            {
3453            HSPACE_CASES: break;
3454            default: return FALSE;
3455            }
3456          break;
3457    
3458          case OP_ANYNL:
3459          case OP_VSPACE:
3460          switch(chr)
3461            {
3462            VSPACE_CASES: return FALSE;
3463            default: break;
3464            }
3465          break;
3466    
3467          case OP_NOT_VSPACE:
3468          switch(chr)
3469            {
3470            VSPACE_CASES: break;
3471            default: return FALSE;
3472            }
3473          break;
3474    
3475          case OP_DOLL:
3476          case OP_EODN:
3477          switch (chr)
3478            {
3479            case CHAR_CR:
3480            case CHAR_LF:
3481            case CHAR_VT:
3482            case CHAR_FF:
3483            case CHAR_NEL:
3484    #ifndef EBCDIC
3485            case 0x2028:
3486            case 0x2029:
3487    #endif  /* Not EBCDIC */
3488            return FALSE;
3489            }
3490          break;
3491    
3492          case OP_EOD:    /* Can always possessify before \z */
3493          break;
3494    
3495  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3496      othercase = UCD_OTHERCASE((unsigned int)next);        case OP_PROP:
3497  #else        case OP_NOTPROP:
3498      othercase = NOTACHAR;        if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3499                list_ptr[0] == OP_NOTPROP))
3500            return FALSE;
3501          break;
3502  #endif  #endif
     return (unsigned int)c != othercase;  
     }  
   else  
 #endif  /* SUPPORT_UTF */  
   return (c != TABLE_GET((unsigned int)next, cd->fcc, next));  /* Non-UTF-8 mode */  
3503    
3504    case OP_NOT:        case OP_NCLASS:
3505  #ifdef SUPPORT_UTF        if (chr > 255) return FALSE;
3506    GETCHARTEST(c, previous);        /* Fall through */
3507  #else  
3508    c = *previous;        case OP_CLASS:
3509  #endif        if (chr > 255) break;
3510    return c == next;        class_bitset = (pcre_uint8 *)
3511            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3512          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3513          break;
3514    
3515    case OP_NOTI:  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3516  #ifdef SUPPORT_UTF        case OP_XCLASS:
3517    GETCHARTEST(c, previous);        if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3518  #else            list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3519    c = *previous;        break;
3520  #endif  #endif
3521    if (c == next) return TRUE;  
3522  #ifdef SUPPORT_UTF        default:
3523    if (utf)        return FALSE;
3524          }
3525    
3526        chr_ptr++;
3527        }
3528      while(*chr_ptr != NOTACHAR);
3529    
3530      /* At least one character must be matched from this opcode. */
3531    
3532      if (list[1] == 0) return TRUE;
3533      }
3534    
3535    return FALSE;
3536    }
3537    
3538    
3539    
3540    /*************************************************
3541    *    Scan compiled regex for auto-possession     *
3542    *************************************************/
3543    
3544    /* Replaces single character iterations with their possessive alternatives
3545    if appropriate. This function modifies the compiled opcode!
3546    
3547    Arguments:
3548      code        points to start of the byte code
3549      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3550      cd          static compile data
3551    
3552    Returns:      nothing
3553    */
3554    
3555    static void
3556    auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3557    {
3558    register pcre_uchar c;
3559    const pcre_uchar *end;
3560    pcre_uchar *repeat_opcode;
3561    pcre_uint32 list[8];
3562    
3563    for (;;)
3564      {
3565      c = *code;
3566    
3567      if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3568      {      {
3569      unsigned int othercase;      c -= get_repeat_base(c) - OP_STAR;
3570      if (next < 128) othercase = cd->fcc[next]; else      end = (c <= OP_MINUPTO) ?
