/[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 964 by ph10, Fri May 4 13:03:39 2012 UTC revision 1387 by ph10, Sat Nov 2 18:29:05 2013 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2012 University of Cambridge             Copyright (c) 1997-2013 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When PCRE_DEBUG is defined, we need the pcre(16)_printint() function, which  /* When PCRE_DEBUG is defined, we need the pcre(16|32)_printint() function, which
57  is also used by pcretest. PCRE_DEBUG is not defined when building a production  is also used by pcretest. PCRE_DEBUG is not defined when building a production
58  library. We do not need to select pcre16_printint.c specially, because the  library. We do not need to select pcre16_printint.c specially, because the
59  COMPILE_PCREx macro will already be appropriately set. */  COMPILE_PCREx macro will already be appropriately set. */
# Line 68  COMPILE_PCREx macro will already be appr Line 68  COMPILE_PCREx macro will already be appr
68    
69  /* Macro for setting individual bits in class bitmaps. */  /* Macro for setting individual bits in class bitmaps. */
70    
71  #define SETBIT(a,b) a[b/8] |= (1 << (b%8))  #define SETBIT(a,b) a[(b)/8] |= (1 << ((b)&7))
72    
73  /* Maximum length value to check against when making sure that the integer that  /* Maximum length value to check against when making sure that the integer that
74  holds the compiled pattern length does not overflow. We make it a bit less than  holds the compiled pattern length does not overflow. We make it a bit less than
# Line 77  to check them every time. */ Line 77  to check them every time. */
77    
78  #define OFLOW_MAX (INT_MAX - 20)  #define OFLOW_MAX (INT_MAX - 20)
79    
80    /* Definitions to allow mutual recursion */
81    
82    static int
83      add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84        const pcre_uint32 *, unsigned int);
85    
86    static BOOL
87      compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
88        pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
89        compile_data *, int *);
90    
91    
92    
93  /*************************************************  /*************************************************
94  *      Code parameters and static tables         *  *      Code parameters and static tables         *
# Line 103  kicks in at the same number of forward r Line 115  kicks in at the same number of forward r
115  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
116  #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)  #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
117    
118    /* This value determines the size of the initial vector that is used for
119    remembering named groups during the pre-compile. It is allocated on the stack,
120    but if it is too small, it is expanded using malloc(), in a similar way to the
121    workspace. The value is the number of slots in the list. */
122    
123    #define NAMED_GROUP_LIST_SIZE  20
124    
125  /* The overrun tests check for a slightly smaller size so that they detect the  /* The overrun tests check for a slightly smaller size so that they detect the
126  overrun before it actually does run off the end of the data block. */  overrun before it actually does run off the end of the data block. */
127    
# Line 110  overrun before it actually does run off Line 129  overrun before it actually does run off
129    
130  /* Private flags added to firstchar and reqchar. */  /* Private flags added to firstchar and reqchar. */
131    
132  #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */  #define REQ_CASELESS    (1 << 0)        /* Indicates caselessness */
133  #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */  #define REQ_VARY        (1 << 1)        /* Reqchar followed non-literal item */
134    /* Negative values for the firstchar and reqchar flags */
135    #define REQ_UNSET       (-2)
136    #define REQ_NONE        (-1)
137    
138  /* Repeated character flags. */  /* Repeated character flags. */
139    
# Line 242  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 253  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 280  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 307  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 360  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 375  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 440  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 472  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 490  static const char error_texts[] = Line 523  static const char error_texts[] =
523    "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"    "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
524    "invalid UTF-16 string\0"    "invalid UTF-16 string\0"
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"
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    ;    ;
535    
536  /* 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 630  static const pcre_uint8 ebcdic_chartab[] Line 670  static const pcre_uint8 ebcdic_chartab[]
670  #endif  #endif
671    
672    
673  /* Definition to allow mutual recursion */  /* This table is used to check whether auto-possessification is possible
674    between adjacent character-type opcodes. The left-hand (repeated) opcode is
675    used to select the row, and the right-hand opcode is use to select the column.
676    A value of 1 means that auto-possessification is OK. For example, the second
677    value in the first row means that \D+\d can be turned into \D++\d.
678    
679    The Unicode property types (\P and \p) have to be present to fill out the table
680    because of what their opcode values are, but the table values should always be
681    zero because property types are handled separately in the code. The last four
682    columns apply to items that cannot be repeated, so there is no need to have
683    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
684    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
685    
686    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
687    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
688    
689    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
690    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
691      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
692      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
693      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
694      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
695      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
696      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
697      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
698      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
699      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
700      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
701      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
702      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
703      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
704      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
705      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
706      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
707      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
708    };
709    
710  static BOOL  
711    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,  /* This table is used to check whether auto-possessification is possible
712      int *, int *, branch_chain *, compile_data *, int *);  between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
713    left-hand (repeated) opcode is used to select the row, and the right-hand
714    opcode is used to select the column. The values are as follows:
715    
716      0   Always return FALSE (never auto-possessify)
717      1   Character groups are distinct (possessify if both are OP_PROP)
718      2   Check character categories in the same group (general or particular)
719      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
720    
721      4   Check left general category vs right particular category
722      5   Check right general category vs left particular category
723    
724      6   Left alphanum vs right general category
725      7   Left space vs right general category
726      8   Left word vs right general category
727    
728      9   Right alphanum vs left general category
729     10   Right space vs left general category
730     11   Right word vs left general category
731    
732     12   Left alphanum vs right particular category
733     13   Left space vs right particular category
734     14   Left word vs right particular category
735    
736     15   Right alphanum vs left particular category
737     16   Right space vs left particular category
738     17   Right word vs left particular category
739    */
740    
741    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
742    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
743      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
744      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
745      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
746      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
747      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
748      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
749      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
750      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
751      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
752      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
753      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
754    };
755    
756    /* This table is used to check whether auto-possessification is possible
757    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
758    specifies a general category and the other specifies a particular category. The
759    row is selected by the general category and the column by the particular
760    category. The value is 1 if the particular category is not part of the general
761    category. */
762    
763    static const pcre_uint8 catposstab[7][30] = {
764    /* 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 */
765      { 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 */
766      { 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 */
767      { 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 */
768      { 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 */
769      { 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 */
770      { 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 */
771      { 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 */
772    };
773    
774    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
775    a general or particular category. The properties in each row are those
776    that apply to the character set in question. Duplication means that a little
777    unnecessary work is done when checking, but this keeps things much simpler
778    because they can all use the same code. For more details see the comment where
779    this table is used.
780    
781    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
782    "space", but from Perl 5.18 it's included, so both categories are treated the
783    same here. */
784    
785    static const pcre_uint8 posspropstab[3][4] = {
786      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
787      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
788      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
789    };
790    
791    /* This table is used when converting repeating opcodes into possessified
792    versions as a result of an explicit possessive quantifier such as ++. A zero
793    value means there is no possessified version - in those cases the item in
794    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
795    because all relevant opcodes are less than that. */
796    
797    static const pcre_uint8 opcode_possessify[] = {
798      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
799      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
800    
801      0,                       /* NOTI */
802      OP_POSSTAR, 0,           /* STAR, MINSTAR */
803      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
804      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
805      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
806      0,                       /* EXACT */
807      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
808    
809      OP_POSSTARI, 0,          /* STARI, MINSTARI */
810      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
811      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
812      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
813      0,                       /* EXACTI */
814      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
815    
816      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
817      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
818      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
819      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
820      0,                       /* NOTEXACT */
821      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
822    
823      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
824      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
825      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
826      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
827      0,                       /* NOTEXACTI */
828      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
829    
830      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
831      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
832      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
833      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
834      0,                       /* TYPEEXACT */
835      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
836    
837      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
838      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
839      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
840      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
841      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
842    
843      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
844      0, 0,                    /* REF, REFI */
845      0, 0,                    /* DNREF, DNREFI */
846      0, 0                     /* RECURSE, CALLOUT */
847    };
848    
849    
850    
# Line 657  find_error_text(int n) Line 867  find_error_text(int n)
867  const char *s = error_texts;  const char *s = error_texts;
868  for (; n > 0; n--)  for (; n > 0; n--)
869    {    {
870    while (*s++ != 0) {};    while (*s++ != CHAR_NULL) {};
871    if (*s == 0) return "Error text not found (please report)";    if (*s == CHAR_NULL) return "Error text not found (please report)";
872    }    }
873  return s;  return s;
874  }  }
875    
876    
877    
878  /*************************************************  /*************************************************
879  *           Expand the workspace                 *  *           Expand the workspace                 *
880  *************************************************/  *************************************************/
# Line 741  return (*p == CHAR_RIGHT_CURLY_BRACKET); Line 952  return (*p == CHAR_RIGHT_CURLY_BRACKET);
952  *************************************************/  *************************************************/
953    
954  /* 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
955  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
956  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.
957  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
958  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
959  ptr is pointing at the \. On exit, it is on the final character of the escape  character of the escape sequence.
 sequence.  
960    
961  Arguments:  Arguments:
962    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
963      chptr          points to a returned data character
964    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
965    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
966    options        the options bits    options        the options bits
967    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
968    
969  Returns:         zero or positive => a data character  Returns:         zero => a data character
970                   negative => a special escape sequence                   positive => a special escape sequence
971                     negative => a back reference
972                   on error, errorcodeptr is set                   on error, errorcodeptr is set
973  */  */
974    
975  static int  static int
976  check_escape(const pcre_uchar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
977    int options, BOOL isclass)    int bracount, int options, BOOL isclass)
978  {  {
979  /* PCRE_UTF16 has the same value as PCRE_UTF8. */  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
980  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
981  const pcre_uchar *ptr = *ptrptr + 1;  const pcre_uchar *ptr = *ptrptr + 1;
982  pcre_int32 c;  pcre_uint32 c;
983    int escape = 0;
984  int i;  int i;
985    
986  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
# Line 775  ptr--;                            /* Set Line 988  ptr--;                            /* Set
988    
989  /* 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. */
990    
991  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
992    
993  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
994  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 784  Otherwise further processing may be requ Line 997  Otherwise further processing may be requ
997  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
998  /* Not alphanumeric */  /* Not alphanumeric */
999  else if (c < CHAR_0 || c > CHAR_z) {}  else if (c < CHAR_0 || c > CHAR_z) {}
1000  else if ((i = escapes[c - CHAR_0]) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0)
1001      { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1002    
1003  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1004  /* Not alphanumeric */  /* Not alphanumeric */
1005  else if (c < 'a' || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}  else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
1006  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; }
1007  #endif  #endif
1008    
1009  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
# Line 797  else if ((i = escapes[c - 0x48]) != 0) Line 1011  else if ((i = escapes[c - 0x48]) != 0)
1011  else  else
1012    {    {
1013    const pcre_uchar *oldptr;    const pcre_uchar *oldptr;
1014    BOOL braced, negated;    BOOL braced, negated, overflow;
1015      int s;
1016    
1017    switch (c)    switch (c)
1018      {      {
# Line 822  else Line 1037  else
1037          c = 0;          c = 0;
1038          for (i = 0; i < 4; ++i)          for (i = 0; i < 4; ++i)
1039            {            {
1040            register int cc = *(++ptr);            register pcre_uint32 cc = *(++ptr);
1041  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1042            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1043            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 831  else Line 1046  else
1046            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1047  #endif  #endif
1048            }            }
1049    
1050    #if defined COMPILE_PCRE8
1051            if (c > (utf ? 0x10ffffU : 0xffU))
1052    #elif defined COMPILE_PCRE16
1053            if (c > (utf ? 0x10ffffU : 0xffffU))
1054    #elif defined COMPILE_PCRE32
1055            if (utf && c > 0x10ffffU)
1056    #endif
1057              {
1058              *errorcodeptr = ERR76;
1059              }
1060            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1061          }          }
1062        }        }
1063      else      else
# Line 857  else Line 1084  else
1084      (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
1085      number either in angle brackets or in single quotes. However, these are      number either in angle brackets or in single quotes. However, these are
1086      (possibly recursive) subroutine calls, _not_ backreferences. Just return      (possibly recursive) subroutine calls, _not_ backreferences. Just return
1087      the -ESC_g code (cf \k). */      the ESC_g code (cf \k). */
1088    
1089      case CHAR_g:      case CHAR_g:
1090      if (isclass) break;      if (isclass) break;
1091      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1092        {        {
1093        c = -ESC_g;        escape = ESC_g;
1094        break;        break;
1095        }        }
1096    
# Line 872  else Line 1099  else
1099      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1100        {        {
1101        const pcre_uchar *p;        const pcre_uchar *p;
1102        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++)
1103          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1104        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)        if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1105          {          {
1106          c = -ESC_k;          escape = ESC_k;
1107          break;          break;
1108          }          }
1109        braced = TRUE;        braced = TRUE;
# Line 892  else Line 1119  else
1119      else negated = FALSE;      else negated = FALSE;
1120    
1121      /* The integer range is limited by the machine's int representation. */      /* The integer range is limited by the machine's int representation. */
1122      c = 0;      s = 0;
1123        overflow = FALSE;
1124      while (IS_DIGIT(ptr[1]))      while (IS_DIGIT(ptr[1]))
1125        {        {
1126        if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */        if (s > INT_MAX / 10 - 1) /* Integer overflow */
1127          {          {
1128          c = -1;          overflow = TRUE;
1129          break;          break;
1130          }          }
1131        c = c * 10 + *(++ptr) - CHAR_0;        s = s * 10 + (int)(*(++ptr) - CHAR_0);
1132        }        }
1133      if (((unsigned int)c) > INT_MAX) /* Integer overflow */      if (overflow) /* Integer overflow */
1134        {        {
1135        while (IS_DIGIT(ptr[1]))        while (IS_DIGIT(ptr[1]))
1136          ptr++;          ptr++;
# Line 916  else Line 1144  else
1144        break;        break;
1145        }        }
1146    
1147      if (c == 0)      if (s == 0)
1148        {        {
1149        *errorcodeptr = ERR58;        *errorcodeptr = ERR58;
1150        break;        break;
# Line 924  else Line 1152  else
1152    
1153      if (negated)      if (negated)
1154        {        {
1155        if (c > bracount)        if (s > bracount)
1156          {          {
1157          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
1158          break;          break;
1159          }          }
1160        c = bracount - (c - 1);        s = bracount - (s - 1);
1161        }        }
1162    
1163      c = -(ESC_REF + c);      escape = -s;
1164      break;      break;
1165    
1166      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1167      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
1168      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1169        recommended to avoid the ambiguities in the old syntax.
1170    
1171      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
1172      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
1173      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
1174      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
1175      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
1176      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
1177      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1178    
1179        Inside a character class, \ followed by a digit is always either a literal
1180        8 or 9 or an octal number. */
1181    
1182      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:
1183      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 954  else Line 1186  else
1186        {        {
1187        oldptr = ptr;        oldptr = ptr;
1188        /* The integer range is limited by the machine's int representation. */        /* The integer range is limited by the machine's int representation. */
1189        c -= CHAR_0;        s = (int)(c -CHAR_0);
1190          overflow = FALSE;
1191        while (IS_DIGIT(ptr[1]))        while (IS_DIGIT(ptr[1]))
1192          {          {
1193          if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */          if (s > INT_MAX / 10 - 1) /* Integer overflow */
1194            {            {
1195            c = -1;            overflow = TRUE;
1196            break;            break;
1197            }            }
1198          c = c * 10 + *(++ptr) - CHAR_0;          s = s * 10 + (int)(*(++ptr) - CHAR_0);
1199          }          }
1200        if (((unsigned int)c) > INT_MAX) /* Integer overflow */        if (overflow) /* Integer overflow */
1201          {          {
1202          while (IS_DIGIT(ptr[1]))          while (IS_DIGIT(ptr[1]))
1203            ptr++;            ptr++;
1204          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1205          break;          break;
1206          }          }
1207        if (c < 10 || c <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1208          {          {
1209          c = -(ESC_REF + c);          escape = -s;
1210          break;          break;
1211          }          }
1212        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1213        }        }
1214    
1215      /* 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
1216      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
1217      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
1218        changed so as not to insert the binary zero. */
1219    
1220      if ((c = *ptr) >= CHAR_8)      if ((c = *ptr) >= CHAR_8) break;
1221        {  
1222        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1223    
1224      /* \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
1225      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 1005  else Line 1236  else
1236  #endif  #endif
1237      break;      break;
1238    
1239      /* \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
1240      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}. */
1241      If not, { is treated as a data character. */  
1242        case CHAR_o:
1243        if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1244          {
1245          ptr += 2;
1246          c = 0;
1247          overflow = FALSE;
1248          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1249            {
1250            register pcre_uint32 cc = *ptr++;
1251            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1252    #ifdef COMPILE_PCRE32
1253            if (c >= 0x20000000l) { overflow = TRUE; break; }
1254    #endif
1255            c = (c << 3) + cc - CHAR_0 ;
1256    #if defined COMPILE_PCRE8
1257            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1258    #elif defined COMPILE_PCRE16
1259            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1260    #elif defined COMPILE_PCRE32
1261            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1262    #endif
1263            }
1264          if (overflow)
1265            {
1266            while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1267            *errorcodeptr = ERR34;
1268            }
1269          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1270            {
1271            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1272            }
1273          else *errorcodeptr = ERR80;
1274          }
1275        break;
1276    
1277        /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1278        numbers. Otherwise it is a lowercase x letter. */
1279    
1280      case CHAR_x:      case CHAR_x:
1281      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1282        {        {
       /* In JavaScript, \x must be followed by two hexadecimal numbers.  
       Otherwise it is a lowercase x letter. */  
1283        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1284          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1285          {          {
1286          c = 0;          c = 0;
1287          for (i = 0; i < 2; ++i)          for (i = 0; i < 2; ++i)
1288            {            {
1289            register int cc = *(++ptr);            register pcre_uint32 cc = *(++ptr);
1290  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1291            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1292            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 1030  else Line 1296  else
1296  #endif  #endif
1297            }            }
1298          }          }
1299        break;        }    /* End JavaScript handling */
       }  
1300    
1301      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1302        {      greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1303        const pcre_uchar *pt = ptr + 2;      digits. If not, { used to be treated as a data character. However, Perl
1304        seems to read hex digits up to the first non-such, and ignore the rest, so
1305        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1306        now gives an error. */
1307    
1308        c = 0;      else
1309        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)        {
1310          if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1311          {          {
1312          register int cc = *pt++;          ptr += 2;
1313          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          c = 0;
1314            overflow = FALSE;
1315            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1316              {
1317              register pcre_uint32 cc = *ptr++;
1318              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1319    
1320    #ifdef COMPILE_PCRE32
1321              if (c >= 0x10000000l) { overflow = TRUE; break; }
1322    #endif
1323    
1324  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1325          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1326          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1327  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1328          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 */
1329          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1330  #endif  #endif
1331    
1332  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
1333          if (c > (utf ? 0x10ffff : 0xff)) { c = -1; break; }            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1334  #else  #elif defined COMPILE_PCRE16
1335  #ifdef COMPILE_PCRE16            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1336          if (c > (utf ? 0x10ffff : 0xffff)) { c = -1; break; }  #elif defined COMPILE_PCRE32
1337  #endif            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1338  #endif  #endif
1339          }            }
1340    
1341        if (c < 0)          if (overflow)
1342          {            {
1343          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1344          *errorcodeptr = ERR34;            *errorcodeptr = ERR34;
1345          }            }
1346    
1347        if (*pt == CHAR_RIGHT_CURLY_BRACKET)          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1348          {            {
1349          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1350          ptr = pt;            }
         break;  
         }  
1351    
1352        /* 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.
1353        recognize this construct; fall through to the normal \x handling. */          We used just to recognize this construct and fall through to the normal
1354        }          \x handling, but nowadays Perl gives an error, which seems much more
1355            sensible, so we do too. */
1356    
1357      /* Read just a single-byte hex-defined char */          else *errorcodeptr = ERR79;
1358            }   /* End of \x{} processing */
1359    
1360      c = 0;        /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1361      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)  
1362        {        else
1363        int cc;                                  /* Some compilers don't like */          {
1364        cc = *(++ptr);                           /* ++ in initializers */          c = 0;
1365            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1366              {
1367              pcre_uint32 cc;                          /* Some compilers don't like */
1368              cc = *(++ptr);                           /* ++ in initializers */
1369  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1370        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1371        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1372  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1373        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */            if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1374        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1375  #endif  #endif
1376        }            }
1377            }     /* End of \xdd handling */
1378          }       /* End of Perl-style \x handling */
1379      break;      break;
1380    
1381      /* 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 1101  else Line 1385  else
1385    
1386      case CHAR_c:      case CHAR_c:
1387      c = *(++ptr);      c = *(++ptr);
1388      if (c == 0)      if (c == CHAR_NULL)
1389        {        {
1390        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1391        break;        break;
# Line 1141  else Line 1425  else
1425  newline". PCRE does not support \N{name}. However, it does support  newline". PCRE does not support \N{name}. However, it does support
1426  quantification such as \N{2,3}. */  quantification such as \N{2,3}. */
1427    
1428  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&  if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1429       !is_counted_repeat(ptr+2))       !is_counted_repeat(ptr+2))
1430    *errorcodeptr = ERR37;    *errorcodeptr = ERR37;
1431    
1432  /* If PCRE_UCP is set, we change the values for \d etc. */  /* If PCRE_UCP is set, we change the values for \d etc. */
1433    
1434  if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)  if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1435    c -= (ESC_DU - ESC_D);    escape += (ESC_DU - ESC_D);
1436    
1437  /* Set the pointer to the final character before returning. */  /* Set the pointer to the final character before returning. */
1438    
1439  *ptrptr = ptr;  *ptrptr = ptr;
1440  return c;  *chptr = c;
1441    return escape;
1442  }  }
1443    
1444    
# Line 1171  escape sequence. Line 1456  escape sequence.
1456  Argument:  Argument:
1457    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1458    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
1459    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
1460      pdataptr       points to an unsigned int that is set to the detailed property value
1461    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1462    
1463  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
1464  */  */
1465    
1466  static int  static BOOL
1467  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1468      unsigned int *pdataptr, int *errorcodeptr)
1469  {  {
1470  int c, i, bot, top;  pcre_uchar c;
1471    int i, bot, top;
1472  const pcre_uchar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
1473  pcre_uchar name[32];  pcre_uchar name[32];
1474    
1475  c = *(++ptr);  c = *(++ptr);
1476  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1477    
1478  *negptr = FALSE;  *negptr = FALSE;
1479    
# Line 1202  if (c == CHAR_LEFT_CURLY_BRACKET) Line 1490  if (c == CHAR_LEFT_CURLY_BRACKET)
1490    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1491      {      {
1492      c = *(++ptr);      c = *(++ptr);
1493      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1494      if (c == CHAR_RIGHT_CURLY_BRACKET) break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1495      name[i] = c;      name[i] = c;
1496      }      }
# Line 1227  top = PRIV(utt_size); Line 1515  top = PRIV(utt_size);
1515    
1516  while (bot < top)  while (bot < top)
1517    {    {
1518      int r;
1519    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1520    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);
1521    if (c == 0)    if (r == 0)
1522      {      {
1523      *dptr = PRIV(utt)[i].value;      *ptypeptr = PRIV(utt)[i].type;
1524      return PRIV(utt)[i].type;      *pdataptr = PRIV(utt)[i].value;
1525        return TRUE;
1526      }      }
1527    if (c > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1528    }    }
1529    
1530  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1531  *ptrptr = ptr;  *ptrptr = ptr;
1532  return -1;  return FALSE;
1533    
1534  ERROR_RETURN:  ERROR_RETURN:
1535  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1536  *ptrptr = ptr;  *ptrptr = ptr;
1537  return -1;  return FALSE;
1538  }  }
1539  #endif  #endif
1540    
1541    
1542    
   