3571  #ifdef SUPPORT_UCP        get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3572      othercase = UCD_OTHERCASE((unsigned int)next);      list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3573  #else  
3574      othercase = NOTACHAR;      if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3575          {
3576          switch(c)
3577            {
3578            case OP_STAR:
3579            *code += OP_POSSTAR - OP_STAR;
3580            break;
3581    
3582            case OP_MINSTAR:
3583            *code += OP_POSSTAR - OP_MINSTAR;
3584            break;
3585    
3586            case OP_PLUS:
3587            *code += OP_POSPLUS - OP_PLUS;
3588            break;
3589    
3590            case OP_MINPLUS:
3591            *code += OP_POSPLUS - OP_MINPLUS;
3592            break;
3593    
3594            case OP_QUERY:
3595            *code += OP_POSQUERY - OP_QUERY;
3596            break;
3597    
3598            case OP_MINQUERY:
3599            *code += OP_POSQUERY - OP_MINQUERY;
3600            break;
3601    
3602            case OP_UPTO:
3603            *code += OP_POSUPTO - OP_UPTO;
3604            break;
3605    
3606            case OP_MINUPTO:
3607            *code += OP_MINUPTO - OP_UPTO;
3608            break;
3609            }
3610          }
3611        c = *code;
3612        }
3613      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3614        {
3615    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3616        if (c == OP_XCLASS)
3617          repeat_opcode = code + GET(code, 1);
3618        else
3619  #endif  #endif
3620      return (unsigned int)c == othercase;        repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3621    
3622        c = *repeat_opcode;
3623        if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3624          {
3625          /* end must not be NULL. */
3626          end = get_chr_property_list(code, utf, cd->fcc, list);
3627    
3628          list[1] = (c & 1) == 0;
3629    
3630          if (compare_opcodes(end, utf, cd, list, end))
3631            {
3632            switch (c)
3633              {
3634              case OP_CRSTAR:
3635              case OP_CRMINSTAR:
3636              *repeat_opcode = OP_CRPOSSTAR;
3637              break;
3638    
3639              case OP_CRPLUS:
3640              case OP_CRMINPLUS:
3641              *repeat_opcode = OP_CRPOSPLUS;
3642              break;
3643    
3644              case OP_CRQUERY:
3645              case OP_CRMINQUERY:
3646              *repeat_opcode = OP_CRPOSQUERY;
3647              break;
3648    
3649              case OP_CRRANGE:
3650              case OP_CRMINRANGE:
3651              *repeat_opcode = OP_CRPOSRANGE;
3652              break;
3653              }
3654            }
3655          }
3656        c = *code;
3657      }      }
   else  
 #endif  /* SUPPORT_UTF */  
   return (c == TABLE_GET((unsigned int)next, cd->fcc, next));  /* Non-UTF-8 mode */  
3658    
3659    /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.    switch(c)
3660    When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */      {
3661        case OP_END:
3662        return;
3663    
3664    case OP_DIGIT:      case OP_TYPESTAR:
3665    return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;      case OP_TYPEMINSTAR:
3666        case OP_TYPEPLUS:
3667        case OP_TYPEMINPLUS:
3668        case OP_TYPEQUERY:
3669        case OP_TYPEMINQUERY:
3670        case OP_TYPEPOSSTAR:
3671        case OP_TYPEPOSPLUS:
3672        case OP_TYPEPOSQUERY:
3673        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3674        break;
3675    
3676    case OP_NOT_DIGIT:      case OP_TYPEUPTO:
3677    return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;      case OP_TYPEMINUPTO:
3678        case OP_TYPEEXACT:
3679        case OP_TYPEPOSUPTO:
3680        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3681          code += 2;
3682        break;
3683    
3684    case OP_WHITESPACE:  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3685    return next > 255 || (cd->ctypes[next] & ctype_space) == 0;      case OP_XCLASS:
3686        code += GET(code, 1);
3687        break;
3688    #endif
3689    
3690    case OP_NOT_WHITESPACE:      case OP_MARK:
3691    return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;      case OP_PRUNE_ARG:
3692        case OP_SKIP_ARG:
3693        case OP_THEN_ARG:
3694        code += code[1];
3695        break;
3696        }
3697    
3698    case OP_WORDCHAR:    /* Add in the fixed length from the table */
   return next > 255 || (cd->ctypes[next] & ctype_word) == 0;  
3699    
3700    case OP_NOT_WORDCHAR:    code += PRIV(OP_lengths)[c];
   return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;  
3701    
3702    case OP_HSPACE:    /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3703    case OP_NOT_HSPACE:    a multi-byte character. The length in the table is a minimum, so we have to
3704    switch(next)    arrange to skip the extra bytes. */
3705    
3706    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3707      if (utf) switch(c)
3708      {      {
3709      HSPACE_CASES:      case OP_CHAR:
3710      return op_code == OP_NOT_HSPACE;      case OP_CHARI:
3711        case OP_NOT:
3712      default:      case OP_NOTI:
3713      return op_code != OP_NOT_HSPACE;      case OP_STAR:
3714        case OP_MINSTAR:
3715        case OP_PLUS:
3716        case OP_MINPLUS:
3717        case OP_QUERY:
3718        case OP_MINQUERY:
3719        case OP_UPTO:
3720        case OP_MINUPTO:
3721        case OP_EXACT:
3722        case OP_POSSTAR:
3723        case OP_POSPLUS:
3724        case OP_POSQUERY:
3725        case OP_POSUPTO:
3726        case OP_STARI:
3727        case OP_MINSTARI:
3728        case OP_PLUSI:
3729        case OP_MINPLUSI:
3730        case OP_QUERYI:
3731        case OP_MINQUERYI:
3732        case OP_UPTOI:
3733        case OP_MINUPTOI:
3734        case OP_EXACTI:
3735        case OP_POSSTARI:
3736        case OP_POSPLUSI:
3737        case OP_POSQUERYI:
3738        case OP_POSUPTOI:
3739        case OP_NOTSTAR:
3740        case OP_NOTMINSTAR:
3741        case OP_NOTPLUS:
3742        case OP_NOTMINPLUS:
3743        case OP_NOTQUERY:
3744        case OP_NOTMINQUERY:
3745        case OP_NOTUPTO:
3746        case OP_NOTMINUPTO:
3747        case OP_NOTEXACT:
3748        case OP_NOTPOSSTAR:
3749        case OP_NOTPOSPLUS:
3750        case OP_NOTPOSQUERY:
3751        case OP_NOTPOSUPTO:
3752        case OP_NOTSTARI:
3753        case OP_NOTMINSTARI:
3754        case OP_NOTPLUSI:
3755        case OP_NOTMINPLUSI:
3756        case OP_NOTQUERYI:
3757        case OP_NOTMINQUERYI:
3758        case OP_NOTUPTOI:
3759        case OP_NOTMINUPTOI:
3760        case OP_NOTEXACTI:
3761        case OP_NOTPOSSTARI:
3762        case OP_NOTPOSPLUSI:
3763        case OP_NOTPOSQUERYI:
3764        case OP_NOTPOSUPTOI:
3765        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3766        break;
3767      }      }
3768    #else
3769      (void)(utf);  /* Keep compiler happy by referencing function argument */
3770    #endif
3771      }
3772    }
3773    
3774    case OP_ANYNL:  
3775    case OP_VSPACE:  
3776    case OP_NOT_VSPACE:  /*************************************************
3777    switch(next)  *           Check for POSIX class syntax         *
3778    *************************************************/
3779    
3780    /* This function is called when the sequence "[:" or "[." or "[=" is
3781    encountered in a character class. It checks whether this is followed by a
3782    sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3783    reach an unescaped ']' without the special preceding character, return FALSE.
3784    
3785    Originally, this function only recognized a sequence of letters between the
3786    terminators, but it seems that Perl recognizes any sequence of characters,
3787    though of course unknown POSIX names are subsequently rejected. Perl gives an
3788    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3789    didn't consider this to be a POSIX class. Likewise for [:1234:].
3790    
3791    The problem in trying to be exactly like Perl is in the handling of escapes. We
3792    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3793    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3794    below handles the special case of \], but does not try to do any other escape
3795    processing. This makes it different from Perl for cases such as [:l\ower:]
3796    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3797    "l\ower". This is a lesser evil than not diagnosing bad classes when Perl does,
3798    I think.
3799    
3800    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3801    It seems that the appearance of a nested POSIX class supersedes an apparent
3802    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3803    a digit.
3804    
3805    In Perl, unescaped square brackets may also appear as part of class names. For
3806    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3807    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3808    seem right at all. PCRE does not allow closing square brackets in POSIX class
3809    names.
3810    
3811    Arguments:
3812      ptr      pointer to the initial [
3813      endptr   where to return the end pointer
3814    
3815    Returns:   TRUE or FALSE
3816    */
3817    
3818    static BOOL
3819    check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3820    {
3821    pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
3822    terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
3823    for (++ptr; *ptr != CHAR_NULL; ptr++)
3824      {
3825      if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3826        ptr++;
3827      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3828      else
3829      {      {
3830      VSPACE_CASES:      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3831      return op_code == OP_NOT_VSPACE;        {
3832          *endptr = ptr;
3833      default:        return TRUE;
3834      return op_code != OP_NOT_VSPACE;        }
3835        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3836             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3837              ptr[1] == CHAR_EQUALS_SIGN) &&
3838            check_posix_syntax(ptr, endptr))
3839          return FALSE;
3840      }      }
3841      }
3842    return FALSE;
3843    }
3844    
 #ifdef SUPPORT_UCP  
   case OP_PROP:  
   return check_char_prop(next, previous[0], previous[1], FALSE);  
3845    
   case OP_NOTPROP:  
   return check_char_prop(next, previous[0], previous[1], TRUE);  
 #endif  
3846    
3847    default:  
3848    return FALSE;  /*************************************************
3849    *          Check POSIX class name                *
3850    *************************************************/
3851    
3852    /* This function is called to check the name given in a POSIX-style class entry
3853    such as [:alnum:].
3854    
3855    Arguments:
3856      ptr        points to the first letter
3857      len        the length of the name
3858    
3859    Returns:     a value representing the name, or -1 if unknown
3860    */
3861    
3862    static int
3863    check_posix_name(const pcre_uchar *ptr, int len)
3864    {
3865    const char *pn = posix_names;
3866    register int yield = 0;
3867    while (posix_name_lengths[yield] != 0)
3868      {
3869      if (len == posix_name_lengths[yield] &&
3870        STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3871      pn += posix_name_lengths[yield] + 1;
3872      yield++;
3873    }    }
3874    return -1;
3875    }
3876    
3877    
3878  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP  /*************************************************
3879  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are  *    Adjust OP_RECURSE items in repeated group   *
3880  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. */  
3881    
3882  switch(op_code)  /* OP_RECURSE items contain an offset from the start of the regex to the group
3883    that is referenced. This means that groups can be replicated for fixed
3884    repetition simply by copying (because the recursion is allowed to refer to
3885    earlier groups that are outside the current group). However, when a group is
3886    optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3887    inserted before it, after it has been compiled. This means that any OP_RECURSE
3888    items within it that refer to the group itself or any contained groups have to
3889    have their offsets adjusted. That one of the jobs of this function. Before it
3890    is called, the partially compiled regex must be temporarily terminated with
3891    OP_END.
3892    
3893    This function has been extended with the possibility of forward references for
3894    recursions and subroutine calls. It must also check the list of such references
3895    for the group we are dealing with. If it finds that one of the recursions in
3896    the current group is on this list, it adjusts the offset in the list, not the
3897    value in the reference (which is a group number).