1543  /*************************************************  /*************************************************
1544  *         Read repeat counts                     *  *         Read repeat counts                     *
1545  *************************************************/  *************************************************/
# Line 1279  int max = -1; Line 1568  int max = -1;
1568  /* 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
1569  an integer overflow. */  an integer overflow. */
1570    
1571  while (IS_DIGIT(*p)) min = min * 10 + *p++ - CHAR_0;  while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);
1572  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1573    {    {
1574    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 1294  if (*p == CHAR_RIGHT_CURLY_BRACKET) max Line 1583  if (*p == CHAR_RIGHT_CURLY_BRACKET) max
1583    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1584      {      {
1585      max = 0;      max = 0;
1586      while(IS_DIGIT(*p)) max = max * 10 + *p++ - CHAR_0;      while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);
1587      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1588        {        {
1589        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 1319  return p; Line 1608  return p;
1608    
1609    
1610  /*************************************************  /*************************************************
 *  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;  
 }  
   
   
   
   
 /*************************************************  
1611  *      Find first significant op code            *  *      Find first significant op code            *
1612  *************************************************/  *************************************************/
1613    
# Line 1652  for (;;) Line 1646  for (;;)
1646    
1647      case OP_CALLOUT:      case OP_CALLOUT:
1648      case OP_CREF:      case OP_CREF:
1649      case OP_NCREF:      case OP_DNCREF:
1650      case OP_RREF:      case OP_RREF:
1651      case OP_NRREF:      case OP_DNRREF:
1652      case OP_DEF:      case OP_DEF:
1653      code += PRIV(OP_lengths)[*code];      code += PRIV(OP_lengths)[*code];
1654      break;      break;
# Line 1668  for (;;) Line 1662  for (;;)
1662    
1663    
1664    
   
1665  /*************************************************  /*************************************************
1666  *        Find the fixed length of a branch       *  *        Find the fixed length of a branch       *
1667  *************************************************/  *************************************************/
# Line 1686  and doing the check at the end; a flag s Line 1679  and doing the check at the end; a flag s
1679    
1680  Arguments:  Arguments:
1681    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1682    utf      TRUE in UTF-8 / UTF-16 mode    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1683    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1684    cd       the "compile data" structure    cd       the "compile data" structure
1685    
# Line 1712  for (;;) Line 1705  for (;;)
1705    {    {
1706    int d;    int d;
1707    pcre_uchar *ce, *cs;    pcre_uchar *ce, *cs;
1708    register int op = *cc;    register pcre_uchar op = *cc;
1709    
1710    switch (op)    switch (op)
1711      {      {
# Line 1792  for (;;) Line 1785  for (;;)
1785      case OP_COMMIT:      case OP_COMMIT:
1786      case OP_CREF:      case OP_CREF:
1787      case OP_DEF:      case OP_DEF:
1788        case OP_DNCREF:
1789        case OP_DNRREF:
1790      case OP_DOLL:      case OP_DOLL:
1791      case OP_DOLLM:      case OP_DOLLM:
1792      case OP_EOD:      case OP_EOD:
1793      case OP_EODN:      case OP_EODN:
1794      case OP_FAIL:      case OP_FAIL:
     case OP_NCREF:  
     case OP_NRREF:  
1795      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1796      case OP_PRUNE:      case OP_PRUNE:
1797      case OP_REVERSE:      case OP_REVERSE:
# Line 1832  for (;;) Line 1825  for (;;)
1825      case OP_EXACTI:      case OP_EXACTI:
1826      case OP_NOTEXACT:      case OP_NOTEXACT:
1827      case OP_NOTEXACTI:      case OP_NOTEXACTI:
1828      branchlength += GET2(cc,1);      branchlength += (int)GET2(cc,1);
1829      cc += 2 + IMM2_SIZE;      cc += 2 + IMM2_SIZE;
1830  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
1831      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
# Line 1841  for (;;) Line 1834  for (;;)
1834    
1835      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1836      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1837      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)
1838          cc += 2;
1839      cc += 1 + IMM2_SIZE + 1;      cc += 1 + IMM2_SIZE + 1;
1840      break;      break;
1841    
# Line 1876  for (;;) Line 1870  for (;;)
1870    
1871      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1872    
 #if defined SUPPORT_UTF || defined COMPILE_PCRE16  
     case OP_XCLASS:  
     cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];  
     /* Fall through */  
 #endif  
   
1873      case OP_CLASS:      case OP_CLASS:
1874      case OP_NCLASS:      case OP_NCLASS:
1875    #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1876        case OP_XCLASS:
1877        /* The original code caused an unsigned overflow in 64 bit systems,
1878        so now we use a conditional statement. */
1879        if (op == OP_XCLASS)
1880          cc += GET(cc, 1);
1881        else
1882          cc += PRIV(OP_lengths)[OP_CLASS];
1883    #else
1884      cc += PRIV(OP_lengths)[OP_CLASS];      cc += PRIV(OP_lengths)[OP_CLASS];
1885    #endif
1886    
1887      switch (*cc)      switch (*cc)
1888        {        {
       case OP_CRPLUS:  
       case OP_CRMINPLUS:  
1889        case OP_CRSTAR:        case OP_CRSTAR:
1890        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1891          case OP_CRPLUS:
1892          case OP_CRMINPLUS:
1893        case OP_CRQUERY:        case OP_CRQUERY:
1894        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1895          case OP_CRPOSSTAR:
1896          case OP_CRPOSPLUS:
1897          case OP_CRPOSQUERY:
1898        return -1;        return -1;
1899    
1900        case OP_CRRANGE:        case OP_CRRANGE:
1901        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1902          case OP_CRPOSRANGE:
1903        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1904        branchlength += GET2(cc,1);        branchlength += (int)GET2(cc,1);
1905        cc += 1 + 2 * IMM2_SIZE;        cc += 1 + 2 * IMM2_SIZE;
1906        break;        break;
1907    
# Line 1966  for (;;) Line 1968  for (;;)
1968      case OP_QUERYI:      case OP_QUERYI:
1969      case OP_REF:      case OP_REF:
1970      case OP_REFI:      case OP_REFI:
1971        case OP_DNREF:
1972        case OP_DNREFI:
1973      case OP_SBRA:      case OP_SBRA:
1974      case OP_SBRAPOS:      case OP_SBRAPOS:
1975      case OP_SCBRA:      case OP_SCBRA:
# Line 2002  for (;;) Line 2006  for (;;)
2006    
2007    
2008    
   
2009  /*************************************************  /*************************************************
2010  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
2011  *************************************************/  *************************************************/
# Line 2015  length. Line 2018  length.
2018    
2019  Arguments:  Arguments:
2020    code        points to start of expression    code        points to start of expression
2021    utf         TRUE in UTF-8 / UTF-16 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2022    number      the required bracket number or negative to find a lookbehind    number      the required bracket number or negative to find a lookbehind
2023    
2024  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 2026  PRIV(find_bracket)(const pcre_uchar *cod Line 2029  PRIV(find_bracket)(const pcre_uchar *cod
2029  {  {
2030  for (;;)  for (;;)
2031    {    {
2032    register int c = *code;    register pcre_uchar c = *code;
2033    
2034    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2035    
# Line 2049  for (;;) Line 2052  for (;;)
2052    else if (c == OP_CBRA || c == OP_SCBRA ||    else if (c == OP_CBRA || c == OP_SCBRA ||
2053             c == OP_CBRAPOS || c == OP_SCBRAPOS)             c == OP_CBRAPOS || c == OP_SCBRAPOS)
2054      {      {
2055      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
2056      if (n == number) return (pcre_uchar *)code;      if (n == number) return (pcre_uchar *)code;
2057      code += PRIV(OP_lengths)[c];      code += PRIV(OP_lengths)[c];
2058      }      }
# Line 2079  for (;;) Line 2082  for (;;)
2082        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2083        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2084        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2085        if (code[1 + IMM2_SIZE] == OP_PROP        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2086          || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;          code += 2;
2087        break;        break;
2088    
2089        case OP_MARK:        case OP_MARK:
2090        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2091        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2092        case OP_THEN_ARG:        case OP_THEN_ARG:
2093        code += code[1];        code += code[1];
2094        break;        break;
# Line 2102  for (;;) Line 2102  for (;;)
2102    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
2103    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2104    
2105  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2106      if (utf) switch(c)      if (utf) switch(c)
2107        {        {
2108        case OP_CHAR:        case OP_CHAR:
# Line 2154  instance of OP_RECURSE. Line 2154  instance of OP_RECURSE.
2154    
2155  Arguments:  Arguments:
2156    code        points to start of expression    code        points to start of expression
2157    utf         TRUE in UTF-8 / UTF-16 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2158    
2159  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
2160  */  */
# Line 2164  find_recurse(const pcre_uchar *code, BOO Line 2164  find_recurse(const pcre_uchar *code, BOO
2164  {  {
2165  for (;;)  for (;;)
2166    {    {
2167    register int c = *code;    register pcre_uchar c = *code;
2168    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2169    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2170    
# Line 2199  for (;;) Line 2199  for (;;)
2199        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2200        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2201        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2202        if (code[1 + IMM2_SIZE] == OP_PROP        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2203          || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;          code += 2;
2204        break;        break;
2205    
2206        case OP_MARK:        case OP_MARK:
2207        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2208        case OP_SKIP_ARG:        case OP_SKIP_ARG:
2209          case OP_THEN_ARG:
2210        code += code[1];        code += code[1];
2211        break;        break;
2212          }
       case OP_THEN_ARG:  
       code += code[1];  
       break;  
       }  
2213    
2214      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2215    
# Line 2222  for (;;) Line 2219  for (;;)
2219      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
2220      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2221    
2222  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2223      if (utf) switch(c)      if (utf) switch(c)
2224        {        {
2225        case OP_CHAR:        case OP_CHAR:
# Line 2308  bracket whose current branch will alread Line 2305  bracket whose current branch will alread
2305  Arguments:  Arguments:
2306    code        points to start of search    code        points to start of search
2307    endcode     points to where to stop    endcode     points to where to stop
2308    utf         TRUE if in UTF-8 / UTF-16 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2309    cd          contains pointers to tables etc.    cd          contains pointers to tables etc.
2310      recurses    chain of recurse_check to catch mutual recursion
2311    
2312  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2313  */  */
2314    
2315    typedef struct recurse_check {
2316      struct recurse_check *prev;
2317      const pcre_uchar *group;
2318    } recurse_check;
2319    
2320  static BOOL  static BOOL
2321  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2322    BOOL utf, compile_data *cd)    BOOL utf, compile_data *cd, recurse_check *recurses)
2323  {  {
2324  register int c;  register pcre_uchar c;
2325    recurse_check this_recurse;
2326    
2327  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2328       code < endcode;       code < endcode;
2329       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
# Line 2346  for (code = first_significant_code(code Line 2351  for (code = first_significant_code(code
2351    
2352    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2353      {      {
2354      const pcre_uchar *scode;      const pcre_uchar *scode = cd->start_code + GET(code, 1);
2355      BOOL empty_branch;      BOOL empty_branch;
2356    
2357      /* Test for forward reference */      /* Test for forward reference or uncompleted reference. This is disabled
2358        when called to scan a completed pattern by setting cd->start_workspace to
2359        NULL. */
2360    
2361        if (cd->start_workspace != NULL)
2362          {
2363          const pcre_uchar *tcode;
2364          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2365            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2366          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2367          }
2368    
2369        /* If we are scanning a completed pattern, there are no forward references
2370        and all groups are complete. We need to detect whether this is a recursive
2371        call, as otherwise there will be an infinite loop. If it is a recursion,
2372        just skip over it. Simple recursions are easily detected. For mutual
2373        recursions we keep a chain on the stack. */
2374    
2375      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)      else
2376        if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;        {
2377          recurse_check *r = recurses;
2378          const pcre_uchar *endgroup = scode;
2379    
2380      /* Not a forward reference, test for completed backward reference */        do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2381          if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2382    
2383      empty_branch = FALSE;        for (r = recurses; r != NULL; r = r->prev)
2384      scode = cd->start_code + GET(code, 1);          if (r->group == scode) break;
2385      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */        if (r != NULL) continue;   /* Mutual recursion */
2386          }
2387    
2388      /* Completed backwards reference */      /* Completed reference; scan the referenced group, remembering it on the
2389        stack chain to detect mutual recursions. */
2390    
2391        empty_branch = FALSE;
2392        this_recurse.prev = recurses;
2393        this_recurse.group = scode;
2394    
2395      do      do
2396        {        {
2397        if (could_be_empty_branch(scode, endcode, utf, cd))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2398          {          {
2399          empty_branch = TRUE;          empty_branch = TRUE;
2400          break;          break;
# Line 2420  for (code = first_significant_code(code Line 2450  for (code = first_significant_code(code
2450        empty_branch = FALSE;        empty_branch = FALSE;
2451        do        do
2452          {          {
2453          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2454            empty_branch = TRUE;            empty_branch = TRUE;
2455          code += GET(code, 1);          code += GET(code, 1);
2456          }          }
# Line 2462  for (code = first_significant_code(code Line 2492  for (code = first_significant_code(code
2492        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2493        case OP_CRQUERY:        case OP_CRQUERY:
2494        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2495          case OP_CRPOSSTAR:
2496          case OP_CRPOSQUERY:
2497        break;        break;
2498    
2499        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2500        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2501        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2502          case OP_CRPOSPLUS:
2503        return FALSE;        return FALSE;
2504    
2505        case OP_CRRANGE:        case OP_CRRANGE:
2506        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2507          case OP_CRPOSRANGE:
2508        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2509        break;        break;
2510        }        }
# Line 2478  for (code = first_significant_code(code Line 2512  for (code = first_significant_code(code
2512    
2513      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2514    
2515        case OP_ANY:
2516        case OP_ALLANY:
2517        case OP_ANYBYTE:
2518    
2519      case OP_PROP:      case OP_PROP:
2520      case OP_NOTPROP:      case OP_NOTPROP:
2521        case OP_ANYNL:
2522    
2523        case OP_NOT_HSPACE:
2524        case OP_HSPACE:
2525        case OP_NOT_VSPACE:
2526        case OP_VSPACE:
2527      case OP_EXTUNI:      case OP_EXTUNI:
2528    
2529      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2530      case OP_DIGIT:      case OP_DIGIT:
2531      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2532      case OP_WHITESPACE:      case OP_WHITESPACE:
2533      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2534      case OP_WORDCHAR:      case OP_WORDCHAR:
2535      case OP_ANY:  
     case OP_ALLANY:  
     case OP_ANYBYTE:  
2536      case OP_CHAR:      case OP_CHAR:
2537      case OP_CHARI:      case OP_CHARI:
2538      case OP_NOT:      case OP_NOT:
2539      case OP_NOTI:      case OP_NOTI:
2540    
2541      case OP_PLUS:      case OP_PLUS:
2542        case OP_PLUSI:
2543      case OP_MINPLUS:      case OP_MINPLUS:
2544      case OP_POSPLUS:      case OP_MINPLUSI:
2545      case OP_EXACT:  
2546      case OP_NOTPLUS:      case OP_NOTPLUS:
2547        case OP_NOTPLUSI:
2548      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2549        case OP_NOTMINPLUSI:
2550    
2551        case OP_POSPLUS:
2552        case OP_POSPLUSI:
2553      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2554        case OP_NOTPOSPLUSI:
2555    
2556        case OP_EXACT:
2557        case OP_EXACTI:
2558      case OP_NOTEXACT:      case OP_NOTEXACT:
2559        case OP_NOTEXACTI:
2560    
2561      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2562      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2563      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2564      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2565    
2566      return FALSE;      return FALSE;
2567    
2568      /* 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 2525  for (code = first_significant_code(code Line 2582  for (code = first_significant_code(code
2582      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2583      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2584      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2585      if (code[1 + IMM2_SIZE] == OP_PROP      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2586        || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;        code += 2;
2587      break;      break;
2588    
2589      /* End of branch */      /* End of branch */
# Line 2539  for (code = first_significant_code(code Line 2596  for (code = first_significant_code(code
2596      return TRUE;      return TRUE;
2597    
2598      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2599      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2600        followed by a multibyte character. */
2601    
2602  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2603      case OP_STAR:      case OP_STAR:
2604      case OP_STARI:      case OP_STARI:
2605        case OP_NOTSTAR:
2606        case OP_NOTSTARI:
2607    
2608      case OP_MINSTAR:      case OP_MINSTAR:
2609      case OP_MINSTARI:      case OP_MINSTARI:
2610        case OP_NOTMINSTAR:
2611        case OP_NOTMINSTARI:
2612    
2613      case OP_POSSTAR:      case OP_POSSTAR:
2614      case OP_POSSTARI:      case OP_POSSTARI:
2615        case OP_NOTPOSSTAR:
2616        case OP_NOTPOSSTARI:
2617    
2618      case OP_QUERY:      case OP_QUERY:
2619      case OP_QUERYI:      case OP_QUERYI:
2620        case OP_NOTQUERY:
2621        case OP_NOTQUERYI:
2622    
2623      case OP_MINQUERY:      case OP_MINQUERY:
2624      case OP_MINQUERYI:      case OP_MINQUERYI:
2625        case OP_NOTMINQUERY:
2626        case OP_NOTMINQUERYI:
2627    
2628      case OP_POSQUERY:      case OP_POSQUERY:
2629      case OP_POSQUERYI:      case OP_POSQUERYI:
2630        case OP_NOTPOSQUERY:
2631        case OP_NOTPOSQUERYI:
2632    
2633      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2634      break;      break;
2635    
2636      case OP_UPTO:      case OP_UPTO:
2637      case OP_UPTOI:      case OP_UPTOI:
2638        case OP_NOTUPTO:
2639        case OP_NOTUPTOI:
2640    
2641      case OP_MINUPTO:      case OP_MINUPTO:
2642      case OP_MINUPTOI:      case OP_MINUPTOI:
2643        case OP_NOTMINUPTO:
2644        case OP_NOTMINUPTOI:
2645    
2646      case OP_POSUPTO:      case OP_POSUPTO:
2647      case OP_POSUPTOI:      case OP_POSUPTOI:
2648        case OP_NOTPOSUPTO:
2649        case OP_NOTPOSUPTOI:
2650    
2651      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]);
2652      break;      break;
2653  #endif  #endif
# Line 2573  for (code = first_significant_code(code Line 2658  for (code = first_significant_code(code
2658      case OP_MARK:      case OP_MARK:
2659      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
2660      case OP_SKIP_ARG:      case OP_SKIP_ARG:
     code += code[1];  
     break;  
   