3898    
3899    Arguments:
3900      group      points to the start of the group
3901      adjust     the amount by which the group is to be moved
3902      utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
3903      cd         contains pointers to tables etc.
3904      save_hwm   the hwm forward reference pointer at the start of the group
3905    
3906    Returns:     nothing
3907    */
3908    
3909    static void
3910    adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3911      pcre_uchar *save_hwm)
3912    {
3913    pcre_uchar *ptr = group;
3914    
3915    while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3916    {    {
3917    case OP_CHAR:    int offset;
3918    case OP_CHARI:    pcre_uchar *hc;
3919  #ifdef SUPPORT_UTF  
3920    GETCHARTEST(c, previous);    /* See if this recursion is on the forward reference list. If so, adjust the
3921  #else    reference. */
3922    c = *previous;  
3923  #endif    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
   switch(-next)  
3924      {      {
3925      case ESC_d:      offset = (int)GET(hc, 0);
3926      return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;      if (cd->start_code + offset == ptr + 1)
3927          {
3928          PUT(hc, 0, offset + adjust);
3929          break;
3930          }
3931        }
3932    
3933      case ESC_D:    /* Otherwise, adjust the recursion offset if it's after the start of this
3934      return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;    group. */
3935    
3936      case ESC_s:    if (hc >= cd->hwm)
3937      return c > 255 || (cd->ctypes[c] & ctype_space) == 0;      {
3938        offset = (int)GET(ptr, 1);
3939        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3940        }
3941    
3942      ptr += 1 + LINK_SIZE;
3943      }
3944    }
3945    
     case ESC_S:  
     return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;  
3946    
     case ESC_w:  
     return c > 255 || (cd->ctypes[c] & ctype_word) == 0;  
3947    
3948      case ESC_W:  /*************************************************
3949      return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;  *        Insert an automatic callout point       *
3950    *************************************************/
3951    
3952      case ESC_h:  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
3953      case ESC_H:  callout points before each pattern item.
     switch(c)  
       {  
       HSPACE_CASES:  
       return -next != ESC_h;  
   
       default:  
       return -next == ESC_h;  
       }  
3954    
3955      case ESC_v:  Arguments:
3956      case ESC_V:    code           current code pointer
3957      switch(c)    ptr            current pattern pointer
3958        {    cd             pointers to tables etc
       VSPACE_CASES:  
       return -next != ESC_v;  
   
       default:  
       return -next == ESC_v;  
       }  
3959    
3960      /* When PCRE_UCP is set, these values get generated for \d etc. Find  Returns:         new code pointer
3961      their substitutions and process them. The result will always be either  */
     -ESC_p or -ESC_P. Then fall through to process those values. */  
3962    
3963  #ifdef SUPPORT_UCP  static pcre_uchar *
3964      case ESC_du:  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
3965      case ESC_DU:  {
3966      case ESC_wu:  *code++ = OP_CALLOUT;
3967      case ESC_WU:  *code++ = 255;
3968      case ESC_su:  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
3969      case ESC_SU:  PUT(code, LINK_SIZE, 0);                       /* Default length */
3970        {  return code + 2 * LINK_SIZE;
3971        int temperrorcode = 0;  }
       ptr = substitutes[-next - ESC_DU];  
       next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);  
       if (temperrorcode != 0) return FALSE;  
       ptr++;    /* For compatibility */  
       }  
     /* Fall through */  
3972    
     case ESC_p:  
     case ESC_P:  
       {  
       int ptype, pdata, errorcodeptr;  
       BOOL negated;  
3973    
       ptr--;      /* Make ptr point at the p or P */  
       ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);  
       if (ptype < 0) return FALSE;  
       ptr++;      /* Point past the final curly ket */  
3974    
3975        /* If the property item is optional, we have to give up. (When generated  /*************************************************
3976        from \d etc by PCRE_UCP, this test will have been applied much earlier,  *         Complete a callout item                *
3977        to the original \d etc. At this point, ptr will point to a zero byte. */  *************************************************/
3978    
3979        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  /* A callout item contains the length of the next item in the pattern, which
3980          STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)  we can't fill in till after we have reached the relevant point. This is used
3981            return FALSE;  for both automatic and manual callouts.
3982    
3983        /* Do the property check. */  Arguments:
3984      previous_callout   points to previous callout item
3985      ptr                current pattern pointer
3986      cd                 pointers to tables etc
3987    
3988        return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);  Returns:             nothing
3989        }  */
 #endif  
3990    
3991      default:  static void
3992      return FALSE;  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
3993      }  {
3994    int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
3995    PUT(previous_callout, 2 + LINK_SIZE, length);
3996    }
3997    
   /* In principle, support for Unicode properties should be integrated here as  
   well. It means re-organizing the above code so as to get hold of the property  
   values before switching on the op-code. However, I wonder how many patterns  
   combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,  
   these op-codes are never generated.) */  
3998    
   case OP_DIGIT:  
   return next == -ESC_D || next == -ESC_s || next == -ESC_W ||  
          next == -ESC_h || next == -ESC_v || next == -ESC_R;  
3999    
4000    case OP_NOT_DIGIT:  #ifdef SUPPORT_UCP
4001    return next == -ESC_d;  /*************************************************
4002    *           Get othercase range                  *
4003    *************************************************/
4004    
4005    case OP_WHITESPACE:  /* This function is passed the start and end of a class range, in UTF-8 mode
4006    return next == -ESC_S || next == -ESC_d || next == -ESC_w;  with UCP support. It searches up the characters, looking for ranges of
4007    characters in the "other" case. Each call returns the next one, updating the
4008    start address. A character with multiple other cases is returned on its own
4009    with a special return value.