2661      case OP_THEN_ARG:      case OP_THEN_ARG:
2662      code += code[1];      code += code[1];
2663      break;      break;
# Line 2607  Arguments: Line 2689  Arguments:
2689    code        points to start of the recursion    code        points to start of the recursion
2690    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2691    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2692    utf         TRUE if in UTF-8 / UTF-16 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2693    cd          pointers to tables etc    cd          pointers to tables etc
2694    
2695  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
# Line 2619  could_be_empty(const pcre_uchar *code, c Line 2701  could_be_empty(const pcre_uchar *code, c
2701  {  {
2702  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2703    {    {
2704    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2705      return FALSE;      return FALSE;
2706    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2707    }    }
# Line 2629  return TRUE; Line 2711  return TRUE;
2711    
2712    
2713  /*************************************************  /*************************************************
2714  *           Check for POSIX class syntax         *  *        Base opcode of repeated opcodes         *
2715  *************************************************/  *************************************************/
2716    
2717  /* This function is called when the sequence "[:" or "[." or "[=" is  /* Returns the base opcode for repeated single character type opcodes. If the
2718  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  
2719    
2720  Returns:   TRUE or FALSE  Arguments:  c opcode
2721    Returns:    base opcode for the type
2722  */  */
2723    
2724  static BOOL  static pcre_uchar
2725  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)  get_repeat_base(pcre_uchar c)
2726  {  {
2727  int terminator;          /* Don't combine these lines; the Solaris cc */  return (c > OP_TYPEPOSUPTO)? c :
2728  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */         (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2729  for (++ptr; *ptr != 0; ptr++)         (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2730    {         (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2731    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)         (c >= OP_STARI)?      OP_STARI :
2732      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;  
2733  }  }
2734    
2735    
2736    
2737    #ifdef SUPPORT_UCP
2738  /*************************************************  /*************************************************
2739  *          Check POSIX class name                *  *        Check a character and a property        *
2740  *************************************************/  *************************************************/
2741    
2742  /* 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
2743  such as [:alnum:].  is adjacent to a fixed character.
2744    
2745  Arguments:  Arguments:
2746    ptr        points to the first letter    c            the character
2747    len        the length of the name    ptype        the property type
2748      pdata        the data for the type
2749      negated      TRUE if it's a negated property (\P or \p{^)
2750    
2751  Returns:     a value representing the name, or -1 if unknown  Returns:       TRUE if auto-possessifying is OK
2752  */  */
2753    
2754  static int  static BOOL
2755  check_posix_name(const pcre_uchar *ptr, int len)  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2756      BOOL negated)
2757  {  {
2758  const char *pn = posix_names;  const pcre_uint32 *p;
2759  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   *  
 *************************************************/  
2760    
2761  /* OP_RECURSE items contain an offset from the start of the regex to the group  switch(ptype)
2762  that is referenced. This means that groups can be replicated for fixed    {
2763  repetition simply by copying (because the recursion is allowed to refer to    case PT_LAMP:
2764  earlier groups that are outside the current group). However, when a group is    return (prop->chartype == ucp_Lu ||
2765  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is            prop->chartype == ucp_Ll ||
2766  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.  
2767    
2768  This function has been extended with the possibility of forward references for    case PT_GC:
2769  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).  
2770    
2771  Arguments:    case PT_PC:
2772    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  
2773    
2774  Returns:     nothing    case PT_SC:
2775  */    return (pdata == prop->script) == negated;
2776    
2777  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;  
2778    
2779  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)    case PT_ALNUM:
2780    {    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2781    int offset;            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
   pcre_uchar *hc;  
2782    
2783    /* 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
2784    reference. */    means that Perl space and POSIX space are now identical. PCRE was changed
2785      at release 8.34. */
2786    
2787    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    case PT_SPACE:    /* Perl space */
2788      case PT_PXSPACE:  /* POSIX space */
2789      switch(c)
2790      {      {
2791      offset = GET(hc, 0);      HSPACE_CASES:
2792      if (cd->start_code + offset == ptr + 1)      VSPACE_CASES:
2793        {      return negated;
2794        PUT(hc, 0, offset + adjust);  
2795        break;      default:
2796        }      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2797      }      }
2798      break;  /* Control never reaches here */
2799    
2800    /* Otherwise, adjust the recursion offset if it's after the start of this    case PT_WORD:
2801    group. */    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2802              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2803              c == CHAR_UNDERSCORE) == negated;
2804    
2805    if (hc >= cd->hwm)    case PT_CLIST:
2806      p = PRIV(ucd_caseless_sets) + prop->caseset;
2807      for (;;)
2808      {      {
2809      offset = GET(ptr, 1);      if (c < *p) return !negated;
2810      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      if (c == *p++) return negated;
2811      }      }
2812      break;  /* Control never reaches here */
   ptr += 1 + LINK_SIZE;  
2813    }    }
2814    
2815    return FALSE;
2816  }  }
2817    #endif  /* SUPPORT_UCP */
2818    
2819    
2820    
2821  /*************************************************  /*************************************************
2822  *        Insert an automatic callout point       *  *        Fill the character property list        *
2823  *************************************************/  *************************************************/
2824    
2825  /* 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-
2826  callout points before each pattern item.  possessification, and if so, fills a list with its properties.
2827    
2828  Arguments:  Arguments:
2829    code           current code pointer    code        points to start of expression
2830    ptr            current pattern pointer    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2831    cd             pointers to tables etc    fcc         points to case-flipping table
2832      list        points to output list
2833                  list[0] will be filled with the opcode
2834                  list[1] will be non-zero if this opcode
2835                    can match an empty character string
2836                  list[2..7] depends on the opcode
2837    
2838  Returns:         new code pointer  Returns:      points to the start of the next opcode if *code is accepted
2839                  NULL if *code is not accepted
2840  */  */
2841    
2842  static pcre_uchar *  static const pcre_uchar *
2843  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)  get_chr_property_list(const pcre_uchar *code, BOOL utf,
2844      const pcre_uint8 *fcc, pcre_uint32 *list)
2845    {
2846    pcre_uchar c = *code;
2847    const pcre_uchar *end;
2848    const pcre_uint32 *clist_src;
2849    pcre_uint32 *clist_dest;
2850    pcre_uint32 chr;
2851    pcre_uchar base;
2852    
2853    list[0] = c;
2854    list[1] = FALSE;
2855    code++;
2856    
2857    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2858      {
2859      base = get_repeat_base(c);
2860      c -= (base - OP_STAR);
2861    
2862      if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2863        code += IMM2_SIZE;
2864    
2865      list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2866    
2867      switch(base)
2868        {
2869        case OP_STAR:
2870        list[0] = OP_CHAR;
2871        break;
2872    
2873        case OP_STARI:
2874        list[0] = OP_CHARI;
2875        break;
2876    
2877        case OP_NOTSTAR:
2878        list[0] = OP_NOT;
2879        break;
2880    
2881        case OP_NOTSTARI:
2882        list[0] = OP_NOTI;
2883        break;
2884    
2885        case OP_TYPESTAR:
2886        list[0] = *code;
2887        code++;
2888        break;
2889        }
2890      c = list[0];
2891      }
2892    
2893    switch(c)
2894      {
2895      case OP_NOT_DIGIT:
2896      case OP_DIGIT:
2897      case OP_NOT_WHITESPACE:
2898      case OP_WHITESPACE:
2899      case OP_NOT_WORDCHAR:
2900      case OP_WORDCHAR:
2901      case OP_ANY:
2902      case OP_ALLANY:
2903      case OP_ANYNL:
2904      case OP_NOT_HSPACE:
2905      case OP_HSPACE:
2906      case OP_NOT_VSPACE:
2907      case OP_VSPACE:
2908      case OP_EXTUNI:
2909      case OP_EODN:
2910      case OP_EOD:
2911      case OP_DOLL:
2912      case OP_DOLLM:
2913      return code;
2914    
2915      case OP_CHAR:
2916      case OP_NOT:
2917      GETCHARINCTEST(chr, code);
2918      list[2] = chr;
2919      list[3] = NOTACHAR;
2920      return code;
2921    
2922      case OP_CHARI:
2923      case OP_NOTI:
2924      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2925      GETCHARINCTEST(chr, code);
2926      list[2] = chr;
2927    
2928    #ifdef SUPPORT_UCP
2929      if (chr < 128 || (chr < 256 && !utf))
2930        list[3] = fcc[chr];
2931      else
2932        list[3] = UCD_OTHERCASE(chr);
2933    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2934      list[3] = (chr < 256) ? fcc[chr] : chr;
2935    #else
2936      list[3] = fcc[chr];
2937    #endif
2938    
2939      /* The othercase might be the same value. */
2940    
2941      if (chr == list[3])
2942        list[3] = NOTACHAR;
2943      else
2944        list[4] = NOTACHAR;
2945      return code;
2946    
2947    #ifdef SUPPORT_UCP
2948      case OP_PROP:
2949      case OP_NOTPROP:
2950      if (code[0] != PT_CLIST)
2951        {
2952        list[2] = code[0];
2953        list[3] = code[1];
2954        return code + 2;
2955        }
2956    
2957      /* Convert only if we have enough space. */
2958    
2959      clist_src = PRIV(ucd_caseless_sets) + code[1];
2960      clist_dest = list + 2;
2961      code += 2;
2962    
2963      do {
2964         if (clist_dest >= list + 8)
2965           {
2966           /* Early return if there is not enough space. This should never
2967           happen, since all clists are shorter than 5 character now. */
2968           list[2] = code[0];
2969           list[3] = code[1];
2970           return code;
2971           }
2972         *clist_dest++ = *clist_src;
2973         }
2974      while(*clist_src++ != NOTACHAR);
2975    
2976      /* All characters are stored. The terminating NOTACHAR
2977      is copied form the clist itself. */
2978    
2979      list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
2980      return code;
2981    #endif
2982    
2983      case OP_NCLASS:
2984      case OP_CLASS:
2985    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2986      case OP_XCLASS:
2987      if (c == OP_XCLASS)
2988        end = code + GET(code, 0) - 1;
2989      else
2990    #endif
2991        end = code + 32 / sizeof(pcre_uchar);
2992    
2993      switch(*end)
2994        {
2995        case OP_CRSTAR:
2996        case OP_CRMINSTAR:
2997        case OP_CRQUERY:
2998        case OP_CRMINQUERY:
2999        case OP_CRPOSSTAR:
3000        case OP_CRPOSQUERY:
3001        list[1] = TRUE;
3002        end++;
3003        break;
3004    
3005        case OP_CRPLUS:
3006        case OP_CRMINPLUS:
3007        case OP_CRPOSPLUS:
3008        end++;
3009        break;
3010    
3011        case OP_CRRANGE:
3012        case OP_CRMINRANGE:
3013        case OP_CRPOSRANGE:
3014        list[1] = (GET2(end, 1) == 0);
3015        end += 1 + 2 * IMM2_SIZE;
3016        break;
3017        }
3018      list[2] = end - code;
3019      return end;
3020      }
3021    return NULL;    /* Opcode not accepted */
3022    }
3023    
3024    
3025    
3026    /*************************************************
3027    *    Scan further character sets for match       *
3028    *************************************************/
3029    
3030    /* Checks whether the base and the current opcode have a common character, in
3031    which case the base cannot be possessified.
3032    
3033    Arguments:
3034      code        points to the byte code
3035      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3036      cd          static compile data
3037      base_list   the data list of the base opcode
3038    
3039    Returns:      TRUE if the auto-possessification is possible
3040    */
3041    
3042    static BOOL
3043    compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3044      const pcre_uint32 *base_list, const pcre_uchar *base_end)
3045    {
3046    pcre_uchar c;
3047    pcre_uint32 list[8];
3048    const pcre_uint32 *chr_ptr;
3049    const pcre_uint32 *ochr_ptr;
3050    const pcre_uint32 *list_ptr;
3051    const pcre_uchar *next_code;
3052    const pcre_uint8 *class_bitset;
3053    const pcre_uint32 *set1, *set2, *set_end;
3054    pcre_uint32 chr;
3055    BOOL accepted, invert_bits;
3056    
3057    /* Note: the base_list[1] contains whether the current opcode has greedy
3058    (represented by a non-zero value) quantifier. This is a different from
3059    other character type lists, which stores here that the character iterator
3060    matches to an empty string (also represented by a non-zero value). */
3061    
3062    for(;;)
3063      {
3064      /* All operations move the code pointer forward.
3065      Therefore infinite recursions are not possible. */
3066    
3067      c = *code;
3068    
3069      /* Skip over callouts */
3070    
3071      if (c == OP_CALLOUT)
3072        {
3073        code += PRIV(OP_lengths)[c];
3074        continue;
3075        }
3076    
3077      if (c == OP_ALT)
3078        {
3079        do code += GET(code, 1); while (*code == OP_ALT);
3080        c = *code;
3081        }
3082    
3083      switch(c)
3084        {
3085        case OP_END:
3086        case OP_KETRPOS:
3087        /* TRUE only in greedy case. The non-greedy case could be replaced by
3088        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3089        uses more memory, which we cannot get at this stage.) */
3090    
3091        return base_list[1] != 0;
3092    
3093        case OP_KET:
3094        /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3095        it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3096        cannot be converted to a possessive form. */
3097    
3098        if (base_list[1] == 0) return FALSE;
3099    
3100        switch(*(code - GET(code, 1)))
3101          {
3102          case OP_ASSERT:
3103          case OP_ASSERT_NOT:
3104          case OP_ASSERTBACK:
3105          case OP_ASSERTBACK_NOT:
3106          case OP_ONCE:
3107          case OP_ONCE_NC:
3108          /* Atomic sub-patterns and assertions can always auto-possessify their
3109          last iterator. */
3110          return TRUE;
3111          }
3112    
3113        code += PRIV(OP_lengths)[c];
3114        continue;
3115    
3116        case OP_ONCE:
3117        case OP_ONCE_NC:
3118        case OP_BRA:
3119        case OP_CBRA:
3120        next_code = code + GET(code, 1);
3121        code += PRIV(OP_lengths)[c];
3122    
3123        while (*next_code == OP_ALT)
3124          {
3125          if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3126          code = next_code + 1 + LINK_SIZE;
3127          next_code += GET(next_code, 1);
3128          }
3129        continue;
3130    
3131        case OP_BRAZERO:
3132        case OP_BRAMINZERO:
3133    
3134        next_code = code + 1;
3135        if (*next_code != OP_BRA && *next_code != OP_CBRA
3136            && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3137    
3138        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3139    
3140        /* The bracket content will be checked by the
3141        OP_BRA/OP_CBRA case above. */
3142        next_code += 1 + LINK_SIZE;
3143        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3144          return FALSE;
3145    
3146        code += PRIV(OP_lengths)[c];
3147        continue;
3148        }
3149    
3150      /* Check for a supported opcode, and load its properties. */
3151    
3152      code = get_chr_property_list(code, utf, cd->fcc, list);
3153      if (code == NULL) return FALSE;    /* Unsupported */
3154    
3155      /* If either opcode is a small character list, set pointers for comparing
3156      characters from that list with another list, or with a property. */
3157    
3158      if (base_list[0] == OP_CHAR)
3159        {
3160        chr_ptr = base_list + 2;
3161        list_ptr = list;
3162        }
3163      else if (list[0] == OP_CHAR)
3164        {
3165        chr_ptr = list + 2;
3166        list_ptr = base_list;
3167        }
3168    
3169      /* Character bitsets can also be compared to certain opcodes. */
3170    
3171      else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3172    #ifdef COMPILE_PCRE8
3173          /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3174          || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3175    #endif
3176          )
3177        {
3178    #ifdef COMPILE_PCRE8
3179        if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3180    #else
3181        if (base_list[0] == OP_CLASS)
3182    #endif
3183          {
3184          set1 = (pcre_uint32 *)(base_end - base_list[2]);
3185          list_ptr = list;
3186          }
3187        else
3188          {
3189          set1 = (pcre_uint32 *)(code - list[2]);
3190          list_ptr = base_list;
3191          }
3192    
3193        invert_bits = FALSE;
3194        switch(list_ptr[0])
3195          {
3196          case OP_CLASS:
3197          case OP_NCLASS:
3198          set2 = (pcre_uint32 *)
3199            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3200          break;
3201    
3202          /* OP_XCLASS cannot be supported here, because its bitset
3203          is not necessarily complete. E.g: [a-\0x{200}] is stored
3204          as a character range, and the appropriate bits are not set. */
3205    
3206          case OP_NOT_DIGIT:
3207            invert_bits = TRUE;
3208            /* Fall through */
3209          case OP_DIGIT:
3210            set2 = (pcre_uint32 *)(cd->cbits + cbit_digit);
3211            break;
3212    
3213          case OP_NOT_WHITESPACE:
3214            invert_bits = TRUE;
3215            /* Fall through */
3216          case OP_WHITESPACE:
3217            set2 = (pcre_uint32 *)(cd->cbits + cbit_space);
3218            break;
3219    
3220          case OP_NOT_WORDCHAR:
3221            invert_bits = TRUE;
3222            /* Fall through */
3223          case OP_WORDCHAR:
3224            set2 = (pcre_uint32 *)(cd->cbits + cbit_word);
3225            break;
3226    
3227          default:
3228          return FALSE;
3229          }
3230    
3231        /* Compare 4 bytes to improve speed. */
3232        set_end = set1 + (32 / 4);
3233        if (invert_bits)
3234          {
3235          do
3236            {
3237            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3238            }
3239          while (set1 < set_end);
3240          }
3241        else
3242          {
3243          do
3244            {
3245            if ((*set1++ & *set2++) != 0) return FALSE;
3246            }
3247          while (set1 < set_end);
3248          }
3249    
3250        if (list[1] == 0) return TRUE;
3251        /* Might be an empty repeat. */
3252        continue;
3253        }
3254    
3255      /* Some property combinations also acceptable. Unicode property opcodes are
3256      processed specially; the rest can be handled with a lookup table. */
3257    
3258      else
3259        {
3260        pcre_uint32 leftop, rightop;
3261    
3262        leftop = base_list[0];
3263        rightop = list[0];
3264    
3265    #ifdef SUPPORT_UCP
3266        accepted = FALSE; /* Always set in non-unicode case. */
3267        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3268          {
3269          if (rightop == OP_EOD)
3270            accepted = TRUE;
3271          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3272            {
3273            int n;
3274            const pcre_uint8 *p;
3275            BOOL same = leftop == rightop;
3276            BOOL lisprop = leftop == OP_PROP;
3277            BOOL risprop = rightop == OP_PROP;
3278            BOOL bothprop = lisprop && risprop;
3279    
3280            /* There's a table that specifies how each combination is to be
3281            processed:
3282              0   Always return FALSE (never auto-possessify)
3283              1   Character groups are distinct (possessify if both are OP_PROP)
3284              2   Check character categories in the same group (general or particular)
3285              3   Return TRUE if the two opcodes are not the same
3286              ... see comments below
3287            */
3288    
3289            n = propposstab[base_list[2]][list[2]];
3290            switch(n)
3291              {
3292              case 0: break;
3293              case 1: accepted = bothprop; break;
3294              case 2: accepted = (base_list[3] == list[3]) != same; break;
3295              case 3: accepted = !same; break;
3296    
3297              case 4:  /* Left general category, right particular category */
3298              accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3299              break;
3300    
3301              case 5:  /* Right general category, left particular category */
3302              accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3303              break;
3304    
3305              /* This code is logically tricky. Think hard before fiddling with it.
3306              The posspropstab table has four entries per row. Each row relates to
3307              one of PCRE's special properties such as ALNUM or SPACE or WORD.
3308              Only WORD actually needs all four entries, but using repeats for the
3309              others means they can all use the same code below.
3310    
3311              The first two entries in each row are Unicode general categories, and
3312              apply always, because all the characters they include are part of the
3313              PCRE character set. The third and fourth entries are a general and a
3314              particular category, respectively, that include one or more relevant
3315              characters. One or the other is used, depending on whether the check
3316              is for a general or a particular category. However, in both cases the
3317              category contains more characters than the specials that are defined
3318              for the property being tested against. Therefore, it cannot be used
3319              in a NOTPROP case.
3320    
3321              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3322              Underscore is covered by ucp_P or ucp_Po. */
3323    
3324              case 6:  /* Left alphanum vs right general category */
3325              case 7:  /* Left space vs right general category */
3326              case 8:  /* Left word vs right general category */
3327              p = posspropstab[n-6];
3328              accepted = risprop && lisprop ==
3329                (list[3] != p[0] &&
3330                 list[3] != p[1] &&
3331                (list[3] != p[2] || !lisprop));
3332              break;
3333    
3334              case 9:   /* Right alphanum vs left general category */
3335              case 10:  /* Right space vs left general category */
3336              case 11:  /* Right word vs left general category */
3337              p = posspropstab[n-9];
3338              accepted = lisprop && risprop ==
3339                (base_list[3] != p[0] &&
3340                 base_list[3] != p[1] &&
3341                (base_list[3] != p[2] || !risprop));
3342              break;
3343    
3344              case 12:  /* Left alphanum vs right particular category */
3345              case 13:  /* Left space vs right particular category */
3346              case 14:  /* Left word vs right particular category */
3347              p = posspropstab[n-12];
3348              accepted = risprop && lisprop ==
3349                (catposstab[p[0]][list[3]] &&
3350                 catposstab[p[1]][list[3]] &&
3351                (list[3] != p[3] || !lisprop));
3352              break;
3353    
3354              case 15:  /* Right alphanum vs left particular category */
3355              case 16:  /* Right space vs left particular category */
3356              case 17:  /* Right word vs left particular category */
3357              p = posspropstab[n-15];
3358              accepted = lisprop && risprop ==
3359                (catposstab[p[0]][base_list[3]] &&
3360                 catposstab[p[1]][base_list[3]] &&
3361                (base_list[3] != p[3] || !risprop));
3362              break;
3363              }
3364            }
3365          }
3366    
3367        else
3368    #endif  /* SUPPORT_UCP */
3369    
3370        accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3371               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3372               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3373    
3374        if (!accepted)
3375          return FALSE;
3376    
3377        if (list[1] == 0) return TRUE;
3378        /* Might be an empty repeat. */
3379        continue;
3380        }
3381    
3382      /* Control reaches here only if one of the items is a small character list.
3383      All characters are checked against the other side. */
3384    
3385      do
3386        {
3387        chr = *chr_ptr;
3388    
3389        switch(list_ptr[0])
3390          {
3391          case OP_CHAR:
3392          ochr_ptr = list_ptr + 2;
3393          do
3394            {
3395            if (chr == *ochr_ptr) return FALSE;
3396            ochr_ptr++;
3397            }
3398          while(*ochr_ptr != NOTACHAR);
3399          break;
3400    
3401          case OP_NOT:
3402          ochr_ptr = list_ptr + 2;
3403          do
3404            {
3405            if (chr == *ochr_ptr)
3406              break;
3407            ochr_ptr++;
3408            }
3409          while(*ochr_ptr != NOTACHAR);
3410          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3411          break;
3412    
3413          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3414          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3415    
3416          case OP_DIGIT:
3417          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3418          break;
3419    
3420          case OP_NOT_DIGIT:
3421          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3422          break;
3423    
3424          case OP_WHITESPACE:
3425          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3426          break;
3427    
3428          case OP_NOT_WHITESPACE:
3429          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3430          break;
3431    
3432          case OP_WORDCHAR:
3433          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3434          break;
3435    
3436          case OP_NOT_WORDCHAR:
3437          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3438          break;
3439    
3440          case OP_HSPACE:
3441          switch(chr)
3442            {
3443            HSPACE_CASES: return FALSE;
3444            default: break;
3445            }
3446          break;
3447    
3448          case OP_NOT_HSPACE:
3449          switch(chr)
3450            {
3451            HSPACE_CASES: break;
3452            default: return FALSE;
3453            }
3454          break;
3455    
3456          case OP_ANYNL:
3457          case OP_VSPACE:
3458          switch(chr)
3459            {
3460            VSPACE_CASES: return FALSE;
3461            default: break;
3462            }
3463          break;
3464    
3465          case OP_NOT_VSPACE:
3466          switch(chr)
3467            {
3468            VSPACE_CASES: break;
3469            default: return FALSE;
3470            }
3471          break;
3472    
3473          case OP_DOLL:
3474          case OP_EODN:
3475          switch (chr)
3476            {
3477            case CHAR_CR:
3478            case CHAR_LF:
3479            case CHAR_VT:
3480            case CHAR_FF:
3481            case CHAR_NEL:
3482    #ifndef EBCDIC
3483            case 0x2028:
3484            case 0x2029:
3485    #endif  /* Not EBCDIC */
3486            return FALSE;
3487            }
3488          break;
3489    
3490          case OP_EOD:    /* Can always possessify before \z */
3491          break;
3492    
3493    #ifdef SUPPORT_UCP
3494          case OP_PROP:
3495          case OP_NOTPROP:
3496          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3497                list_ptr[0] == OP_NOTPROP))
3498            return FALSE;
3499          break;
3500    #endif
3501    
3502          case OP_NCLASS:
3503          if (chr > 255) return FALSE;
3504          /* Fall through */
3505    
3506          case OP_CLASS:
3507          if (chr > 255) break;
3508          class_bitset = (pcre_uint8 *)
3509            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3510          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3511          break;
3512    
3513    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3514          case OP_XCLASS:
3515          if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3516              list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3517          break;
3518    #endif
3519    
3520          default:
3521          return FALSE;
3522          }
3523    
3524        chr_ptr++;
3525        }
3526      while(*chr_ptr != NOTACHAR);
3527    
3528      /* At least one character must be matched from this opcode. */
3529    
3530      if (list[1] == 0) return TRUE;
3531      }
3532    
3533    return FALSE;
3534    }
3535    
3536    
3537    
3538    /*************************************************
3539    *    Scan compiled regex for auto-possession     *
3540    *************************************************/
3541    
3542    /* Replaces single character iterations with their possessive alternatives
3543    if appropriate. This function modifies the compiled opcode!
3544    
3545    Arguments:
3546      code        points to start of the byte code
3547      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3548      cd          static compile data
3549    
3550    Returns:      nothing
3551    */
3552    
3553    static void
3554    auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3555    {
3556    register pcre_uchar c;
3557    const pcre_uchar *end;
3558    pcre_uchar *repeat_opcode;
3559    pcre_uint32 list[8];
3560    
3561    for (;;)
3562      {
3563      c = *code;
3564    
3565      if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3566        {
3567        c -= get_repeat_base(c) - OP_STAR;
3568        end = (c <= OP_MINUPTO) ?
3569          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3570        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3571    
3572        if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3573          {
3574          switch(c)
3575            {
3576            case OP_STAR:
3577            *code += OP_POSSTAR - OP_STAR;
3578            break;
3579    
3580            case OP_MINSTAR:
3581            *code += OP_POSSTAR - OP_MINSTAR;
3582            break;
3583    
3584            case OP_PLUS:
3585            *code += OP_POSPLUS - OP_PLUS;
3586            break;
3587    
3588            case OP_MINPLUS:
3589            *code += OP_POSPLUS - OP_MINPLUS;
3590            break;
3591    
3592            case OP_QUERY:
3593            *code += OP_POSQUERY - OP_QUERY;
3594            break;
3595    
3596            case OP_MINQUERY:
3597            *code += OP_POSQUERY - OP_MINQUERY;
3598            break;
3599    
3600            case OP_UPTO:
3601            *code += OP_POSUPTO - OP_UPTO;
3602            break;
3603    
3604            case OP_MINUPTO:
3605            *code += OP_MINUPTO - OP_UPTO;
3606            break;
3607            }
3608          }
3609        c = *code;
3610        }
3611      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3612        {
3613    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3614        if (c == OP_XCLASS)
3615          repeat_opcode = code + GET(code, 1);
3616        else
3617    #endif
3618          repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3619    
3620        c = *repeat_opcode;
3621        if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3622          {
3623          /* end must not be NULL. */
3624          end = get_chr_property_list(code, utf, cd->fcc, list);
3625    
3626          list[1] = (c & 1) == 0;
3627    
3628          if (compare_opcodes(end, utf, cd, list, end))
3629            {
3630            switch (c)
3631              {
3632              case OP_CRSTAR:
3633              case OP_CRMINSTAR:
3634              *repeat_opcode = OP_CRPOSSTAR;
3635              break;
3636    
3637              case OP_CRPLUS:
3638              case OP_CRMINPLUS:
3639              *repeat_opcode = OP_CRPOSPLUS;
3640              break;
3641    
3642              case OP_CRQUERY:
3643              case OP_CRMINQUERY:
3644              *repeat_opcode = OP_CRPOSQUERY;
3645              break;
3646    
3647              case OP_CRRANGE:
3648              case OP_CRMINRANGE:
3649              *repeat_opcode = OP_CRPOSRANGE;
3650              break;
3651              }
3652            }
3653          }
3654        c = *code;
3655        }
3656    
3657      switch(c)
3658        {
3659        case OP_END:
3660        return;
3661    
3662        case OP_TYPESTAR:
3663        case OP_TYPEMINSTAR:
3664        case OP_TYPEPLUS:
3665        case OP_TYPEMINPLUS:
3666        case OP_TYPEQUERY:
3667        case OP_TYPEMINQUERY:
3668        case OP_TYPEPOSSTAR:
3669        case OP_TYPEPOSPLUS:
3670        case OP_TYPEPOSQUERY:
3671        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3672        break;
3673    
3674        case OP_TYPEUPTO:
3675        case OP_TYPEMINUPTO:
3676        case OP_TYPEEXACT:
3677        case OP_TYPEPOSUPTO:
3678        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3679          code += 2;
3680        break;
3681    
3682    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3683        case OP_XCLASS:
3684        code += GET(code, 1);
3685        break;
3686    #endif
3687    
3688        case OP_MARK:
3689        case OP_PRUNE_ARG:
3690        case OP_SKIP_ARG:
3691        case OP_THEN_ARG:
3692        code += code[1];
3693        break;
3694        }
3695    
3696      /* Add in the fixed length from the table */
3697    
3698      code += PRIV(OP_lengths)[c];
3699    
3700      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3701      a multi-byte character. The length in the table is a minimum, so we have to
3702      arrange to skip the extra bytes. */
3703    
3704    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3705      if (utf) switch(c)
3706        {
3707        case OP_CHAR:
3708        case OP_CHARI:
3709        case OP_NOT:
3710        case OP_NOTI:
3711        case OP_STAR:
3712        case OP_MINSTAR:
3713        case OP_PLUS:
3714        case OP_MINPLUS:
3715        case OP_QUERY:
3716        case OP_MINQUERY:
3717        case OP_UPTO:
3718        case OP_MINUPTO:
3719        case OP_EXACT:
3720        case OP_POSSTAR:
3721        case OP_POSPLUS:
3722        case OP_POSQUERY:
3723        case OP_POSUPTO:
3724        case OP_STARI:
3725        case OP_MINSTARI:
3726        case OP_PLUSI:
3727        case OP_MINPLUSI:
3728        case OP_QUERYI:
3729        case OP_MINQUERYI:
3730        case OP_UPTOI:
3731        case OP_MINUPTOI:
3732        case OP_EXACTI:
3733        case OP_POSSTARI:
3734        case OP_POSPLUSI:
3735        case OP_POSQUERYI:
3736        case OP_POSUPTOI:
3737        case OP_NOTSTAR:
3738        case OP_NOTMINSTAR:
3739        case OP_NOTPLUS:
3740        case OP_NOTMINPLUS:
3741        case OP_NOTQUERY:
3742        case OP_NOTMINQUERY:
3743        case OP_NOTUPTO:
3744        case OP_NOTMINUPTO:
3745        case OP_NOTEXACT:
3746        case OP_NOTPOSSTAR:
3747        case OP_NOTPOSPLUS:
3748        case OP_NOTPOSQUERY:
3749        case OP_NOTPOSUPTO:
3750        case OP_NOTSTARI:
3751        case OP_NOTMINSTARI:
3752        case OP_NOTPLUSI:
3753        case OP_NOTMINPLUSI:
3754        case OP_NOTQUERYI:
3755        case OP_NOTMINQUERYI:
3756        case OP_NOTUPTOI:
3757        case OP_NOTMINUPTOI:
3758        case OP_NOTEXACTI:
3759        case OP_NOTPOSSTARI:
3760        case OP_NOTPOSPLUSI:
3761        case OP_NOTPOSQUERYI:
3762        case OP_NOTPOSUPTOI:
3763        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3764        break;
3765        }
3766    #else
3767      (void)(utf);  /* Keep compiler happy by referencing function argument */
3768    #endif
3769      }
3770    }
3771    
3772    
3773    
3774    /*************************************************
3775    *           Check for POSIX class syntax         *
3776    *************************************************/
3777    
3778    /* This function is called when the sequence "[:" or "[." or "[=" is
3779    encountered in a character class. It checks whether this is followed by a
3780    sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3781    reach an unescaped ']' without the special preceding character, return FALSE.
3782    
3783    Originally, this function only recognized a sequence of letters between the
3784    terminators, but it seems that Perl recognizes any sequence of characters,
3785    though of course unknown POSIX names are subsequently rejected. Perl gives an
3786    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3787    didn't consider this to be a POSIX class. Likewise for [:1234:].
3788    
3789    The problem in trying to be exactly like Perl is in the handling of escapes. We
3790    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3791    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3792    below handles the special case of \], but does not try to do any other escape
3793    processing. This makes it different from Perl for cases such as [:l\ower:]
3794    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3795    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
3796    I think.
3797    
3798    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3799    It seems that the appearance of a nested POSIX class supersedes an apparent
3800    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3801    a digit.
3802    
3803    In Perl, unescaped square brackets may also appear as part of class names. For
3804    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3805    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3806    seem right at all. PCRE does not allow closing square brackets in POSIX class
3807    names.
3808    
3809    Arguments:
3810      ptr      pointer to the initial [
3811      endptr   where to return the end pointer
3812    
3813    Returns:   TRUE or FALSE
3814    */
3815    
3816    static BOOL
3817    check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3818    {
3819    pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
3820    terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
3821    for (++ptr; *ptr != CHAR_NULL; ptr++)
3822      {
3823      if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3824        ptr++;
3825      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3826      else
3827        {
3828        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3829          {
3830          *endptr = ptr;
3831          return TRUE;
3832          }
3833        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3834             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3835              ptr[1] == CHAR_EQUALS_SIGN) &&
3836            check_posix_syntax(ptr, endptr))
3837          return FALSE;
3838        }
3839      }
3840    return FALSE;
3841    }
3842    
3843    
3844    
3845    
3846    /*************************************************
3847    *          Check POSIX class name                *
3848    *************************************************/
3849    
3850    /* This function is called to check the name given in a POSIX-style class entry
3851    such as [:alnum:].
3852    
3853    Arguments:
3854      ptr        points to the first letter
3855      len        the length of the name
3856    
3857    Returns:     a value representing the name, or -1 if unknown
3858    */
3859    
3860    static int
3861    check_posix_name(const pcre_uchar *ptr, int len)
3862    {
3863    const char *pn = posix_names;
3864    register int yield = 0;
3865    while (posix_name_lengths[yield] != 0)
3866      {
3867      if (len == posix_name_lengths[yield] &&
3868        STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3869      pn += posix_name_lengths[yield] + 1;
3870      yield++;
3871      }
3872    return -1;
3873    }
3874    
3875    
3876    /*************************************************
3877    *    Adjust OP_RECURSE items in repeated group   *
3878    *************************************************/
3879    
3880    /* OP_RECURSE items contain an offset from the start of the regex to the group
3881    that is referenced. This means that groups can be replicated for fixed
3882    repetition simply by copying (because the recursion is allowed to refer to
3883    earlier groups that are outside the current group). However, when a group is
3884    optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3885    inserted before it, after it has been compiled. This means that any OP_RECURSE
3886    items within it that refer to the group itself or any contained groups have to
3887    have their offsets adjusted. That one of the jobs of this function. Before it
3888    is called, the partially compiled regex must be temporarily terminated with
3889    OP_END.
3890    
3891    This function has been extended with the possibility of forward references for
3892    recursions and subroutine calls. It must also check the list of such references
3893    for the group we are dealing with. If it finds that one of the recursions in
3894    the current group is on this list, it adjusts the offset in the list, not the
3895    value in the reference (which is a group number).
3896    
3897    Arguments:
3898      group      points to the start of the group
3899      adjust     the amount by which the group is to be moved
3900      utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
3901      cd         contains pointers to tables etc.
3902      save_hwm   the hwm forward reference pointer at the start of the group
3903    
3904    Returns:     nothing
3905    */
3906    
3907    static void
3908    adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3909      pcre_uchar *save_hwm)
3910    {
3911    pcre_uchar *ptr = group;
3912    
3913    while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3914      {
3915      int offset;
3916      pcre_uchar *hc;
3917    
3918      /* See if this recursion is on the forward reference list. If so, adjust the
3919      reference. */
3920    
3921      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3922        {
3923        offset = (int)GET(hc, 0);
3924        if (cd->start_code + offset == ptr + 1)
3925          {
3926          PUT(hc, 0, offset + adjust);
3927          break;
3928          }
3929        }
3930    
3931      /* Otherwise, adjust the recursion offset if it's after the start of this
3932      group. */
3933    
3934      if (hc >= cd->hwm)
3935        {
3936        offset = (int)GET(ptr, 1);
3937        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3938        }
3939    
3940      ptr += 1 + LINK_SIZE;
3941      }
3942    }
3943    
3944    
3945    
3946    /*************************************************
3947    *        Insert an automatic callout point       *
3948    *************************************************/
3949    
3950    /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
3951    callout points before each pattern item.
3952    
3953    Arguments:
3954      code           current code pointer
3955      ptr            current pattern pointer
3956      cd             pointers to tables etc
3957    
3958    Returns:         new code pointer
3959    */
3960    
3961    static pcre_uchar *
3962    auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
3963  {  {
3964  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
3965  *code++ = 255;  *code++ = 255;
# Line 2858  PUT(previous_callout, 2 + LINK_SIZE, len Line 4001  PUT(previous_callout, 2 + LINK_SIZE, len
4001  *************************************************/  *************************************************/
4002    
4003  /* This function is passed the start and end of a class range, in UTF-8 mode  /* This function is passed the start and end of a class range, in UTF-8 mode
4004  with UCP support. It searches up the characters, looking for internal ranges of  with UCP support. It searches up the characters, looking for ranges of
4005  characters in the "other" case. Each call returns the next one, updating the  characters in the "other" case. Each call returns the next one, updating the
4006  start address.  start address. A character with multiple other cases is returned on its own
4007    with a special return value.
4008    
4009  Arguments:  Arguments:
4010    cptr        points to starting character value; updated    cptr        points to starting character value; updated
# Line 2868  Arguments: Line 4012  Arguments:
4012    ocptr       where to put start of othercase range    ocptr       where to put start of othercase range
4013    odptr       where to put end of othercase range    odptr       where to put end of othercase range
4014    
4015  Yield:        TRUE when range returned; FALSE when no more  Yield:        -1 when no more
4016                   0 when a range is returned
4017                  >0 the CASESET offset for char with multiple other cases
4018                    in this case, ocptr contains the original
4019  */  */
4020    
4021  static BOOL  static int
4022  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4023    unsigned int *odptr)    pcre_uint32 *odptr)
4024  {  {
4025  unsigned int c, othercase, next;  pcre_uint32 c, othercase, next;
4026    unsigned int co;
4027    
4028    /* Find the first character that has an other case. If it has multiple other
4029    cases, return its case offset value. */
4030    
4031  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
4032    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }    {
4033      if ((co = UCD_CASESET(c)) != 0)
4034        {
4035        *ocptr = c++;   /* Character that has the set */
4036        *cptr = c;      /* Rest of input range */
4037        return (int)co;
4038        }
4039      if ((othercase = UCD_OTHERCASE(c)) != c) break;
4040      }
4041    
4042  if (c > d) return FALSE;  if (c > d) return -1;  /* Reached end of range */
4043    
4044  *ocptr = othercase;  *ocptr = othercase;
4045  next = othercase + 1;  next = othercase + 1;
# Line 2891  for (++c; c <= d; c++) Line 4050  for (++c; c <= d; c++)
4050    next++;    next++;
4051    }    }
4052    
4053  *odptr = next - 1;  *odptr = next - 1;     /* End of othercase range */
4054  *cptr = c;  *cptr = c;             /* Rest of input range */
4055    return 0;
 return TRUE;  
 }  
   