4010    
4011    case OP_NOT_WHITESPACE:  Arguments:
4012    return next == -ESC_s || next == -ESC_h || next == -ESC_v || next == -ESC_R;    cptr        points to starting character value; updated
4013      d           end value
4014      ocptr       where to put start of othercase range
4015      odptr       where to put end of othercase range
4016    
4017    case OP_HSPACE:  Yield:        -1 when no more
4018    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||                 0 when a range is returned
4019           next == -ESC_w || next == -ESC_v || next == -ESC_R;                >0 the CASESET offset for char with multiple other cases
4020                    in this case, ocptr contains the original
4021    */
4022    
4023    case OP_NOT_HSPACE:  static int
4024    return next == -ESC_h;  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4025      pcre_uint32 *odptr)
4026    {
4027    pcre_uint32 c, othercase, next;
4028    unsigned int co;
4029    
4030    /* 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
4031    case OP_ANYNL:  cases, return its case offset value. */
   case OP_VSPACE:  
   return next == -ESC_V || next == -ESC_d || next == -ESC_w;  
4032    
4033    case OP_NOT_VSPACE:  for (c = *cptr; c <= d; c++)
4034    return next == -ESC_v || next == -ESC_R;    {
4035      if ((co = UCD_CASESET(c)) != 0)
4036        {
4037        *ocptr = c++;   /* Character that has the set */
4038        *cptr = c;      /* Rest of input range */
4039        return (int)co;
4040        }
4041      if ((othercase = UCD_OTHERCASE(c)) != c) break;
4042      }
4043    
4044    case OP_WORDCHAR:  if (c > d) return -1;  /* Reached end of range */
   return next == -ESC_W || next == -ESC_s || next == -ESC_h ||  
          next == -ESC_v || next == -ESC_R;  
4045    
4046    case OP_NOT_WORDCHAR:  *ocptr = othercase;
4047    return next == -ESC_w || next == -ESC_d;  next = othercase + 1;
4048    
4049    default:  for (++c; c <= d; c++)
4050    return FALSE;    {
4051      if (UCD_OTHERCASE(c) != next) break;
4052      next++;
4053    }    }
4054    
4055  /* Control does not reach here */  *odptr = next - 1;     /* End of othercase range */
4056    *cptr = c;             /* Rest of input range */
4057    return 0;
4058  }  }
4059    #endif  /* SUPPORT_UCP */
4060    
4061    
4062    
# Line 3381  switch(op_code) Line 4065  switch(op_code)
4065  *************************************************/  *************************************************/
4066    
4067  /* 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
4068  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
4069  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
4070  mutually recursive with the function immediately below.  mutually recursive with the function immediately below.
4071    
4072  Arguments:  Arguments:
4073    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4074    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4075    options       the options word    options       the options word
4076    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4077    start         start of range character    start         start of range character
4078    end           end of range character    end           end of range character
4079    
4080  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4081                  the pointer to extra data is updated                  the pointer to extra data is updated
4082  */  */
4083    
4084  static int  static int
4085  add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,  add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4086    compile_data *cd, unsigned int start, unsigned int end)    compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4087  {  {
4088  unsigned int c;  pcre_uint32 c;
4089  int n8 = 0;  int n8 = 0;
4090    
4091  /* If caseless matching is required, scan the range and process alternate  /* If caseless matching is required, scan the range and process alternate
4092  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
4093  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
4094  range. */  range. */
4095    
4096  if ((options & PCRE_CASELESS) != 0)  if ((options & PCRE_CASELESS) != 0)
4097    {    {
4098  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4099    if ((options & PCRE_UTF8) != 0)    if ((options & PCRE_UTF8) != 0)
4100      {      {
4101      int rc;      int rc;
4102      unsigned int oc, od;      pcre_uint32 oc, od;
4103    
4104      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
4105      c = start;      c = start;
4106    
4107      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
4108        {        {
4109        /* Handle a single character that has more than one other case. */        /* Handle a single character that has more than one other case. */
4110    
4111        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4112          PRIV(ucd_caseless_sets) + rc, oc);          PRIV(ucd_caseless_sets) + rc, oc);
4113    
4114        /* Do nothing if the other case range is within the original range. */        /* Do nothing if the other case range is within the original range. */
4115    
4116        else if (oc >= start && od <= end) continue;        else if (oc >= start && od <= end) continue;
4117    
4118        /* 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
4119        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
4120        range. Otherwise, use a recursive call to add the additional range. */        range. Otherwise, use a recursive call to add the additional range. */
4121    
4122        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4123        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
4124        else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);        else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
# Line 3444  if ((options & PCRE_CASELESS) != 0) Line 4128  if ((options & PCRE_CASELESS) != 0)
4128  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
4129    
4130    /* Not UTF-mode, or no UCP */    /* Not UTF-mode, or no UCP */
4131    
4132    for (c = start; c <= end && c < 256; c++)    for (c = start; c <= end && c < 256; c++)
4133      {      {
4134      SETBIT(classbits, cd->fcc[c]);      SETBIT(classbits, cd->fcc[c]);
4135      n8++;      n8++;
4136      }      }
4137    }    }
4138    
4139  /* 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
4140  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
4141  in all cases. */  in all cases. */
4142    
4143  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
4144  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4145    if ((options & PCRE_UTF8) == 0)    if ((options & PCRE_UTF8) == 0)
4146  #endif  #endif
4147    if (end > 0xff) end = 0xff;    if (end > 0xff) end = 0xff;
 #endif  
4148    
4149  #ifdef COMPILE_PCRE16  #elif defined COMPILE_PCRE16
4150  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4151    if ((options & PCRE_UTF16) == 0)    if ((options & PCRE_UTF16) == 0)
4152  #endif  #endif
4153    if (end > 0xffff) end = 0xffff;    if (end > 0xffff) end = 0xffff;
4154  #endif  
4155    #endif /* COMPILE_PCRE[8|16] */
4156    
4157  /* If all characters are less than 256, use the bit map. Otherwise use extra  /* If all characters are less than 256, use the bit map. Otherwise use extra
4158  data. */  data. */
# Line 3477  if (end < 0x100) Line 4161  if (end < 0x100)
4161    {    {
4162    for (c = start; c <= end; c++)    for (c = start; c <= end; c++)
4163      {      {
4164      n8++;      n8++;
4165      SETBIT(classbits, c);      SETBIT(classbits, c);
4166      }      }
4167    }    }
4168    
4169  else  else
4170    {    {
4171    pcre_uchar *uchardata = *uchardptr;    pcre_uchar *uchardata = *uchardptr;
4172    
4173  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4174    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
4175      {      {
4176      if (start < end)      if (start < end)
4177        {        {
4178        *uchardata++ = XCL_RANGE;        *uchardata++ = XCL_RANGE;
4179        uchardata += PRIV(ord2utf)(start, uchardata);        uchardata += PRIV(ord2utf)(start, uchardata);
4180        uchardata += PRIV(ord2utf)(end, uchardata);        uchardata += PRIV(ord2utf)(end, uchardata);
4181        }        }
4182      else if (start == end)      else if (start == end)
4183        {        {
4184        *uchardata++ = XCL_SINGLE;        *uchardata++ = XCL_SINGLE;
4185        uchardata += PRIV(ord2utf)(start, uchardata);        uchardata += PRIV(ord2utf)(start, uchardata);
4186        }        }
4187      }      }
4188    else    else
4189  #endif  /* SUPPORT_UTF */  #endif  /* SUPPORT_UTF */
4190    
4191    /* Without UTF support, character values are constrained by the bit length,    /* Without UTF support, character values are constrained by the bit length,
4192    and can only be > 256 for 16-bit and 32-bit libraries. */    and can only be > 256 for 16-bit and 32-bit libraries. */
4193    
4194  #ifdef COMPILE_PCRE8  #ifdef COMPILE_PCRE8
4195      {}      {}
4196  #else  #else
4197    if (start < end)    if (start < end)
4198      {      {
4199      *uchardata++ = XCL_RANGE;      *uchardata++ = XCL_RANGE;
# Line 3520  else Line 4204  else
4204      {      {
4205      *uchardata++ = XCL_SINGLE;      *uchardata++ = XCL_SINGLE;
4206      *uchardata++ = start;      *uchardata++ = start;
4207      }      }
4208  #endif  #endif
4209    
4210    *uchardptr = uchardata;   /* Updata extra data pointer */    *uchardptr = uchardata;   /* Updata extra data pointer */
4211    }    }
4212    
4213  return n8;    /* Number of 8-bit characters */  return n8;    /* Number of 8-bit characters */
4214  }  }
4215    
4216    
4217    
4218    
4219  /*************************************************  /*************************************************
4220  *        Add a list of characters to a class     *  *        Add a list of characters to a class     *
4221  *************************************************/  *************************************************/
4222    
4223  /* 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
4224  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
4225  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
4226  handled appropriately. This function is mutually recursive with the function  handled appropriately. This function is mutually recursive with the function
# Line 3546  Arguments: Line 4230  Arguments:
4230    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4231    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4232    options       the options word    options       the options word
4233    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4234    p             points to row of 32-bit values, terminated by NOTACHAR    p             points to row of 32-bit values, terminated by NOTACHAR
4235    except        character to omit; this is used when adding lists of    except        character to omit; this is used when adding lists of
4236                    case-equivalent characters to avoid including the one we                    case-equivalent characters to avoid including the one we
4237                    already know about                    already know about
4238    
4239  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4240                  the pointer to extra data is updated                  the pointer to extra data is updated
4241  */  */