   
   
 /*************************************************  
 *        Check a character and a property        *  
 *************************************************/  
   
 /* This function is called by check_auto_possessive() when a property item  
 is adjacent to a fixed character.  
   
 Arguments:  
   c            the character  
   ptype        the property type  
   pdata        the data for the type  
   negated      TRUE if it's a negated property (\P or \p{^)  
   
 Returns:       TRUE if auto-possessifying is OK  
 */  
   
 static BOOL  
 check_char_prop(int c, int ptype, int pdata, BOOL negated)  
 {  
 const ucd_record *prop = GET_UCD(c);  
 switch(ptype)  
   {  
   case PT_LAMP:  
   return (prop->chartype == ucp_Lu ||  
           prop->chartype == ucp_Ll ||  
           prop->chartype == ucp_Lt) == negated;  
   
   case PT_GC:  
   return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;  
   
   case PT_PC:  
   return (pdata == prop->chartype) == negated;  
   
   case PT_SC:  
   return (pdata == prop->script) == negated;  
   
   /* These are specials */  
   
   case PT_ALNUM:  
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||  
           PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;  
   
   case PT_SPACE:    /* Perl space */  
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||  
           c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)  
           == negated;  
   
   case PT_PXSPACE:  /* POSIX space */  
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||  
           c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||  
           c == CHAR_FF || c == CHAR_CR)  
           == negated;  
   
   case PT_WORD:  
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||  
           PRIV(ucp_gentype)[prop->chartype] == ucp_N ||  
           c == CHAR_UNDERSCORE) == negated;  
   }  
 return FALSE;  
4056  }  }
4057  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
4058    
4059    
4060    
4061  /*************************************************  /*************************************************
4062  *     Check if auto-possessifying is possible    *  *        Add a character or range to a class     *
4063  *************************************************/  *************************************************/
4064    
4065  /* This function is called for unlimited repeats of certain items, to see  /* This function packages up the logic of adding a character or range of
4066  whether the next thing could possibly match the repeated item. If not, it makes  characters to a class. The character values in the arguments will be within the
4067  sense to automatically possessify the repeated item.  valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
4068    mutually recursive with the function immediately below.
4069    
4070  Arguments:  Arguments:
4071    previous      pointer to the repeated opcode    classbits     the bit map for characters < 256
4072    utf           TRUE in UTF-8 / UTF-16 mode    uchardptr     points to the pointer for extra data
4073    ptr           next character in pattern    options       the options word
   options       options bits  
4074    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4075      start         start of range character
4076      end           end of range character
4077    
4078  Returns:        TRUE if possessifying is wanted  Returns:        the number of < 256 characters added
4079                    the pointer to extra data is updated
4080  */  */
4081    
4082  static BOOL  static int
4083  check_auto_possessive(const pcre_uchar *previous, BOOL utf,  add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4084    const pcre_uchar *ptr, int options, compile_data *cd)    compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4085  {  {
4086  pcre_int32 c, next;  pcre_uint32 c;
4087  int op_code = *previous++;  int n8 = 0;
4088    
4089  /* Skip whitespace and comments in extended mode */  /* If caseless matching is required, scan the range and process alternate
4090    cases. In Unicode, there are 8-bit characters that have alternate cases that
4091    are greater than 255 and vice-versa. Sometimes we can just extend the original
4092    range. */
4093    
4094  if ((options & PCRE_EXTENDED) != 0)  if ((options & PCRE_CASELESS) != 0)
4095    {    {
4096    for (;;)  #ifdef SUPPORT_UCP
4097      if ((options & PCRE_UTF8) != 0)
4098      {      {
4099      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      int rc;
4100      if (*ptr == CHAR_NUMBER_SIGN)      pcre_uint32 oc, od;
       {  
       ptr++;  
       while (*ptr != 0)  
         {  
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
       }  
     else break;  
     }  
   }  
   
 /* If the next item is one that we can handle, get its value. A non-negative  
 value is a character, a negative value is an escape value. */  
4101    
4102  if (*ptr == CHAR_BACKSLASH)      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
4103    {      c = start;
   int temperrorcode = 0;  
   next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);  
   if (temperrorcode != 0) return FALSE;  
   ptr++;    /* Point after the escape sequence */  
   }  
 else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)  
   {  
 #ifdef SUPPORT_UTF  
   if (utf) { GETCHARINC(next, ptr); } else  
 #endif  
   next = *ptr++;  
   }  
 else return FALSE;  
4104    
4105  /* Skip whitespace and comments in extended mode */      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
   
 if ((options & PCRE_EXTENDED) != 0)  
   {  
   for (;;)  
     {  
     while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;  
     if (*ptr == CHAR_NUMBER_SIGN)  
4106        {        {
4107        ptr++;        /* Handle a single character that has more than one other case. */
       while (*ptr != 0)  
         {  
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
       }  
     else break;  
     }  
   }  
   
 /* If the next thing is itself optional, we have to give up. */  
   
 if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  
   STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)  
     return FALSE;  
   
 /* Now compare the next item with the previous opcode. First, handle cases when  
 the next item is a character. */  
4108    
4109  if (next >= 0) switch(op_code)        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4110    {          PRIV(ucd_caseless_sets) + rc, oc);
   case OP_CHAR:  
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   return c != next;  
4111    
4112    /* For CHARI (caseless character) we must check the other case. If we have        /* Do nothing if the other case range is within the original range. */
   Unicode property support, we can use it to test the other case of  
   high-valued characters. */  
4113    
4114    case OP_CHARI:        else if (oc >= start && od <= end) continue;
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   if (c == next) return FALSE;  
 #ifdef SUPPORT_UTF  
   if (utf)  
     {  
     unsigned int othercase;  
     if (next < 128) othercase = cd->fcc[next]; else  
 #ifdef SUPPORT_UCP  
     othercase = UCD_OTHERCASE((unsigned int)next);  
 #else  
     othercase = NOTACHAR;  
 #endif  
     return (unsigned int)c != othercase;  
     }  
   else  
 #endif  /* SUPPORT_UTF */  
   return (c != TABLE_GET((unsigned int)next, cd->fcc, next));  /* Non-UTF-8 mode */  
4115    
4116    case OP_NOT:        /* Extend the original range if there is overlap, noting that if oc < c, we
4117  #ifdef SUPPORT_UTF        can't have od > end because a subrange is always shorter than the basic
4118    GETCHARTEST(c, previous);        range. Otherwise, use a recursive call to add the additional range. */
 #else  
   c = *previous;  
 #endif  
   return c == next;  
4119    
4120    case OP_NOTI:        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4121  #ifdef SUPPORT_UTF        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
4122    GETCHARTEST(c, previous);        else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
4123  #else        }
   c = *previous;  
 #endif  
   if (c == next) return TRUE;  
 #ifdef SUPPORT_UTF  
   if (utf)  
     {  
     unsigned int othercase;  
     if (next < 128) othercase = cd->fcc[next]; else  
 #ifdef SUPPORT_UCP  
     othercase = UCD_OTHERCASE((unsigned int)next);  
 #else  
     othercase = NOTACHAR;  
 #endif  
     return (unsigned int)c == othercase;  
4124      }      }
4125    else    else
4126  #endif  /* SUPPORT_UTF */  #endif  /* SUPPORT_UCP */
   return (c == TABLE_GET((unsigned int)next, cd->fcc, next));  /* Non-UTF-8 mode */  
   
   /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.  
   When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */  
   
   case OP_DIGIT:  
   return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;  
   
   case OP_NOT_DIGIT:  
   return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;  
   
   case OP_WHITESPACE:  
   return next > 255 || (cd->ctypes[next] & ctype_space) == 0;  
   
   case OP_NOT_WHITESPACE:  
   return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;  
   
   case OP_WORDCHAR:  
   return next > 255 || (cd->ctypes[next] & ctype_word) == 0;  
   
   case OP_NOT_WORDCHAR:  
   return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;  
4127    
4128    case OP_HSPACE:    /* Not UTF-mode, or no UCP */
   case OP_NOT_HSPACE:  
   switch(next)  
     {  
     case 0x09:  
     case 0x20:  
     case 0xa0:  
     case 0x1680:  
     case 0x180e:  
     case 0x2000:  
     case 0x2001:  
     case 0x2002:  
     case 0x2003:  
     case 0x2004:  
     case 0x2005:  
     case 0x2006:  
     case 0x2007:  
     case 0x2008:  
     case 0x2009:  
     case 0x200A:  
     case 0x202f:  
     case 0x205f:  
     case 0x3000:  
     return op_code == OP_NOT_HSPACE;  
     default:  
     return op_code != OP_NOT_HSPACE;  
     }  
4129    
4130    case OP_ANYNL:    for (c = start; c <= end && c < 256; c++)
   case OP_VSPACE:  
   case OP_NOT_VSPACE:  
   switch(next)  
4131      {      {
4132      case 0x0a:      SETBIT(classbits, cd->fcc[c]);
4133      case 0x0b:      n8++;
     case 0x0c:  
     case 0x0d:  
     case 0x85:  
     case 0x2028:  
     case 0x2029:  
     return op_code == OP_NOT_VSPACE;  
     default:  
     return op_code != OP_NOT_VSPACE;  
4134      }      }
   