# Line 3565  while (p[0] < NOTACHAR) Line 4249  while (p[0] < NOTACHAR)
4249    {    {
4250    int n = 0;    int n = 0;
4251    if (p[0] != except)    if (p[0] != except)
4252      {      {
4253      while(p[n+1] == p[0] + n + 1) n++;      while(p[n+1] == p[0] + n + 1) n++;
4254      n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);      n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4255      }      }
4256    p += n + 1;    p += n + 1;
4257    }    }
4258  return n8;  return n8;
4259  }  }
4260    
4261    
4262    
# Line 3587  Arguments: Line 4271  Arguments:
4271    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4272    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4273    options       the options word    options       the options word
4274    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4275    p             points to row of 32-bit values, terminated by NOTACHAR    p             points to row of 32-bit values, terminated by NOTACHAR
4276    
4277  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4278                  the pointer to extra data is updated                  the pointer to extra data is updated
4279  */  */
4280    
4281  static int  static int
4282  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
4283    int options, compile_data *cd, const pcre_uint32 *p)    int options, compile_data *cd, const pcre_uint32 *p)
4284  {  {
4285    BOOL utf = (options & PCRE_UTF8) != 0;
4286  int n8 = 0;  int n8 = 0;
4287  if (p[0] > 0)  if (p[0] > 0)
4288    n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);    n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
4289  while (p[0] < NOTACHAR)  while (p[0] < NOTACHAR)
4290    {    {
4291    while (p[1] == p[0] + 1) p++;    while (p[1] == p[0] + 1) p++;
4292    n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,    n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
4293      (p[1] == NOTACHAR)? 0x10ffff : p[1] - 1);      (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
4294    p++;    p++;
4295    }    }
4296  return n8;  return n8;
4297  }  }
4298    
4299    
4300    
# Line 3624  to find out the amount of memory needed, Line 4309  to find out the amount of memory needed,
4309  phase. The value of lengthptr distinguishes the two phases.  phase. The value of lengthptr distinguishes the two phases.
4310    
4311  Arguments:  Arguments:
4312    optionsptr     pointer to the option bits    optionsptr        pointer to the option bits
4313    codeptr        points to the pointer to the current code point    codeptr           points to the pointer to the current code point
4314    ptrptr         points to the current pattern pointer    ptrptr            points to the current pattern pointer
4315    errorcodeptr   points to error code variable    errorcodeptr      points to error code variable
4316    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr      place to put the first required character
4317    reqcharptr     set to the last literal character required, else < 0    firstcharflagsptr place to put the first character flags, or a negative number
4318    bcptr          points to current branch chain    reqcharptr        place to put the last required character
4319    cond_depth     conditional nesting depth    reqcharflagsptr   place to put the last required character flags, or a negative number
4320    cd             contains pointers to tables etc.    bcptr             points to current branch chain
4321    lengthptr      NULL during the real compile phase    cond_depth        conditional nesting depth
4322                   points to length accumulator during pre-compile phase    cd                contains pointers to tables etc.
4323      lengthptr         NULL during the real compile phase
4324                        points to length accumulator during pre-compile phase
4325    
4326  Returns:         TRUE on success  Returns:            TRUE on success
4327                   FALSE, with *errorcodeptr set non-zero on error                      FALSE, with *errorcodeptr set non-zero on error
4328  */  */
4329    
4330  static BOOL  static BOOL
4331  compile_branch(int *optionsptr, pcre_uchar **codeptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
4332    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,    const pcre_uchar **ptrptr, int *errorcodeptr,
4333    pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,    pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
4334      pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
4335      branch_chain *bcptr, int cond_depth,
4336    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
4337  {  {
4338  int repeat_type, op_type;  int repeat_type, op_type;
4339  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
4340  int bravalue = 0;  int bravalue = 0;
4341  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
4342  pcre_int32 firstchar, reqchar;  pcre_uint32 firstchar, reqchar;
4343  pcre_int32 zeroreqchar, zerofirstchar;  pcre_int32 firstcharflags, reqcharflags;
4344    pcre_uint32 zeroreqchar, zerofirstchar;
4345    pcre_int32 zeroreqcharflags, zerofirstcharflags;
4346  pcre_int32 req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
4347  int options = *optionsptr;               /* May change dynamically */  int options = *optionsptr;               /* May change dynamically */
4348  int after_manual_callout = 0;  int after_manual_callout = 0;
4349  int length_prevgroup = 0;  int length_prevgroup = 0;
4350  register int c;  register pcre_uint32 c;
4351    int escape;
4352  register pcre_uchar *code = *codeptr;  register pcre_uchar *code = *codeptr;
4353  pcre_uchar *last_code = code;  pcre_uchar *last_code = code;
4354  pcre_uchar *orig_code = code;  pcre_uchar *orig_code = code;
# Line 3676  must not do this for other options (e.g. Line 4368  must not do this for other options (e.g.
4368  dynamically as we process the pattern. */  dynamically as we process the pattern. */
4369    
4370  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4371  /* PCRE_UTF16 has the same value as PCRE_UTF8. */  /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
4372  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
4373    #ifndef COMPILE_PCRE32
4374  pcre_uchar utf_chars[6];  pcre_uchar utf_chars[6];
4375    #endif
4376  #else  #else
4377  BOOL utf = FALSE;  BOOL utf = FALSE;
4378  #endif  #endif
4379    
4380  /* Helper variables for OP_XCLASS opcode (for characters > 255). */  /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
4381    class_uchardata always so that it can be passed to add_to_class() always,
4382    though it will not be used in non-UTF 8-bit cases. This avoids having to supply
4383    alternative calls for the different cases. */
4384    
4385    pcre_uchar *class_uchardata;
4386  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4387  BOOL xclass;  BOOL xclass;
 pcre_uchar *class_uchardata;  
4388  pcre_uchar *class_uchardata_base;  pcre_uchar *class_uchardata_base;
4389  #endif  #endif
4390    
# Line 3710  to take the zero repeat into account. Th Line 4407  to take the zero repeat into account. Th
4407  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
4408  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
4409    
4410  firstchar = reqchar = zerofirstchar = zeroreqchar = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
4411    firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
4412    
4413  /* The variable req_caseopt contains either the REQ_CASELESS value  /* The variable req_caseopt contains either the REQ_CASELESS value
4414  or zero, according to the current setting of the caseless flag. The  or zero, according to the current setting of the caseless flag. The
# Line 3736  for (;; ptr++) Line 4434  for (;; ptr++)
4434    int recno;    int recno;
4435    int refsign;    int refsign;
4436    int skipbytes;    int skipbytes;
4437    int subreqchar;    pcre_uint32 subreqchar, subfirstchar;
4438    int subfirstchar;    pcre_int32 subreqcharflags, subfirstcharflags;
4439    int terminator;    int terminator;
4440    int mclength;    unsigned int mclength;
4441    int tempbracount;    unsigned int tempbracount;
4442      pcre_uint32 ec;
4443    pcre_uchar mcbuffer[8];    pcre_uchar mcbuffer[8];
4444    
4445    /* Get next character in the pattern */    /* Get next character in the pattern */
# Line 3750  for (;; ptr++) Line 4449  for (;; ptr++)
4449    /* 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
4450    string. Nesting only happens one level deep. */    string. Nesting only happens one level deep. */
4451    
4452    if (c == 0 && nestptr != NULL)    if (c == CHAR_NULL && nestptr != NULL)
4453      {      {
4454      ptr = nestptr;      ptr = nestptr;
4455      nestptr = NULL;      nestptr = NULL;
# Line 3825  for (;; ptr++) Line 4524  for (;; ptr++)
4524    
4525    /* 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 */
4526    
4527    if (inescq && c != 0)    if (inescq && c != CHAR_NULL)
4528      {      {
4529      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4530        {        {
# Line 3850  for (;; ptr++) Line 4549  for (;; ptr++)
4549        }        }
4550      }      }
4551    
   /* Fill in length of a previous callout, except when the next thing is  
   a quantifier. */  
   
4552    is_quantifier =    is_quantifier =
4553      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4554      (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));      (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4555    
4556    if (!is_quantifier && previous_callout != NULL &&    /* Fill in length of a previous callout, except when the next thing is a
4557      quantifier or when processing a property substitution string in UCP mode. */
4558    
4559      if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4560         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
4561      {      {
4562      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
# Line 3873  for (;; ptr++) Line 4572  for (;; ptr++)
4572      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
4573        {        {
4574        ptr++;        ptr++;
4575        while (*ptr != 0)        while (*ptr != CHAR_NULL)
4576          {          {
4577          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
4578          ptr++;          ptr++;
# Line 3881  for (;; ptr++) Line 4580  for (;; ptr++)
4580          if (utf) FORWARDCHAR(ptr);          if (utf) FORWARDCHAR(ptr);
4581  #endif  #endif
4582          }          }
4583        if (*ptr != 0) continue;        if (*ptr != CHAR_NULL) continue;
4584    
4585        /* Else fall through to handle end of string */        /* Else fall through to handle end of string */
4586        c = 0;        c = 0;
4587        }        }
4588      }      }
4589    
4590    /* No auto callout for quantifiers. */    /* No auto callout for quantifiers, or while processing property strings that
4591      are substituted for \w etc in UCP mode. */
4592    
4593    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4594      {      {
4595      previous_callout = code;      previous_callout = code;
4596      code = auto_callout(code, ptr, cd);      code = auto_callout(code, ptr, cd);
# Line 3903  for (;; ptr++) Line 4603  for (;; ptr++)
4603      case CHAR_VERTICAL_LINE:       /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
4604      case CHAR_RIGHT_PARENTHESIS:      case CHAR_RIGHT_PARENTHESIS:
4605      *firstcharptr = firstchar;      *firstcharptr = firstchar;
4606        *firstcharflagsptr = firstcharflags;
4607      *reqcharptr = reqchar;      *reqcharptr = reqchar;
4608        *reqcharflagsptr = reqcharflags;
4609      *codeptr = code;      *codeptr = code;
4610      *ptrptr = ptr;      *ptrptr = ptr;
4611      if (lengthptr != NULL)      if (lengthptr != NULL)
# Line 3927  for (;; ptr++) Line 4629  for (;; ptr++)
4629      previous = NULL;      previous = NULL;
4630      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
4631        {        {