 #ifdef SUPPORT_UCP  
   case OP_PROP:  
   return check_char_prop(next, previous[0], previous[1], FALSE);  
   
   case OP_NOTPROP:  
   return check_char_prop(next, previous[0], previous[1], TRUE);  
 #endif  
   
   default:  
   return FALSE;  
4135    }    }
4136    
4137    /* Now handle the original range. Adjust the final value according to the bit
4138    length - this means that the same lists of (e.g.) horizontal spaces can be used
4139    in all cases. */
4140    
4141  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP  #if defined COMPILE_PCRE8
 is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are  
 generated only when PCRE_UCP is *not* set, that is, when only ASCII  
 characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are  
 replaced by OP_PROP codes when PCRE_UCP is set. */  
   
 switch(op_code)  
   {  
   case OP_CHAR:  
   case OP_CHARI:  
4142  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4143    GETCHARTEST(c, previous);    if ((options & PCRE_UTF8) == 0)
 #else  
   c = *previous;  
4144  #endif  #endif
4145    switch(-next)    if (end > 0xff) end = 0xff;
     {  
     case ESC_d:  
     return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;  
4146    
4147      case ESC_D:  #elif defined COMPILE_PCRE16
4148      return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;  #ifdef SUPPORT_UTF
4149      if ((options & PCRE_UTF16) == 0)
4150      case ESC_s:  #endif
4151      return c > 255 || (cd->ctypes[c] & ctype_space) == 0;    if (end > 0xffff) end = 0xffff;
4152    
4153      case ESC_S:  #endif /* COMPILE_PCRE[8|16] */
     return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;  
4154    
4155      case ESC_w:  /* If all characters are less than 256, use the bit map. Otherwise use extra
4156      return c > 255 || (cd->ctypes[c] & ctype_word) == 0;  data. */
4157    
4158      case ESC_W:  if (end < 0x100)
4159      return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;    {
4160      for (c = start; c <= end; c++)
4161        {
4162        n8++;
4163        SETBIT(classbits, c);
4164        }
4165      }
4166    
4167      case ESC_h:  else
4168      case ESC_H:    {
4169      switch(c)    pcre_uchar *uchardata = *uchardptr;
       {  
       case 0x09:  
       case 0x20:  
       case 0xa0:  
       case 0x1680:  
       case 0x180e:  
       case 0x2000:  
       case 0x2001:  
       case 0x2002:  
       case 0x2003:  
       case 0x2004:  
       case 0x2005:  
       case 0x2006:  
       case 0x2007:  
       case 0x2008:  
       case 0x2009:  
       case 0x200A:  
       case 0x202f:  
       case 0x205f:  
       case 0x3000:  
       return -next != ESC_h;  
       default:  
       return -next == ESC_h;  
       }  
4170    
4171      case ESC_v:  #ifdef SUPPORT_UTF
4172      case ESC_V:    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
4173      switch(c)      {
4174        if (start < end)
4175        {        {
4176        case 0x0a:        *uchardata++ = XCL_RANGE;
4177        case 0x0b:        uchardata += PRIV(ord2utf)(start, uchardata);
4178        case 0x0c:        uchardata += PRIV(ord2utf)(end, uchardata);
       case 0x0d:  
       case 0x85:  
       case 0x2028:  
       case 0x2029:  
       return -next != ESC_v;  
       default:  
       return -next == ESC_v;  
4179        }        }
4180        else if (start == end)
4181      /* When PCRE_UCP is set, these values get generated for \d etc. Find        {
4182      their substitutions and process them. The result will always be either        *uchardata++ = XCL_SINGLE;
4183      -ESC_p or -ESC_P. Then fall through to process those values. */        uchardata += PRIV(ord2utf)(start, uchardata);
   
 #ifdef SUPPORT_UCP  
     case ESC_du:  
     case ESC_DU:  
     case ESC_wu:  
     case ESC_WU:  
     case ESC_su:  
     case ESC_SU:  
       {  
       int temperrorcode = 0;  
       ptr = substitutes[-next - ESC_DU];  
       next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);  
       if (temperrorcode != 0) return FALSE;  
       ptr++;    /* For compatibility */  
4184        }        }
4185      /* Fall through */      }
4186      else
4187    #endif  /* SUPPORT_UTF */
4188    
4189      case ESC_p:    /* Without UTF support, character values are constrained by the bit length,
4190      case ESC_P:    and can only be > 256 for 16-bit and 32-bit libraries. */
       {  
       int ptype, pdata, errorcodeptr;  
       BOOL negated;  
4191    
4192        ptr--;      /* Make ptr point at the p or P */  #ifdef COMPILE_PCRE8
4193        ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);      {}
4194        if (ptype < 0) return FALSE;  #else
4195        ptr++;      /* Point past the final curly ket */    if (start < end)
4196        {
4197        *uchardata++ = XCL_RANGE;
4198        *uchardata++ = start;
4199        *uchardata++ = end;
4200        }
4201      else if (start == end)
4202        {
4203        *uchardata++ = XCL_SINGLE;
4204        *uchardata++ = start;
4205        }
4206    #endif
4207    
4208        /* If the property item is optional, we have to give up. (When generated    *uchardptr = uchardata;   /* Updata extra data pointer */
4209        from \d etc by PCRE_UCP, this test will have been applied much earlier,    }
       to the original \d etc. At this point, ptr will point to a zero byte. */  
4210    
4211        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  return n8;    /* Number of 8-bit characters */
4212          STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)  }
           return FALSE;  
4213    
       /* Do the property check. */  
4214    
       return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);  
       }  
 #endif  
4215    
     default:  
     return FALSE;  
     }  
4216    
4217    /* In principle, support for Unicode properties should be integrated here as  /*************************************************
4218    well. It means re-organizing the above code so as to get hold of the property  *        Add a list of characters to a class     *
4219    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.) */  
4220    
4221    case OP_DIGIT:  /* This function is used for adding a list of case-equivalent characters to a
4222    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||  class, and also for adding a list of horizontal or vertical whitespace. If the
4223           next == -ESC_h || next == -ESC_v || next == -ESC_R;  list is in order (which it should be), ranges of characters are detected and
4224    handled appropriately. This function is mutually recursive with the function
4225    above.
4226    
4227    case OP_NOT_DIGIT:  Arguments:
4228    return next == -ESC_d;    classbits     the bit map for characters < 256
4229      uchardptr     points to the pointer for extra data
4230      options       the options word
4231      cd            contains pointers to tables etc.
4232      p             points to row of 32-bit values, terminated by NOTACHAR
4233      except        character to omit; this is used when adding lists of
4234                      case-equivalent characters to avoid including the one we
4235                      already know about
4236    
4237    case OP_WHITESPACE:  Returns:        the number of < 256 characters added
4238    return next == -ESC_S || next == -ESC_d || next == -ESC_w;                  the pointer to extra data is updated
4239    */
4240    
4241    case OP_NOT_WHITESPACE:  static int
4242    return next == -ESC_s || next == -ESC_h || next == -ESC_v || next == -ESC_R;  add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4243      compile_data *cd, const pcre_uint32 *p, unsigned int except)
4244    {
4245    int n8 = 0;
4246    while (p[0] < NOTACHAR)
4247      {
4248      int n = 0;
4249      if (p[0] != except)
4250        {
4251        while(p[n+1] == p[0] + n + 1) n++;
4252        n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4253        }
4254      p += n + 1;
4255      }
4256    return n8;
4257    }
4258    
   case OP_HSPACE:  
   return next == -ESC_S || next == -ESC_H || next == -ESC_d ||  
          next == -ESC_w || next == -ESC_v || next == -ESC_R;  
4259    
   case OP_NOT_HSPACE:  
   return next == -ESC_h;  
4260    
4261    /* Can't have \S in here because VT matches \S (Perl anomaly) */  /*************************************************
4262    case OP_ANYNL:  *    Add characters not in a list to a class     *
4263    case OP_VSPACE:  *************************************************/
   return next == -ESC_V || next == -ESC_d || next == -ESC_w;  
4264    
4265    case OP_NOT_VSPACE:  /* This function is used for adding the complement of a list of horizontal or
4266    return next == -ESC_v || next == -ESC_R;  vertical whitespace to a class. The list must be in order.
4267    
4268    case OP_WORDCHAR:  Arguments:
4269    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||    classbits     the bit map for characters < 256
4270           next == -ESC_v || next == -ESC_R;    uchardptr     points to the pointer for extra data
4271      options       the options word
4272      cd            contains pointers to tables etc.
4273      p             points to row of 32-bit values, terminated by NOTACHAR
4274    
4275    case OP_NOT_WORDCHAR:  Returns:        the number of < 256 characters added
4276    return next == -ESC_w || next == -ESC_d;                  the pointer to extra data is updated
4277    */
4278    
4279    default:  static int
4280    return FALSE;  add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
4281      int options, compile_data *cd, const pcre_uint32 *p)
4282    {
4283    BOOL utf = (options & PCRE_UTF8) != 0;
4284    int n8 = 0;
4285    if (p[0] > 0)
4286      n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
4287    while (p[0] < NOTACHAR)
4288      {
4289      while (p[1] == p[0] + 1) p++;
4290      n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
4291        (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
4292      p++;
4293    }    }
4294    return n8;
 /* Control does not reach here */  
4295  }  }
4296    
4297    
# Line 3398  to find out the amount of memory needed, Line 4307  to find out the amount of memory needed,
4307  phase. The value of lengthptr distinguishes the two phases.  phase. The value of lengthptr distinguishes the two phases.
4308    
4309  Arguments:  Arguments:
4310    optionsptr     pointer to the option bits    optionsptr        pointer to the option bits
4311    codeptr        points to the pointer to the current code point    codeptr           points to the pointer to the current code point
4312    ptrptr         points to the current pattern pointer    ptrptr            points to the current pattern pointer
4313    errorcodeptr   points to error code variable    errorcodeptr      points to error code variable
4314    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr      place to put the first required character
4315    reqcharptr     set to the last literal character required, else < 0    firstcharflagsptr place to put the first character flags, or a negative number
4316    bcptr          points to current branch chain    reqcharptr        place to put the last required character
4317    cond_depth     conditional nesting depth    reqcharflagsptr   place to put the last required character flags, or a negative number
4318    cd             contains pointers to tables etc.    bcptr             points to current branch chain
4319    lengthptr      NULL during the real compile phase    cond_depth        conditional nesting depth
4320                   points to length accumulator during pre-compile phase    cd                contains pointers to tables etc.
4321      lengthptr         NULL during the real compile phase
4322                        points to length accumulator during pre-compile phase
4323    
4324  Returns:         TRUE on success  Returns:            TRUE on success
4325                   FALSE, with *errorcodeptr set non-zero on error                      FALSE, with *errorcodeptr set non-zero on error
4326  */  */
4327    
4328  static BOOL  static BOOL
4329  compile_branch(int *optionsptr, pcre_uchar **codeptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
4330    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,    const pcre_uchar **ptrptr, int *errorcodeptr,
4331    pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,    pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
4332      pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
4333      branch_chain *bcptr, int cond_depth,
4334    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
4335  {  {
4336  int repeat_type, op_type;  int repeat_type, op_type;
4337  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
4338  int bravalue = 0;  int bravalue = 0;
4339  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
4340  pcre_int32 firstchar, reqchar;  pcre_uint32 firstchar, reqchar;
4341  pcre_int32 zeroreqchar, zerofirstchar;  pcre_int32 firstcharflags, reqcharflags;
4342    pcre_uint32 zeroreqchar, zerofirstchar;
4343    pcre_int32 zeroreqcharflags, zerofirstcharflags;
4344  pcre_int32 req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
4345  int options = *optionsptr;               /* May change dynamically */  int options = *optionsptr;               /* May change dynamically */
4346  int after_manual_callout = 0;  int after_manual_callout = 0;
4347  int length_prevgroup = 0;  int length_prevgroup = 0;
4348  register int c;  register pcre_uint32 c;
4349    int escape;
4350  register pcre_uchar *code = *codeptr;  register pcre_uchar *code = *codeptr;
4351  pcre_uchar *last_code = code;  pcre_uchar *last_code = code;
4352  pcre_uchar *orig_code = code;  pcre_uchar *orig_code = code;
# Line 3450  must not do this for other options (e.g. Line 4366  must not do this for other options (e.g.
4366  dynamically as we process the pattern. */  dynamically as we process the pattern. */
4367    
4368  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4369  /* PCRE_UTF16 has the same value as PCRE_UTF8. */  /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
4370  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
4371    #ifndef COMPILE_PCRE32
4372  pcre_uchar utf_chars[6];  pcre_uchar utf_chars[6];
4373    #endif
4374  #else  #else
4375  BOOL utf = FALSE;  BOOL utf = FALSE;
4376  #endif  #endif
4377    
4378  /* Helper variables for OP_XCLASS opcode (for characters > 255). */  /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
4379    class_uchardata always so that it can be passed to add_to_class() always,
4380    though it will not be used in non-UTF 8-bit cases. This avoids having to supply
4381    alternative calls for the different cases. */
4382    
4383    pcre_uchar *class_uchardata;
4384  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4385  BOOL xclass;  BOOL xclass;
 pcre_uchar *class_uchardata;  
4386  pcre_uchar *class_uchardata_base;  pcre_uchar *class_uchardata_base;
4387  #endif  #endif
4388    
# Line 3484  to take the zero repeat into account. Th Line 4405  to take the zero repeat into account. Th
4405  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
4406  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
4407    
4408  firstchar = reqchar = zerofirstchar = zeroreqchar = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
4409    firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
4410    
4411  /* The variable req_caseopt contains either the REQ_CASELESS value  /* The variable req_caseopt contains either the REQ_CASELESS value
4412  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 3505  for (;; ptr++) Line 4427  for (;; ptr++)
4427    BOOL is_recurse;    BOOL is_recurse;
4428    BOOL reset_bracount;    BOOL reset_bracount;
4429    int class_has_8bitchar;    int class_has_8bitchar;
4430    int class_single_char;    int class_one_char;
4431    int newoptions;    int newoptions;
4432    int recno;    int recno;
4433    int refsign;    int refsign;
4434    int skipbytes;    int skipbytes;
4435    int subreqchar;    pcre_uint32 subreqchar, subfirstchar;
4436    int subfirstchar;    pcre_int32 subreqcharflags, subfirstcharflags;
4437    int terminator;    int terminator;
4438    int mclength;    unsigned int mclength;
4439    int tempbracount;    unsigned int tempbracount;
4440      pcre_uint32 ec;